openai/codex
1
0
Fork 0
mirror of https://github.com/openai/codex.git synced 2026-10-31 19:28:49 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'main' into jif/codex-exec-review

Browse Source
This commit is contained in:
jif-oai
2025-11-18 14:01:06 +00:00
committed by GitHub
parent 8970cccbb1 f3d4e210d8
commit d619a0dc05
211 changed files with 13176 additions and 29831 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.github/codex/home/config.toml
+1 -1
View File
@@ -1,3 +1,3 @@
model = "gpt-5"
model = "gpt-5.1"
# Consider setting [mcp_servers] here!
.github/workflows/cla.yml
+6 -1
View File
@@ -13,6 +13,9 @@ permissions:
jobs:
cla:
# Only run the CLA assistant for the canonical openai repo so forks are not blocked
# and contributors who signed previously do not receive duplicate CLA notifications.
if: ${{ github.repository_owner == 'openai' }}
runs-on: ubuntu-latest
steps:
- uses: contributor-assistant/github-action@v2.6.1
@@ -43,4 +46,6 @@ jobs:
path-to-document: https://github.com/openai/codex/blob/main/docs/CLA.md
path-to-signatures: signatures/cla.json
branch: cla-signatures
allowlist: dependabot[bot]
allowlist: |
codex
dependabot[bot]
.github/workflows/close-stale-contributor-prs.yml
+105
View File
@@ -0,0 +1,105 @@
name: Close stale contributor PRs
on:
workflow_dispatch:
schedule:
- cron: "0 6 * * *"
permissions:
contents: read
issues: write
pull-requests: write
jobs:
close-stale-contributor-prs:
runs-on: ubuntu-latest
steps:
- name: Close inactive PRs from contributors
uses: actions/github-script@v8
with:
github-token: ${{ secrets.GITHUB_TOKEN }}
script: |
const DAYS_INACTIVE = 14;
const cutoff = new Date(Date.now() - DAYS_INACTIVE * 24 * 60 * 60 * 1000);
const { owner, repo } = context.repo;
const dryRun = false;
const stalePrs = [];
core.info(`Dry run mode: ${dryRun}`);
const prs = await github.paginate(github.rest.pulls.list, {
owner,
repo,
state: "open",
per_page: 100,
sort: "updated",
direction: "asc",
});
for (const pr of prs) {
const lastUpdated = new Date(pr.updated_at);
if (lastUpdated > cutoff) {
core.info(`PR ${pr.number} is fresh`);
continue;
}
if (!pr.user || pr.user.type !== "User") {
core.info(`PR ${pr.number} wasn't created by a user`);
continue;
}
let permission;
try {
const permissionResponse = await github.rest.repos.getCollaboratorPermissionLevel({
owner,
repo,
username: pr.user.login,
});
permission = permissionResponse.data.permission;
} catch (error) {
if (error.status === 404) {
core.info(`Author ${pr.user.login} is not a collaborator; skipping #${pr.number}`);
continue;
}
throw error;
}
const hasContributorAccess = ["admin", "maintain", "write"].includes(permission);
if (!hasContributorAccess) {
core.info(`Author ${pr.user.login} has ${permission} access; skipping #${pr.number}`);
continue;
}
stalePrs.push(pr);
}
if (!stalePrs.length) {
core.info("No stale contributor pull requests found.");
return;
}
for (const pr of stalePrs) {
const issue_number = pr.number;
const closeComment = `Closing this pull request because it has had no updates for more than ${DAYS_INACTIVE} days. If you plan to continue working on it, feel free to reopen or open a new PR.`;
if (dryRun) {
core.info(`[dry-run] Would close contributor PR #${issue_number} from ${pr.user.login}`);
continue;
}
await github.rest.issues.createComment({
owner,
repo,
issue_number,
body: closeComment,
});
await github.rest.pulls.update({
owner,
repo,
pull_number: issue_number,
state: "closed",
});
core.info(`Closed contributor PR #${issue_number} from ${pr.user.login}`);
}
.github/workflows/issue-deduplicator.yml
+1 -1
View File
@@ -46,7 +46,7 @@ jobs:
with:
openai-api-key: ${{ secrets.CODEX_OPENAI_API_KEY }}
allow-users: "*"
model: gpt-5
model: gpt-5.1
prompt: |
You are an assistant that triages new GitHub issues by identifying potential duplicates.
.github/workflows/rust-ci.yml
+27 -12
View File
@@ -95,8 +95,8 @@ jobs:
run:
working-directory: codex-rs
env:
# Speed up repeated builds across CI runs by caching compiled objects.
RUSTC_WRAPPER: sccache
# Speed up repeated builds across CI runs by caching compiled objects (non-Windows).
USE_SCCACHE: ${{ startsWith(matrix.runner, 'windows') && 'false' || 'true' }}
CARGO_INCREMENTAL: "0"
SCCACHE_CACHE_SIZE: 10G
@@ -170,12 +170,14 @@ jobs:
# Install and restore sccache cache
- name: Install sccache
if: ${{ env.USE_SCCACHE == 'true' }}
uses: taiki-e/install-action@44c6d64aa62cd779e873306675c7a58e86d6d532 # v2
with:
tool: sccache
version: 0.7.5
- name: Configure sccache backend
if: ${{ env.USE_SCCACHE == 'true' }}
shell: bash
run: |
set -euo pipefail
@@ -188,8 +190,13 @@ jobs:
echo "Using sccache local disk + actions/cache fallback"
fi
- name: Enable sccache wrapper
if: ${{ env.USE_SCCACHE == 'true' }}
shell: bash
run: echo "RUSTC_WRAPPER=sccache" >> "$GITHUB_ENV"
- name: Restore sccache cache (fallback)
if: ${{ env.SCCACHE_GHA_ENABLED != 'true' }}
if: ${{ env.USE_SCCACHE == 'true' && env.SCCACHE_GHA_ENABLED != 'true' }}
id: cache_sccache_restore
uses: actions/cache/restore@v4
with:
@@ -274,7 +281,7 @@ jobs:
key: cargo-home-${{ matrix.runner }}-${{ matrix.target }}-${{ matrix.profile }}-${{ hashFiles('**/Cargo.lock') }}-${{ hashFiles('codex-rs/rust-toolchain.toml') }}
- name: Save sccache cache (fallback)
if: always() && !cancelled() && env.SCCACHE_GHA_ENABLED != 'true'
if: always() && !cancelled() && env.USE_SCCACHE == 'true' && env.SCCACHE_GHA_ENABLED != 'true'
continue-on-error: true
uses: actions/cache/save@v4
with:
@@ -282,12 +289,12 @@ jobs:
key: sccache-${{ matrix.runner }}-${{ matrix.target }}-${{ matrix.profile }}-${{ hashFiles('**/Cargo.lock') }}-${{ github.run_id }}
- name: sccache stats
if: always()
if: always() && env.USE_SCCACHE == 'true'
continue-on-error: true
run: sccache --show-stats || true
- name: sccache summary
if: always()
if: always() && env.USE_SCCACHE == 'true'
shell: bash
run: |
{
@@ -326,7 +333,8 @@ jobs:
run:
working-directory: codex-rs
env:
RUSTC_WRAPPER: sccache
# Speed up repeated builds across CI runs by caching compiled objects (non-Windows).
USE_SCCACHE: ${{ startsWith(matrix.runner, 'windows') && 'false' || 'true' }}
CARGO_INCREMENTAL: "0"
SCCACHE_CACHE_SIZE: 10G
@@ -370,12 +378,14 @@ jobs:
cargo-home-${{ matrix.runner }}-${{ matrix.target }}-${{ matrix.profile }}-
- name: Install sccache
if: ${{ env.USE_SCCACHE == 'true' }}
uses: taiki-e/install-action@44c6d64aa62cd779e873306675c7a58e86d6d532 # v2
with:
tool: sccache
version: 0.7.5
- name: Configure sccache backend
if: ${{ env.USE_SCCACHE == 'true' }}
shell: bash
run: |
set -euo pipefail
@@ -388,8 +398,13 @@ jobs:
echo "Using sccache local disk + actions/cache fallback"
fi
- name: Enable sccache wrapper
if: ${{ env.USE_SCCACHE == 'true' }}
shell: bash
run: echo "RUSTC_WRAPPER=sccache" >> "$GITHUB_ENV"
- name: Restore sccache cache (fallback)
if: ${{ env.SCCACHE_GHA_ENABLED != 'true' }}
if: ${{ env.USE_SCCACHE == 'true' && env.SCCACHE_GHA_ENABLED != 'true' }}
id: cache_sccache_restore
uses: actions/cache/restore@v4
with:
@@ -406,10 +421,10 @@ jobs:
- name: tests
id: test
continue-on-error: true
run: cargo nextest run --all-features --no-fail-fast --target ${{ matrix.target }} --cargo-profile ci-test
env:
RUST_BACKTRACE: 1
NEXTEST_STATUS_LEVEL: leak
- name: Save cargo home cache
if: always() && !cancelled() && steps.cache_cargo_home_restore.outputs.cache-hit != 'true'
@@ -424,7 +439,7 @@ jobs:
key: cargo-home-${{ matrix.runner }}-${{ matrix.target }}-${{ matrix.profile }}-${{ hashFiles('**/Cargo.lock') }}-${{ hashFiles('codex-rs/rust-toolchain.toml') }}
- name: Save sccache cache (fallback)
if: always() && !cancelled() && env.SCCACHE_GHA_ENABLED != 'true'
if: always() && !cancelled() && env.USE_SCCACHE == 'true' && env.SCCACHE_GHA_ENABLED != 'true'
continue-on-error: true
uses: actions/cache/save@v4
with:
@@ -432,12 +447,12 @@ jobs:
key: sccache-${{ matrix.runner }}-${{ matrix.target }}-${{ matrix.profile }}-${{ hashFiles('**/Cargo.lock') }}-${{ github.run_id }}
- name: sccache stats
if: always()
if: always() && env.USE_SCCACHE == 'true'
continue-on-error: true
run: sccache --show-stats || true
- name: sccache summary
if: always()
if: always() && env.USE_SCCACHE == 'true'
shell: bash
run: |
{
.gitignore
+3
View File
@@ -64,6 +64,9 @@ apply_patch/
# coverage
coverage/
# personal files
personal/
# os
.DS_Store
Thumbs.db
AGENTS.md
+1
View File
@@ -85,6 +85,7 @@ If you don’t have the tool:
- `ResponsesRequest` exposes helpers (`body_json`, `input`, `function_call_output`, `custom_tool_call_output`, `call_output`, `header`, `path`, `query_param`) so assertions can target structured payloads instead of manual JSON digging.
- Build SSE payloads with the provided `ev_*` constructors and the `sse(...)`.
- Prefer `wait_for_event` over `wait_for_event_with_timeout`.
- Prefer `mount_sse_once` over `mount_sse_once_match` or `mount_sse_sequence`
- Typical pattern:
codex-rs/.config/nextest.toml
+9
View File
@@ -0,0 +1,9 @@
[profile.default]
# Do not increase, fix your test instead
slow-timeout = { period = "15s", terminate-after = 2 }
[[profile.default.overrides]]
# Do not add new tests here
filter = 'test(rmcp_client) | test(humanlike_typing_1000_chars_appears_live_no_placeholder)'
slow-timeout = { period = "1m", terminate-after = 4 }
codex-rs/Cargo.lock
Generated
+85 -5
View File
@@ -850,6 +850,7 @@ dependencies = [
"codex-protocol",
"codex-utils-json-to-toml",
"core_test_support",
"mcp-types",
"opentelemetry-appender-tracing",
"os_info",
"pretty_assertions",
@@ -873,7 +874,6 @@ dependencies = [
"clap",
"codex-protocol",
"mcp-types",
"paste",
"pretty_assertions",
"schemars 0.8.22",
"serde",
@@ -883,6 +883,19 @@ dependencies = [
"uuid",
]
[[package]]
name = "codex-app-server-test-client"
version = "0.0.0"
dependencies = [
"anyhow",
"clap",
"codex-app-server-protocol",
"codex-protocol",
"serde",
"serde_json",
"uuid",
]
[[package]]
name = "codex-apply-patch"
version = "0.0.0"
@@ -1047,6 +1060,8 @@ dependencies = [
"clap",
"codex-app-server-protocol",
"codex-core",
"codex-lmstudio",
"codex-ollama",
"codex-protocol",
"once_cell",
"serde",
@@ -1115,6 +1130,7 @@ dependencies = [
"similar",
"strum_macros 0.27.2",
"tempfile",
"test-case",
"test-log",
"thiserror 2.0.17",
"time",
@@ -1144,7 +1160,6 @@ dependencies = [
"codex-arg0",
"codex-common",
"codex-core",
"codex-ollama",
"codex-protocol",
"core_test_support",
"libc",
@@ -1187,6 +1202,21 @@ dependencies = [
"tempfile",
]
[[package]]
name = "codex-execpolicy2"
version = "0.0.0"
dependencies = [
"anyhow",
"clap",
"multimap",
"pretty_assertions",
"serde",
"serde_json",
"shlex",
"starlark",
"thiserror 2.0.17",
]
[[package]]
name = "codex-feedback"
version = "0.0.0"
@@ -1248,6 +1278,19 @@ dependencies = [
"tokio",
]
[[package]]
name = "codex-lmstudio"
version = "0.0.0"
dependencies = [
"codex-core",
"reqwest",
"serde_json",
"tokio",
"tracing",
"which",
"wiremock",
]
[[package]]
name = "codex-login"
version = "0.0.0"
@@ -1412,6 +1455,7 @@ dependencies = [
"tracing",
"urlencoding",
"webbrowser",
"which",
]
[[package]]
@@ -1439,12 +1483,12 @@ dependencies = [
"codex-ansi-escape",
"codex-app-server-protocol",
"codex-arg0",
"codex-backend-client",
"codex-common",
"codex-core",
"codex-feedback",
"codex-file-search",
"codex-login",
"codex-ollama",
"codex-protocol",
"codex-windows-sandbox",
"color-eyre",
@@ -1467,6 +1511,7 @@ dependencies = [
"ratatui",
"ratatui-macros",
"regex-lite",
"reqwest",
"serde",
"serde_json",
"serial_test",
@@ -1549,9 +1594,11 @@ name = "codex-utils-tokenizer"
version = "0.0.0"
dependencies = [
"anyhow",
"codex-utils-cache",
"pretty_assertions",
"thiserror 2.0.17",
"tiktoken-rs",
"tokio",
]
[[package]]
@@ -6161,6 +6208,39 @@ version = "0.5.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "8f50febec83f5ee1df3015341d8bd429f2d1cc62bcba7ea2076759d315084683"
[[package]]
name = "test-case"
version = "3.3.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "eb2550dd13afcd286853192af8601920d959b14c401fcece38071d53bf0768a8"
dependencies = [
"test-case-macros",
]
[[package]]
name = "test-case-core"
version = "3.3.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "adcb7fd841cd518e279be3d5a3eb0636409487998a4aff22f3de87b81e88384f"
dependencies = [
"cfg-if",
"proc-macro2",
"quote",
"syn 2.0.104",
]
[[package]]
name = "test-case-macros"
version = "3.3.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "5c89e72a01ed4c579669add59014b9a524d609c0c88c6a585ce37485879f6ffb"
dependencies = [
"proc-macro2",
"quote",
"syn 2.0.104",
"test-case-core",
]
[[package]]
name = "test-log"
version = "0.2.18"
@@ -6268,9 +6348,9 @@ dependencies = [
[[package]]
name = "tiktoken-rs"
version = "0.7.0"
version = "0.9.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "25563eeba904d770acf527e8b370fe9a5547bacd20ff84a0b6c3bc41288e5625"
checksum = "3a19830747d9034cd9da43a60eaa8e552dfda7712424aebf187b7a60126bae0d"
dependencies = [
"anyhow",
"base64",
codex-rs/Cargo.toml
+6 -2
View File
@@ -5,6 +5,7 @@ members = [
"async-utils",
"app-server",
"app-server-protocol",
"app-server-test-client",
"apply-patch",
"arg0",
"feedback",
@@ -16,9 +17,11 @@ members = [
"core",
"exec",
"execpolicy",
"execpolicy2",
"keyring-store",
"file-search",
"linux-sandbox",
"lmstudio",
"login",
"mcp-server",
"mcp-types",
@@ -68,6 +71,7 @@ codex-file-search = { path = "file-search" }
codex-git = { path = "utils/git" }
codex-keyring-store = { path = "keyring-store" }
codex-linux-sandbox = { path = "linux-sandbox" }
codex-lmstudio = { path = "lmstudio" }
codex-login = { path = "login" }
codex-mcp-server = { path = "mcp-server" }
codex-ollama = { path = "ollama" }
@@ -129,7 +133,7 @@ image = { version = "^0.25.8", default-features = false }
indexmap = "2.12.0"
insta = "1.43.2"
itertools = "0.14.0"
keyring = "3.6"
keyring = { version = "3.6", default-features = false }
landlock = "0.4.1"
lazy_static = "1"
libc = "0.2.175"
@@ -149,7 +153,6 @@ opentelemetry-semantic-conventions = "0.30.0"
opentelemetry_sdk = "0.30.0"
os_info = "3.12.0"
owo-colors = "4.2.0"
paste = "1.0.15"
path-absolutize = "3.1.1"
pathdiff = "0.2"
portable-pty = "0.9.0"
@@ -182,6 +185,7 @@ tempfile = "3.23.0"
test-log = "0.2.18"
textwrap = "0.16.2"
thiserror = "2.0.17"
tiktoken-rs = "0.9"
time = "0.3"
tiny_http = "0.12"
tokio = "1"
codex-rs/app-server-protocol/Cargo.toml
-1
View File
@@ -15,7 +15,6 @@ anyhow = { workspace = true }
clap = { workspace = true, features = ["derive"] }
codex-protocol = { workspace = true }
mcp-types = { workspace = true }
paste = { workspace = true }
schemars = { workspace = true }
serde = { workspace = true, features = ["derive"] }
serde_json = { workspace = true }
codex-rs/app-server-protocol/src/export.rs
+8 -19
View File
@@ -13,10 +13,7 @@ use crate::export_server_responses;
use anyhow::Context;
use anyhow::Result;
use anyhow::anyhow;
use codex_protocol::parse_command::ParsedCommand;
use codex_protocol::protocol::EventMsg;
use codex_protocol::protocol::FileChange;
use codex_protocol::protocol::SandboxPolicy;
use schemars::JsonSchema;
use schemars::schema_for;
use serde::Serialize;
@@ -120,10 +117,6 @@ pub fn generate_json(out_dir: &Path) -> Result<()> {
|d| write_json_schema_with_return::<crate::ClientNotification>(d, "ClientNotification"),
|d| write_json_schema_with_return::<crate::ServerNotification>(d, "ServerNotification"),
|d| write_json_schema_with_return::<EventMsg>(d, "EventMsg"),
|d| write_json_schema_with_return::<FileChange>(d, "FileChange"),
|d| write_json_schema_with_return::<crate::protocol::v1::InputItem>(d, "InputItem"),
|d| write_json_schema_with_return::<ParsedCommand>(d, "ParsedCommand"),
|d| write_json_schema_with_return::<SandboxPolicy>(d, "SandboxPolicy"),
];
let mut schemas: Vec<GeneratedSchema> = Vec::new();
@@ -152,13 +145,10 @@ fn build_schema_bundle(schemas: Vec<GeneratedSchema>) -> Result<Value> {
"ClientNotification",
"ClientRequest",
"EventMsg",
"FileChange",
"InputItem",
"ParsedCommand",
"SandboxPolicy",
"ServerNotification",
"ServerRequest",
];
const IGNORED_DEFINITIONS: &[&str] = &["Option<()>"];
let namespaced_types = collect_namespaced_types(&schemas);
let mut definitions = Map::new();
@@ -171,6 +161,10 @@ fn build_schema_bundle(schemas: Vec<GeneratedSchema>) -> Result<Value> {
in_v1_dir,
} = schema;
if IGNORED_DEFINITIONS.contains(&logical_name.as_str()) {
continue;
}
if let Some(ref ns) = namespace {
rewrite_refs_to_namespace(&mut value, ns);
}
@@ -181,6 +175,9 @@ fn build_schema_bundle(schemas: Vec<GeneratedSchema>) -> Result<Value> {
&& let Value::Object(defs_obj) = defs
{
for (def_name, mut def_schema) in defs_obj {
if IGNORED_DEFINITIONS.contains(&def_name.as_str()) {
continue;
}
if SPECIAL_DEFINITIONS.contains(&def_name.as_str()) {
continue;
}
@@ -386,14 +383,6 @@ fn variant_definition_name(base: &str, variant: &Value) -> Option<String> {
});
}
if let Some(mode_literal) = literal_from_property(props, "mode") {
let pascal = to_pascal_case(mode_literal);
return Some(match base {
"SandboxPolicy" => format!("{pascal}SandboxPolicy"),
_ => format!("{pascal}{base}"),
});
}
if props.len() == 1
&& let Some(key) = props.keys().next()
{
codex-rs/app-server-protocol/src/protocol/common.rs
+69 -85
View File
@@ -1,6 +1,4 @@
use std::collections::HashMap;
use std::path::Path;
use std::path::PathBuf;
use crate::JSONRPCNotification;
use crate::JSONRPCRequest;
@@ -9,12 +7,6 @@ use crate::export::GeneratedSchema;
use crate::export::write_json_schema;
use crate::protocol::v1;
use crate::protocol::v2;
use codex_protocol::ConversationId;
use codex_protocol::parse_command::ParsedCommand;
use codex_protocol::protocol::FileChange;
use codex_protocol::protocol::ReviewDecision;
use codex_protocol::protocol::SandboxCommandAssessment;
use paste::paste;
use schemars::JsonSchema;
use serde::Deserialize;
use serde::Serialize;
@@ -277,34 +269,36 @@ macro_rules! server_request_definitions {
(
$(
$(#[$variant_meta:meta])*
$variant:ident
$variant:ident $(=> $wire:literal)? {
params: $params:ty,
response: $response:ty,
}
),* $(,)?
) => {
paste! {
/// Request initiated from the server and sent to the client.
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(tag = "method", rename_all = "camelCase")]
pub enum ServerRequest {
$(
$(#[$variant_meta])*
$variant {
#[serde(rename = "id")]
request_id: RequestId,
params: [<$variant Params>],
},
)*
}
/// Request initiated from the server and sent to the client.
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(tag = "method", rename_all = "camelCase")]
pub enum ServerRequest {
$(
$(#[$variant_meta])*
$(#[serde(rename = $wire)] #[ts(rename = $wire)])?
$variant {
#[serde(rename = "id")]
request_id: RequestId,
params: $params,
},
)*
}
#[derive(Debug, Clone, PartialEq, JsonSchema)]
pub enum ServerRequestPayload {
$( $variant([<$variant Params>]), )*
}
#[derive(Debug, Clone, PartialEq, JsonSchema)]
pub enum ServerRequestPayload {
$( $variant($params), )*
}
impl ServerRequestPayload {
pub fn request_with_id(self, request_id: RequestId) -> ServerRequest {
match self {
$(Self::$variant(params) => ServerRequest::$variant { request_id, params },)*
}
impl ServerRequestPayload {
pub fn request_with_id(self, request_id: RequestId) -> ServerRequest {
match self {
$(Self::$variant(params) => ServerRequest::$variant { request_id, params },)*
}
}
}
@@ -312,9 +306,9 @@ macro_rules! server_request_definitions {
pub fn export_server_responses(
out_dir: &::std::path::Path,
) -> ::std::result::Result<(), ::ts_rs::ExportError> {
paste! {
$(<[<$variant Response>] as ::ts_rs::TS>::export_all_to(out_dir)?;)*
}
$(
<$response as ::ts_rs::TS>::export_all_to(out_dir)?;
)*
Ok(())
}
@@ -323,9 +317,12 @@ macro_rules! server_request_definitions {
out_dir: &Path,
) -> ::anyhow::Result<Vec<GeneratedSchema>> {
let mut schemas = Vec::new();
paste! {
$(schemas.push(crate::export::write_json_schema::<[<$variant Response>]>(out_dir, stringify!([<$variant Response>]))?);)*
}
$(
schemas.push(crate::export::write_json_schema::<$response>(
out_dir,
concat!(stringify!($variant), "Response"),
)?);
)*
Ok(schemas)
}
@@ -334,9 +331,12 @@ macro_rules! server_request_definitions {
out_dir: &Path,
) -> ::anyhow::Result<Vec<GeneratedSchema>> {
let mut schemas = Vec::new();
paste! {
$(schemas.push(crate::export::write_json_schema::<[<$variant Params>]>(out_dir, stringify!([<$variant Params>]))?);)*
}
$(
schemas.push(crate::export::write_json_schema::<$params>(
out_dir,
concat!(stringify!($variant), "Params"),
)?);
)*
Ok(schemas)
}
};
@@ -426,49 +426,27 @@ impl TryFrom<JSONRPCRequest> for ServerRequest {
}
server_request_definitions! {
/// NEW APIs
/// Sent when approval is requested for a specific command execution.
/// This request is used for Turns started via turn/start.
CommandExecutionRequestApproval => "item/commandExecution/requestApproval" {
params: v2::CommandExecutionRequestApprovalParams,
response: v2::CommandExecutionRequestApprovalResponse,
},
/// DEPRECATED APIs below
/// Request to approve a patch.
ApplyPatchApproval,
/// This request is used for Turns started via the legacy APIs (i.e. SendUserTurn, SendUserMessage).
ApplyPatchApproval {
params: v1::ApplyPatchApprovalParams,
response: v1::ApplyPatchApprovalResponse,
},
/// Request to exec a command.
ExecCommandApproval,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
pub struct ApplyPatchApprovalParams {
pub conversation_id: ConversationId,
/// Use to correlate this with [codex_core::protocol::PatchApplyBeginEvent]
/// and [codex_core::protocol::PatchApplyEndEvent].
pub call_id: String,
pub file_changes: HashMap<PathBuf, FileChange>,
/// Optional explanatory reason (e.g. request for extra write access).
pub reason: Option<String>,
/// When set, the agent is asking the user to allow writes under this root
/// for the remainder of the session (unclear if this is honored today).
pub grant_root: Option<PathBuf>,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
pub struct ExecCommandApprovalParams {
pub conversation_id: ConversationId,
/// Use to correlate this with [codex_core::protocol::ExecCommandBeginEvent]
/// and [codex_core::protocol::ExecCommandEndEvent].
pub call_id: String,
pub command: Vec<String>,
pub cwd: PathBuf,
pub reason: Option<String>,
pub risk: Option<SandboxCommandAssessment>,
pub parsed_cmd: Vec<ParsedCommand>,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
pub struct ExecCommandApprovalResponse {
pub decision: ReviewDecision,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
pub struct ApplyPatchApprovalResponse {
pub decision: ReviewDecision,
/// This request is used for Turns started via the legacy APIs (i.e. SendUserTurn, SendUserMessage).
ExecCommandApproval {
params: v1::ExecCommandApprovalParams,
response: v1::ExecCommandApprovalResponse,
},
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
@@ -508,6 +486,9 @@ server_notification_definitions! {
McpToolCallProgress => "item/mcpToolCall/progress" (v2::McpToolCallProgressNotification),
AccountUpdated => "account/updated" (v2::AccountUpdatedNotification),
AccountRateLimitsUpdated => "account/rateLimits/updated" (v2::AccountRateLimitsUpdatedNotification),
ReasoningSummaryTextDelta => "item/reasoning/summaryTextDelta" (v2::ReasoningSummaryTextDeltaNotification),
ReasoningSummaryPartAdded => "item/reasoning/summaryPartAdded" (v2::ReasoningSummaryPartAddedNotification),
ReasoningTextDelta => "item/reasoning/textDelta" (v2::ReasoningTextDeltaNotification),
#[serde(rename = "account/login/completed")]
#[ts(rename = "account/login/completed")]
@@ -530,17 +511,20 @@ client_notification_definitions! {
mod tests {
use super::*;
use anyhow::Result;
use codex_protocol::ConversationId;
use codex_protocol::account::PlanType;
use codex_protocol::parse_command::ParsedCommand;
use codex_protocol::protocol::AskForApproval;
use pretty_assertions::assert_eq;
use serde_json::json;
use std::path::PathBuf;
#[test]
fn serialize_new_conversation() -> Result<()> {
let request = ClientRequest::NewConversation {
request_id: RequestId::Integer(42),
params: v1::NewConversationParams {
model: Some("gpt-5-codex".to_string()),
model: Some("gpt-5.1-codex".to_string()),
model_provider: None,
profile: None,
cwd: None,
@@ -558,7 +542,7 @@ mod tests {
"method": "newConversation",
"id": 42,
"params": {
"model": "gpt-5-codex",
"model": "gpt-5.1-codex",
"modelProvider": null,
"profile": null,
"cwd": null,
@@ -613,7 +597,7 @@ mod tests {
#[test]
fn serialize_server_request() -> Result<()> {
let conversation_id = ConversationId::from_string("67e55044-10b1-426f-9247-bb680e5fe0c8")?;
let params = ExecCommandApprovalParams {
let params = v1::ExecCommandApprovalParams {
conversation_id,
call_id: "call-42".to_string(),
command: vec!["echo".to_string(), "hello".to_string()],
codex-rs/app-server-protocol/src/protocol/v1.rs
+44
View File
@@ -8,8 +8,12 @@ use codex_protocol::config_types::ReasoningSummary;
use codex_protocol::config_types::SandboxMode;
use codex_protocol::config_types::Verbosity;
use codex_protocol::models::ResponseItem;
use codex_protocol::parse_command::ParsedCommand;
use codex_protocol::protocol::AskForApproval;
use codex_protocol::protocol::EventMsg;
use codex_protocol::protocol::FileChange;
use codex_protocol::protocol::ReviewDecision;
use codex_protocol::protocol::SandboxCommandAssessment;
use codex_protocol::protocol::SandboxPolicy;
use codex_protocol::protocol::SessionSource;
use codex_protocol::protocol::TurnAbortReason;
@@ -191,6 +195,46 @@ pub struct GitDiffToRemoteResponse {
pub diff: String,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
pub struct ApplyPatchApprovalParams {
pub conversation_id: ConversationId,
/// Use to correlate this with [codex_core::protocol::PatchApplyBeginEvent]
/// and [codex_core::protocol::PatchApplyEndEvent].
pub call_id: String,
pub file_changes: HashMap<PathBuf, FileChange>,
/// Optional explanatory reason (e.g. request for extra write access).
pub reason: Option<String>,
/// When set, the agent is asking the user to allow writes under this root
/// for the remainder of the session (unclear if this is honored today).
pub grant_root: Option<PathBuf>,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
pub struct ApplyPatchApprovalResponse {
pub decision: ReviewDecision,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
pub struct ExecCommandApprovalParams {
pub conversation_id: ConversationId,
/// Use to correlate this with [codex_core::protocol::ExecCommandBeginEvent]
/// and [codex_core::protocol::ExecCommandEndEvent].
pub call_id: String,
pub command: Vec<String>,
pub cwd: PathBuf,
pub reason: Option<String>,
pub risk: Option<SandboxCommandAssessment>,
pub parsed_cmd: Vec<ParsedCommand>,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
pub struct ExecCommandApprovalResponse {
pub decision: ReviewDecision,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
pub struct CancelLoginChatGptParams {
codex-rs/app-server-protocol/src/protocol/v2.rs
+226 -44
View File
@@ -4,11 +4,13 @@ use std::path::PathBuf;
use crate::protocol::common::AuthMode;
use codex_protocol::ConversationId;
use codex_protocol::account::PlanType;
use codex_protocol::approvals::SandboxCommandAssessment as CoreSandboxCommandAssessment;
use codex_protocol::config_types::ReasoningEffort;
use codex_protocol::config_types::ReasoningSummary;
use codex_protocol::items::AgentMessageContent as CoreAgentMessageContent;
use codex_protocol::items::TurnItem as CoreTurnItem;
use codex_protocol::models::ResponseItem;
use codex_protocol::parse_command::ParsedCommand as CoreParsedCommand;
use codex_protocol::protocol::RateLimitSnapshot as CoreRateLimitSnapshot;
use codex_protocol::protocol::RateLimitWindow as CoreRateLimitWindow;
use codex_protocol::user_input::UserInput as CoreUserInput;
@@ -20,7 +22,7 @@ use serde_json::Value as JsonValue;
use ts_rs::TS;
// Macro to declare a camelCased API v2 enum mirroring a core enum which
// tends to use kebab-case.
// tends to use either snake_case or kebab-case.
macro_rules! v2_enum_from_core {
(
pub enum $Name:ident from $Src:path { $( $Variant:ident ),+ $(,)? }
@@ -56,13 +58,32 @@ v2_enum_from_core!(
}
);
v2_enum_from_core!(
pub enum CommandRiskLevel from codex_protocol::approvals::SandboxRiskLevel {
Low,
Medium,
High
}
);
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Eq, JsonSchema, TS)]
#[serde(tag = "mode", rename_all = "camelCase")]
#[ts(tag = "mode")]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub enum ApprovalDecision {
Accept,
Decline,
Cancel,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Eq, JsonSchema, TS)]
#[serde(tag = "type", rename_all = "camelCase")]
#[ts(tag = "type")]
#[ts(export_to = "v2/")]
pub enum SandboxPolicy {
DangerFullAccess,
ReadOnly,
#[serde(rename_all = "camelCase")]
#[ts(rename_all = "camelCase")]
WorkspaceWrite {
#[serde(default)]
writable_roots: Vec<PathBuf>,
@@ -119,6 +140,98 @@ impl From<codex_protocol::protocol::SandboxPolicy> for SandboxPolicy {
}
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct SandboxCommandAssessment {
pub description: String,
pub risk_level: CommandRiskLevel,
}
impl SandboxCommandAssessment {
pub fn into_core(self) -> CoreSandboxCommandAssessment {
CoreSandboxCommandAssessment {
description: self.description,
risk_level: self.risk_level.to_core(),
}
}
}
impl From<CoreSandboxCommandAssessment> for SandboxCommandAssessment {
fn from(value: CoreSandboxCommandAssessment) -> Self {
Self {
description: value.description,
risk_level: CommandRiskLevel::from(value.risk_level),
}
}
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(tag = "type", rename_all = "camelCase")]
#[ts(tag = "type")]
#[ts(export_to = "v2/")]
pub enum CommandAction {
Read {
command: String,
name: String,
path: PathBuf,
},
ListFiles {
command: String,
path: Option<String>,
},
Search {
command: String,
query: Option<String>,
path: Option<String>,
},
Unknown {
command: String,
},
}
impl CommandAction {
pub fn into_core(self) -> CoreParsedCommand {
match self {
CommandAction::Read {
command: cmd,
name,
path,
} => CoreParsedCommand::Read { cmd, name, path },
CommandAction::ListFiles { command: cmd, path } => {
CoreParsedCommand::ListFiles { cmd, path }
}
CommandAction::Search {
command: cmd,
query,
path,
} => CoreParsedCommand::Search { cmd, query, path },
CommandAction::Unknown { command: cmd } => CoreParsedCommand::Unknown { cmd },
}
}
}
impl From<CoreParsedCommand> for CommandAction {
fn from(value: CoreParsedCommand) -> Self {
match value {
CoreParsedCommand::Read { cmd, name, path } => CommandAction::Read {
command: cmd,
name,
path,
},
CoreParsedCommand::ListFiles { cmd, path } => {
CommandAction::ListFiles { command: cmd, path }
}
CoreParsedCommand::Search { cmd, query, path } => CommandAction::Search {
command: cmd,
query,
path,
},
CoreParsedCommand::Unknown { cmd } => CommandAction::Unknown { command: cmd },
}
}
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(tag = "type", rename_all = "camelCase")]
#[ts(tag = "type")]
@@ -279,7 +392,7 @@ pub struct ThreadStartParams {
pub cwd: Option<String>,
pub approval_policy: Option<AskForApproval>,
pub sandbox: Option<SandboxMode>,
pub config: Option<HashMap<String, serde_json::Value>>,
pub config: Option<HashMap<String, JsonValue>>,
pub base_instructions: Option<String>,
pub developer_instructions: Option<String>,
}
@@ -506,31 +619,50 @@ impl From<CoreUserInput> for UserInput {
#[ts(tag = "type")]
#[ts(export_to = "v2/")]
pub enum ThreadItem {
UserMessage {
id: String,
content: Vec<UserInput>,
},
AgentMessage {
id: String,
text: String,
},
#[serde(rename_all = "camelCase")]
#[ts(rename_all = "camelCase")]
UserMessage { id: String, content: Vec<UserInput> },
#[serde(rename_all = "camelCase")]
#[ts(rename_all = "camelCase")]
AgentMessage { id: String, text: String },
#[serde(rename_all = "camelCase")]
#[ts(rename_all = "camelCase")]
Reasoning {
id: String,
text: String,
#[serde(default)]
summary: Vec<String>,
#[serde(default)]
content: Vec<String>,
},
#[serde(rename_all = "camelCase")]
#[ts(rename_all = "camelCase")]
CommandExecution {
id: String,
/// The command to be executed.
command: String,
aggregated_output: String,
exit_code: Option<i32>,
/// The command's working directory.
cwd: PathBuf,
status: CommandExecutionStatus,
/// A best-effort parsing of the command to understand the action(s) it will perform.
/// This returns a list of CommandAction objects because a single shell command may
/// be composed of many commands piped together.
command_actions: Vec<CommandAction>,
/// The command's output, aggregated from stdout and stderr.
aggregated_output: Option<String>,
/// The command's exit code.
exit_code: Option<i32>,
/// The duration of the command execution in milliseconds.
duration_ms: Option<i64>,
},
#[serde(rename_all = "camelCase")]
#[ts(rename_all = "camelCase")]
FileChange {
id: String,
changes: Vec<FileUpdateChange>,
status: PatchApplyStatus,
},
#[serde(rename_all = "camelCase")]
#[ts(rename_all = "camelCase")]
McpToolCall {
id: String,
server: String,
@@ -540,22 +672,18 @@ pub enum ThreadItem {
result: Option<McpToolCallResult>,
error: Option<McpToolCallError>,
},
WebSearch {
id: String,
query: String,
},
TodoList {
id: String,
items: Vec<TodoItem>,
},
ImageView {
id: String,
path: String,
},
CodeReview {
id: String,
review: String,
},
#[serde(rename_all = "camelCase")]
#[ts(rename_all = "camelCase")]
WebSearch { id: String, query: String },
#[serde(rename_all = "camelCase")]
#[ts(rename_all = "camelCase")]
TodoList { id: String, items: Vec<TodoItem> },
#[serde(rename_all = "camelCase")]
#[ts(rename_all = "camelCase")]
ImageView { id: String, path: String },
#[serde(rename_all = "camelCase")]
#[ts(rename_all = "camelCase")]
CodeReview { id: String, review: String },
}
impl From<CoreTurnItem> for ThreadItem {
@@ -575,17 +703,11 @@ impl From<CoreTurnItem> for ThreadItem {
.collect::<String>();
ThreadItem::AgentMessage { id: agent.id, text }
}
CoreTurnItem::Reasoning(reasoning) => {
let text = if !reasoning.summary_text.is_empty() {
reasoning.summary_text.join("\n")
} else {
reasoning.raw_content.join("\n")
};
ThreadItem::Reasoning {
id: reasoning.id,
text,
}
}
CoreTurnItem::Reasoning(reasoning) => ThreadItem::Reasoning {
id: reasoning.id,
summary: reasoning.summary_text,
content: reasoning.raw_content,
},
CoreTurnItem::WebSearch(search) => ThreadItem::WebSearch {
id: search.id,
query: search.query,
@@ -643,7 +765,7 @@ pub enum McpToolCallStatus {
#[ts(export_to = "v2/")]
pub struct McpToolCallResult {
pub content: Vec<McpContentBlock>,
pub structured_content: JsonValue,
pub structured_content: Option<JsonValue>,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
@@ -719,6 +841,32 @@ pub struct AgentMessageDeltaNotification {
pub delta: String,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct ReasoningSummaryTextDeltaNotification {
pub item_id: String,
pub delta: String,
pub summary_index: i64,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct ReasoningSummaryPartAddedNotification {
pub item_id: String,
pub summary_index: i64,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct ReasoningTextDeltaNotification {
pub item_id: String,
pub delta: String,
pub content_index: i64,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
@@ -735,6 +883,39 @@ pub struct McpToolCallProgressNotification {
pub message: String,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct CommandExecutionRequestApprovalParams {
pub thread_id: String,
pub turn_id: String,
pub item_id: String,
/// Optional explanatory reason (e.g. request for network access).
pub reason: Option<String>,
/// Optional model-provided risk assessment describing the blocked command.
pub risk: Option<SandboxCommandAssessment>,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct CommandExecutionRequestAcceptSettings {
/// If true, automatically approve this command for the duration of the session.
#[serde(default)]
pub for_session: bool,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct CommandExecutionRequestApprovalResponse {
pub decision: ApprovalDecision,
/// Optional approval settings for when the decision is `accept`.
/// Ignored if the decision is `decline` or `cancel`.
#[serde(default)]
pub accept_settings: Option<CommandExecutionRequestAcceptSettings>,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
@@ -867,7 +1048,8 @@ mod tests {
ThreadItem::from(reasoning_item),
ThreadItem::Reasoning {
id: "reasoning-1".to_string(),
text: "line one\nline two".to_string(),
summary: vec!["line one".to_string(), "line two".to_string()],
content: vec![],
}
);
codex-rs/app-server-test-client/Cargo.lock
Generated
+1298
View File
File diff suppressed because it is too large Load Diff
codex-rs/app-server-test-client/Cargo.toml
+16
View File
@@ -0,0 +1,16 @@
[package]
name = "codex-app-server-test-client"
version = { workspace = true }
edition = "2024"
[lints]
workspace = true
[dependencies]
anyhow = { workspace = true }
clap = { workspace = true, features = ["derive", "env"] }
codex-app-server-protocol = { workspace = true }
codex-protocol = { workspace = true }
serde = { workspace = true, features = ["derive"] }
serde_json = { workspace = true }
uuid = { workspace = true, features = ["v4"] }
codex-rs/app-server-test-client/README.md
+2
View File
@@ -0,0 +1,2 @@
# App Server Test Client
Exercises simple `codex app-server` flows end-to-end, logging JSON-RPC messages sent between client and server to stdout.
codex-rs/app-server-test-client/src/main.rs
+691
View File
@@ -0,0 +1,691 @@
use std::collections::VecDeque;
use std::io::BufRead;
use std::io::BufReader;
use std::io::Write;
use std::process::Child;
use std::process::ChildStdin;
use std::process::ChildStdout;
use std::process::Command;
use std::process::Stdio;
use std::thread;
use std::time::Duration;
use anyhow::Context;
use anyhow::Result;
use anyhow::bail;
use clap::Parser;
use clap::Subcommand;
use codex_app_server_protocol::AddConversationListenerParams;
use codex_app_server_protocol::AddConversationSubscriptionResponse;
use codex_app_server_protocol::AskForApproval;
use codex_app_server_protocol::ClientInfo;
use codex_app_server_protocol::ClientRequest;
use codex_app_server_protocol::GetAccountRateLimitsResponse;
use codex_app_server_protocol::InitializeParams;
use codex_app_server_protocol::InitializeResponse;
use codex_app_server_protocol::InputItem;
use codex_app_server_protocol::JSONRPCMessage;
use codex_app_server_protocol::JSONRPCNotification;
use codex_app_server_protocol::JSONRPCResponse;
use codex_app_server_protocol::LoginChatGptCompleteNotification;
use codex_app_server_protocol::LoginChatGptResponse;
use codex_app_server_protocol::NewConversationParams;
use codex_app_server_protocol::NewConversationResponse;
use codex_app_server_protocol::RequestId;
use codex_app_server_protocol::SandboxPolicy;
use codex_app_server_protocol::SendUserMessageParams;
use codex_app_server_protocol::SendUserMessageResponse;
use codex_app_server_protocol::ServerNotification;
use codex_app_server_protocol::ThreadStartParams;
use codex_app_server_protocol::ThreadStartResponse;
use codex_app_server_protocol::TurnStartParams;
use codex_app_server_protocol::TurnStartResponse;
use codex_app_server_protocol::UserInput as V2UserInput;
use codex_protocol::ConversationId;
use codex_protocol::protocol::Event;
use codex_protocol::protocol::EventMsg;
use serde::de::DeserializeOwned;
use serde_json::Value;
use uuid::Uuid;
/// Minimal launcher that initializes the Codex app-server and logs the handshake.
#[derive(Parser)]
#[command(author = "Codex", version, about = "Bootstrap Codex app-server", long_about = None)]
struct Cli {
/// Path to the `codex` CLI binary.
#[arg(long, env = "CODEX_BIN", default_value = "codex")]
codex_bin: String,
#[command(subcommand)]
command: CliCommand,
}
#[derive(Subcommand)]
enum CliCommand {
/// Send a user message through the Codex app-server.
SendMessage {
/// User message to send to Codex.
#[arg()]
user_message: String,
},
/// Send a user message through the app-server V2 thread/turn APIs.
SendMessageV2 {
/// User message to send to Codex.
#[arg()]
user_message: String,
},
/// Start a V2 turn that elicits an ExecCommand approval.
#[command(name = "trigger-cmd-approval")]
TriggerCmdApproval {
/// Optional prompt; defaults to a simple python command.
#[arg()]
user_message: Option<String>,
},
/// Start a V2 turn that elicits an ApplyPatch approval.
#[command(name = "trigger-patch-approval")]
TriggerPatchApproval {
/// Optional prompt; defaults to creating a file via apply_patch.
#[arg()]
user_message: Option<String>,
},
/// Start a V2 turn that should not elicit an ExecCommand approval.
#[command(name = "no-trigger-cmd-approval")]
NoTriggerCmdApproval,
/// Trigger the ChatGPT login flow and wait for completion.
TestLogin,
/// Fetch the current account rate limits from the Codex app-server.
GetAccountRateLimits,
}
fn main() -> Result<()> {
let Cli { codex_bin, command } = Cli::parse();
match command {
CliCommand::SendMessage { user_message } => send_message(codex_bin, user_message),
CliCommand::SendMessageV2 { user_message } => send_message_v2(codex_bin, user_message),
CliCommand::TriggerCmdApproval { user_message } => {
trigger_cmd_approval(codex_bin, user_message)
}
CliCommand::TriggerPatchApproval { user_message } => {
trigger_patch_approval(codex_bin, user_message)
}
CliCommand::NoTriggerCmdApproval => no_trigger_cmd_approval(codex_bin),
CliCommand::TestLogin => test_login(codex_bin),
CliCommand::GetAccountRateLimits => get_account_rate_limits(codex_bin),
}
}
fn send_message(codex_bin: String, user_message: String) -> Result<()> {
let mut client = CodexClient::spawn(codex_bin)?;
let initialize = client.initialize()?;
println!("< initialize response: {initialize:?}");
let conversation = client.new_conversation()?;
println!("< newConversation response: {conversation:?}");
let subscription = client.add_conversation_listener(&conversation.conversation_id)?;
println!("< addConversationListener response: {subscription:?}");
let send_response = client.send_user_message(&conversation.conversation_id, &user_message)?;
println!("< sendUserMessage response: {send_response:?}");
client.stream_conversation(&conversation.conversation_id)?;
client.remove_conversation_listener(subscription.subscription_id)?;
Ok(())
}
fn send_message_v2(codex_bin: String, user_message: String) -> Result<()> {
send_message_v2_with_policies(codex_bin, user_message, None, None)
}
fn trigger_cmd_approval(codex_bin: String, user_message: Option<String>) -> Result<()> {
let default_prompt =
"Run `touch /tmp/should-trigger-approval` so I can confirm the file exists.";
let message = user_message.unwrap_or_else(|| default_prompt.to_string());
send_message_v2_with_policies(
codex_bin,
message,
Some(AskForApproval::OnRequest),
Some(SandboxPolicy::ReadOnly),
)
}
fn trigger_patch_approval(codex_bin: String, user_message: Option<String>) -> Result<()> {
let default_prompt =
"Create a file named APPROVAL_DEMO.txt containing a short hello message using apply_patch.";
let message = user_message.unwrap_or_else(|| default_prompt.to_string());
send_message_v2_with_policies(
codex_bin,
message,
Some(AskForApproval::OnRequest),
Some(SandboxPolicy::ReadOnly),
)
}
fn no_trigger_cmd_approval(codex_bin: String) -> Result<()> {
let prompt = "Run `touch should_not_trigger_approval.txt`";
send_message_v2_with_policies(codex_bin, prompt.to_string(), None, None)
}
fn send_message_v2_with_policies(
codex_bin: String,
user_message: String,
approval_policy: Option<AskForApproval>,
sandbox_policy: Option<SandboxPolicy>,
) -> Result<()> {
let mut client = CodexClient::spawn(codex_bin)?;
let initialize = client.initialize()?;
println!("< initialize response: {initialize:?}");
let thread_response = client.thread_start(ThreadStartParams::default())?;
println!("< thread/start response: {thread_response:?}");
let mut turn_params = TurnStartParams {
thread_id: thread_response.thread.id.clone(),
input: vec![V2UserInput::Text { text: user_message }],
..Default::default()
};
turn_params.approval_policy = approval_policy;
turn_params.sandbox_policy = sandbox_policy;
let turn_response = client.turn_start(turn_params)?;
println!("< turn/start response: {turn_response:?}");
client.stream_turn(&thread_response.thread.id, &turn_response.turn.id)?;
Ok(())
}
fn test_login(codex_bin: String) -> Result<()> {
let mut client = CodexClient::spawn(codex_bin)?;
let initialize = client.initialize()?;
println!("< initialize response: {initialize:?}");
let login_response = client.login_chat_gpt()?;
println!("< loginChatGpt response: {login_response:?}");
println!(
"Open the following URL in your browser to continue:\n{}",
login_response.auth_url
);
let completion = client.wait_for_login_completion(&login_response.login_id)?;
println!("< loginChatGptComplete notification: {completion:?}");
if completion.success {
println!("Login succeeded.");
Ok(())
} else {
bail!(
"login failed: {}",
completion
.error
.as_deref()
.unwrap_or("unknown error from loginChatGptComplete")
);
}
}
fn get_account_rate_limits(codex_bin: String) -> Result<()> {
let mut client = CodexClient::spawn(codex_bin)?;
let initialize = client.initialize()?;
println!("< initialize response: {initialize:?}");
let response = client.get_account_rate_limits()?;
println!("< account/rateLimits/read response: {response:?}");
Ok(())
}
struct CodexClient {
child: Child,
stdin: Option<ChildStdin>,
stdout: BufReader<ChildStdout>,
pending_notifications: VecDeque<JSONRPCNotification>,
}
impl CodexClient {
fn spawn(codex_bin: String) -> Result<Self> {
let mut codex_app_server = Command::new(&codex_bin)
.arg("app-server")
.stdin(Stdio::piped())
.stdout(Stdio::piped())
.stderr(Stdio::inherit())
.spawn()
.with_context(|| format!("failed to start `{codex_bin}` app-server"))?;
let stdin = codex_app_server
.stdin
.take()
.context("codex app-server stdin unavailable")?;
let stdout = codex_app_server
.stdout
.take()
.context("codex app-server stdout unavailable")?;
Ok(Self {
child: codex_app_server,
stdin: Some(stdin),
stdout: BufReader::new(stdout),
pending_notifications: VecDeque::new(),
})
}
fn initialize(&mut self) -> Result<InitializeResponse> {
let request_id = self.request_id();
let request = ClientRequest::Initialize {
request_id: request_id.clone(),
params: InitializeParams {
client_info: ClientInfo {
name: "codex-toy-app-server".to_string(),
title: Some("Codex Toy App Server".to_string()),
version: env!("CARGO_PKG_VERSION").to_string(),
},
},
};
self.send_request(request, request_id, "initialize")
}
fn new_conversation(&mut self) -> Result<NewConversationResponse> {
let request_id = self.request_id();
let request = ClientRequest::NewConversation {
request_id: request_id.clone(),
params: NewConversationParams::default(),
};
self.send_request(request, request_id, "newConversation")
}
fn add_conversation_listener(
&mut self,
conversation_id: &ConversationId,
) -> Result<AddConversationSubscriptionResponse> {
let request_id = self.request_id();
let request = ClientRequest::AddConversationListener {
request_id: request_id.clone(),
params: AddConversationListenerParams {
conversation_id: *conversation_id,
experimental_raw_events: false,
},
};
self.send_request(request, request_id, "addConversationListener")
}
fn remove_conversation_listener(&mut self, subscription_id: Uuid) -> Result<()> {
let request_id = self.request_id();
let request = ClientRequest::RemoveConversationListener {
request_id: request_id.clone(),
params: codex_app_server_protocol::RemoveConversationListenerParams { subscription_id },
};
self.send_request::<codex_app_server_protocol::RemoveConversationSubscriptionResponse>(
request,
request_id,
"removeConversationListener",
)?;
Ok(())
}
fn send_user_message(
&mut self,
conversation_id: &ConversationId,
message: &str,
) -> Result<SendUserMessageResponse> {
let request_id = self.request_id();
let request = ClientRequest::SendUserMessage {
request_id: request_id.clone(),
params: SendUserMessageParams {
conversation_id: *conversation_id,
items: vec![InputItem::Text {
text: message.to_string(),
}],
},
};
self.send_request(request, request_id, "sendUserMessage")
}
fn thread_start(&mut self, params: ThreadStartParams) -> Result<ThreadStartResponse> {
let request_id = self.request_id();
let request = ClientRequest::ThreadStart {
request_id: request_id.clone(),
params,
};
self.send_request(request, request_id, "thread/start")
}
fn turn_start(&mut self, params: TurnStartParams) -> Result<TurnStartResponse> {
let request_id = self.request_id();
let request = ClientRequest::TurnStart {
request_id: request_id.clone(),
params,
};
self.send_request(request, request_id, "turn/start")
}
fn login_chat_gpt(&mut self) -> Result<LoginChatGptResponse> {
let request_id = self.request_id();
let request = ClientRequest::LoginChatGpt {
request_id: request_id.clone(),
params: None,
};
self.send_request(request, request_id, "loginChatGpt")
}
fn get_account_rate_limits(&mut self) -> Result<GetAccountRateLimitsResponse> {
let request_id = self.request_id();
let request = ClientRequest::GetAccountRateLimits {
request_id: request_id.clone(),
params: None,
};
self.send_request(request, request_id, "account/rateLimits/read")
}
fn stream_conversation(&mut self, conversation_id: &ConversationId) -> Result<()> {
loop {
let notification = self.next_notification()?;
if !notification.method.starts_with("codex/event/") {
continue;
}
if let Some(event) = self.extract_event(notification, conversation_id)? {
match &event.msg {
EventMsg::AgentMessage(event) => {
println!("{}", event.message);
}
EventMsg::AgentMessageDelta(event) => {
print!("{}", event.delta);
std::io::stdout().flush().ok();
}
EventMsg::TaskComplete(event) => {
println!("\n[task complete: {event:?}]");
break;
}
EventMsg::TurnAborted(event) => {
println!("\n[turn aborted: {:?}]", event.reason);
break;
}
EventMsg::Error(event) => {
println!("[error] {event:?}");
}
_ => {
println!("[UNKNOWN EVENT] {:?}", event.msg);
}
}
}
}
Ok(())
}
fn wait_for_login_completion(
&mut self,
expected_login_id: &Uuid,
) -> Result<LoginChatGptCompleteNotification> {
loop {
let notification = self.next_notification()?;
if let Ok(server_notification) = ServerNotification::try_from(notification) {
match server_notification {
ServerNotification::LoginChatGptComplete(completion) => {
if &completion.login_id == expected_login_id {
return Ok(completion);
}
println!(
"[ignoring loginChatGptComplete for unexpected login_id: {}]",
completion.login_id
);
}
ServerNotification::AuthStatusChange(status) => {
println!("< authStatusChange notification: {status:?}");
}
ServerNotification::AccountRateLimitsUpdated(snapshot) => {
println!("< accountRateLimitsUpdated notification: {snapshot:?}");
}
ServerNotification::SessionConfigured(_) => {
// SessionConfigured notifications are unrelated to login; skip.
}
_ => {}
}
}
// Not a server notification (likely a conversation event); keep waiting.
}
}
fn stream_turn(&mut self, thread_id: &str, turn_id: &str) -> Result<()> {
loop {
let notification = self.next_notification()?;
let Ok(server_notification) = ServerNotification::try_from(notification) else {
continue;
};
match server_notification {
ServerNotification::ThreadStarted(payload) => {
if payload.thread.id == thread_id {
println!("< thread/started notification: {:?}", payload.thread);
}
}
ServerNotification::TurnStarted(payload) => {
if payload.turn.id == turn_id {
println!("< turn/started notification: {:?}", payload.turn.status);
}
}
ServerNotification::AgentMessageDelta(delta) => {
print!("{}", delta.delta);
std::io::stdout().flush().ok();
}
ServerNotification::CommandExecutionOutputDelta(delta) => {
print!("{}", delta.delta);
std::io::stdout().flush().ok();
}
ServerNotification::ItemStarted(payload) => {
println!("\n< item started: {:?}", payload.item);
}
ServerNotification::ItemCompleted(payload) => {
println!("< item completed: {:?}", payload.item);
}
ServerNotification::TurnCompleted(payload) => {
if payload.turn.id == turn_id {
println!("\n< turn/completed notification: {:?}", payload.turn.status);
if let Some(error) = payload.turn.error {
println!("[turn error] {}", error.message);
}
println!("< usage: {:?}", payload.usage);
break;
}
}
ServerNotification::McpToolCallProgress(payload) => {
println!("< MCP tool progress: {}", payload.message);
}
_ => {
println!("[UNKNOWN SERVER NOTIFICATION] {server_notification:?}");
}
}
}
Ok(())
}
fn extract_event(
&self,
notification: JSONRPCNotification,
conversation_id: &ConversationId,
) -> Result<Option<Event>> {
let params = notification
.params
.context("event notification missing params")?;
let mut map = match params {
Value::Object(map) => map,
other => bail!("unexpected params shape: {other:?}"),
};
let conversation_value = map
.remove("conversationId")
.context("event missing conversationId")?;
let notification_conversation: ConversationId = serde_json::from_value(conversation_value)
.context("conversationId was not a valid UUID")?;
if &notification_conversation != conversation_id {
return Ok(None);
}
let event_value = Value::Object(map);
let event: Event =
serde_json::from_value(event_value).context("failed to decode event payload")?;
Ok(Some(event))
}
fn send_request<T>(
&mut self,
request: ClientRequest,
request_id: RequestId,
method: &str,
) -> Result<T>
where
T: DeserializeOwned,
{
self.write_request(&request)?;
self.wait_for_response(request_id, method)
}
fn write_request(&mut self, request: &ClientRequest) -> Result<()> {
let request_json = serde_json::to_string(request)?;
let request_pretty = serde_json::to_string_pretty(request)?;
print_multiline_with_prefix("> ", &request_pretty);
if let Some(stdin) = self.stdin.as_mut() {
writeln!(stdin, "{request_json}")?;
stdin
.flush()
.context("failed to flush request to codex app-server")?;
} else {
bail!("codex app-server stdin closed");
}
Ok(())
}
fn wait_for_response<T>(&mut self, request_id: RequestId, method: &str) -> Result<T>
where
T: DeserializeOwned,
{
loop {
let message = self.read_jsonrpc_message()?;
match message {
JSONRPCMessage::Response(JSONRPCResponse { id, result }) => {
if id == request_id {
return serde_json::from_value(result)
.with_context(|| format!("{method} response missing payload"));
}
}
JSONRPCMessage::Error(err) => {
if err.id == request_id {
bail!("{method} failed: {err:?}");
}
}
JSONRPCMessage::Notification(notification) => {
self.pending_notifications.push_back(notification);
}
JSONRPCMessage::Request(_) => {
bail!("unexpected request from codex app-server");
}
}
}
}
fn next_notification(&mut self) -> Result<JSONRPCNotification> {
if let Some(notification) = self.pending_notifications.pop_front() {
return Ok(notification);
}
loop {
let message = self.read_jsonrpc_message()?;
match message {
JSONRPCMessage::Notification(notification) => return Ok(notification),
JSONRPCMessage::Response(_) | JSONRPCMessage::Error(_) => {
// No outstanding requests, so ignore stray responses/errors for now.
continue;
}
JSONRPCMessage::Request(_) => {
bail!("unexpected request from codex app-server");
}
}
}
}
fn read_jsonrpc_message(&mut self) -> Result<JSONRPCMessage> {
loop {
let mut response_line = String::new();
let bytes = self
.stdout
.read_line(&mut response_line)
.context("failed to read from codex app-server")?;
if bytes == 0 {
bail!("codex app-server closed stdout");
}
let trimmed = response_line.trim();
if trimmed.is_empty() {
continue;
}
let parsed: Value =
serde_json::from_str(trimmed).context("response was not valid JSON-RPC")?;
let pretty = serde_json::to_string_pretty(&parsed)?;
print_multiline_with_prefix("< ", &pretty);
let message: JSONRPCMessage = serde_json::from_value(parsed)
.context("response was not a valid JSON-RPC message")?;
return Ok(message);
}
}
fn request_id(&self) -> RequestId {
RequestId::String(Uuid::new_v4().to_string())
}
}
fn print_multiline_with_prefix(prefix: &str, payload: &str) {
for line in payload.lines() {
println!("{prefix}{line}");
}
}
impl Drop for CodexClient {
fn drop(&mut self) {
let _ = self.stdin.take();
if let Ok(Some(status)) = self.child.try_wait() {
println!("[codex app-server exited: {status}]");
return;
}
thread::sleep(Duration::from_millis(100));
if let Ok(Some(status)) = self.child.try_wait() {
println!("[codex app-server exited: {status}]");
return;
}
let _ = self.child.kill();
let _ = self.child.wait();
}
}
codex-rs/app-server/Cargo.toml
+1
View File
@@ -46,6 +46,7 @@ app_test_support = { workspace = true }
assert_cmd = { workspace = true }
base64 = { workspace = true }
core_test_support = { workspace = true }
mcp-types = { workspace = true }
os_info = { workspace = true }
pretty_assertions = { workspace = true }
serial_test = { workspace = true }
codex-rs/app-server/README.md
+50 -2
View File
@@ -2,6 +2,16 @@
`codex app-server` is the interface Codex uses to power rich interfaces such as the [Codex VS Code extension](https://marketplace.visualstudio.com/items?itemName=openai.chatgpt). The message schema is currently unstable, but those who wish to build experimental UIs on top of Codex may find it valuable.
## Table of Contents
- [Protocol](#protocol)
- [Message Schema](#message-schema)
- [Lifecycle Overview](#lifecycle-overview)
- [Initialization](#initialization)
- [Core primitives](#core-primitives)
- [Thread & turn endpoints](#thread--turn-endpoints)
- [Auth endpoints](#auth-endpoints)
- [Events (work-in-progress)](#v2-streaming-events-work-in-progress)
## Protocol
Similar to [MCP](https://modelcontextprotocol.io/), `codex app-server` supports bidirectional communication, streaming JSONL over stdio. The protocol is JSON-RPC 2.0, though the `"jsonrpc":"2.0"` header is omitted.
@@ -15,6 +25,14 @@ codex app-server generate-ts --out DIR
codex app-server generate-json-schema --out DIR
```
## Lifecycle Overview
- Initialize once: Immediately after launching the codex app-server process, send an `initialize` request with your client metadata, then emit an `initialized` notification. Any other request before this handshake gets rejected.
- Start (or resume) a thread: Call `thread/start` to open a fresh conversation. The response returns the thread object and you’ll also get a `thread/started` notification. If you’re continuing an existing conversation, call `thread/resume` with its ID instead.
- Begin a turn: To send user input, call `turn/start` with the target `threadId` and the user's input. Optional fields let you override model, cwd, sandbox policy, etc. This immediately returns the new turn object and triggers a `turn/started` notification.
- Stream events: After `turn/start`, keep reading JSON-RPC notifications on stdout. You’ll see `item/started`, `item/completed`, deltas like `item/agentMessage/delta`, tool progress, etc. These represent streaming model output plus any side effects (commands, tool calls, reasoning notes).
- Finish the turn: When the model is done (or the turn is interrupted via making the `turn/interrupt` call), the server sends `turn/completed` with the final turn state and token usage.
## Initialization
Clients must send a single `initialize` request before invoking any other method, then acknowledge with an `initialized` notification. The server returns the user agent string it will present to upstream services; subsequent requests issued before initialization receive a `"Not initialized"` error, and repeated `initialize` calls receive an `"Already initialized"` error.
@@ -56,7 +74,7 @@ Start a fresh thread when you need a new Codex conversation.
{ "method": "thread/start", "id": 10, "params": {
// Optionally set config settings. If not specified, will use the user's
// current config settings.
"model": "gpt-5-codex",
"model": "gpt-5.1-codex",
"cwd": "/Users/me/project",
"approvalPolicy": "never",
"sandbox": "workspaceWrite",
@@ -137,7 +155,7 @@ You can optionally specify config overrides on the new turn. If specified, these
"writableRoots": ["/Users/me/project"],
"networkAccess": true
},
"model": "gpt-5-codex",
"model": "gpt-5.1-codex",
"effort": "medium",
"summary": "concise"
} }
@@ -258,3 +276,33 @@ Field notes:
- `codex app-server generate-ts --out <dir>` emits v2 types under `v2/`.
- `codex app-server generate-json-schema --out <dir>` outputs `codex_app_server_protocol.schemas.json`.
- See [“Authentication and authorization” in the config docs](../../docs/config.md#authentication-and-authorization) for configuration knobs.
## Events (work-in-progress)
Event notifications are the server-initiated event stream for thread lifecycles, turn lifecycles, and the items within them. After you start or resume a thread, keep reading stdout for `thread/started`, `turn/*`, and `item/*` notifications.
### Turn events
The app-server streams JSON-RPC notifications while a turn is running. Each turn starts with `turn/started` (initial `turn`) and ends with `turn/completed` (final `turn` plus token `usage`), and clients subscribe to the events they care about, rendering each item incrementally as updates arrive. The per-item lifecycle is always: `item/started` → zero or more item-specific deltas → `item/completed`.
#### Thread items
`ThreadItem` is the tagged union carried in turn responses and `item/*` notifications. Currently we support events for the following items:
- `userMessage` — `{id, content}` where `content` is a list of user inputs (`text`, `image`, or `localImage`).
- `agentMessage` — `{id, text}` containing the accumulated agent reply.
- `reasoning` — `{id, summary, content}` where `summary` holds streamed reasoning summaries (applicable for most OpenAI models) and `content` holds raw reasoning blocks (applicable for e.g. open source models).
- `mcpToolCall` — `{id, server, tool, status, arguments, result?, error?}` describing MCP calls; `status` is `inProgress`, `completed`, or `failed`.
- `webSearch` — `{id, query}` for a web search request issued by the agent.
All items emit two shared lifecycle events:
- `item/started` — emits the full `item` when a new unit of work begins so the UI can render it immediately; the `item.id` in this payload matches the `itemId` used by deltas.
- `item/completed` — sends the final `item` once that work finishes (e.g., after a tool call or message completes); treat this as the authoritative state.
There are additional item-specific events:
#### agentMessage
- `item/agentMessage/delta` — appends streamed text for the agent message; concatenate `delta` values for the same `itemId` in order to reconstruct the full reply.
#### reasoning
- `item/reasoning/summaryTextDelta` — streams readable reasoning summaries; `summaryIndex` increments when a new summary section opens.
- `item/reasoning/summaryPartAdded` — marks the boundary between reasoning summary sections for an `itemId`; subsequent `summaryTextDelta` entries share the same `summaryIndex`.
- `item/reasoning/textDelta` — streams raw reasoning text (only applicable for e.g. open source models); use `contentIndex` to group deltas that belong together before showing them in the UI.
codex-rs/app-server/src/bespoke_event_handling.rs
+628
View File
@@ -0,0 +1,628 @@
use crate::codex_message_processor::ApiVersion;
use crate::codex_message_processor::PendingInterrupts;
use crate::outgoing_message::OutgoingMessageSender;
use codex_app_server_protocol::AccountRateLimitsUpdatedNotification;
use codex_app_server_protocol::AgentMessageDeltaNotification;
use codex_app_server_protocol::ApplyPatchApprovalParams;
use codex_app_server_protocol::ApplyPatchApprovalResponse;
use codex_app_server_protocol::ApprovalDecision;
use codex_app_server_protocol::CommandAction as V2ParsedCommand;
use codex_app_server_protocol::CommandExecutionOutputDeltaNotification;
use codex_app_server_protocol::CommandExecutionRequestApprovalParams;
use codex_app_server_protocol::CommandExecutionRequestApprovalResponse;
use codex_app_server_protocol::CommandExecutionStatus;
use codex_app_server_protocol::ExecCommandApprovalParams;
use codex_app_server_protocol::ExecCommandApprovalResponse;
use codex_app_server_protocol::InterruptConversationResponse;
use codex_app_server_protocol::ItemCompletedNotification;
use codex_app_server_protocol::ItemStartedNotification;
use codex_app_server_protocol::McpToolCallError;
use codex_app_server_protocol::McpToolCallResult;
use codex_app_server_protocol::McpToolCallStatus;
use codex_app_server_protocol::ReasoningSummaryPartAddedNotification;
use codex_app_server_protocol::ReasoningSummaryTextDeltaNotification;
use codex_app_server_protocol::ReasoningTextDeltaNotification;
use codex_app_server_protocol::SandboxCommandAssessment as V2SandboxCommandAssessment;
use codex_app_server_protocol::ServerNotification;
use codex_app_server_protocol::ServerRequestPayload;
use codex_app_server_protocol::ThreadItem;
use codex_app_server_protocol::TurnInterruptResponse;
use codex_core::CodexConversation;
use codex_core::parse_command::shlex_join;
use codex_core::protocol::ApplyPatchApprovalRequestEvent;
use codex_core::protocol::Event;
use codex_core::protocol::EventMsg;
use codex_core::protocol::ExecApprovalRequestEvent;
use codex_core::protocol::ExecCommandEndEvent;
use codex_core::protocol::McpToolCallBeginEvent;
use codex_core::protocol::McpToolCallEndEvent;
use codex_core::protocol::Op;
use codex_core::protocol::ReviewDecision;
use codex_protocol::ConversationId;
use std::convert::TryFrom;
use std::sync::Arc;
use tokio::sync::oneshot;
use tracing::error;
type JsonValue = serde_json::Value;
pub(crate) async fn apply_bespoke_event_handling(
event: Event,
conversation_id: ConversationId,
conversation: Arc<CodexConversation>,
outgoing: Arc<OutgoingMessageSender>,
pending_interrupts: PendingInterrupts,
api_version: ApiVersion,
) {
let Event { id: event_id, msg } = event;
match msg {
EventMsg::ApplyPatchApprovalRequest(ApplyPatchApprovalRequestEvent {
call_id,
changes,
reason,
grant_root,
}) => {
let params = ApplyPatchApprovalParams {
conversation_id,
call_id,
file_changes: changes,
reason,
grant_root,
};
let rx = outgoing
.send_request(ServerRequestPayload::ApplyPatchApproval(params))
.await;
tokio::spawn(async move {
on_patch_approval_response(event_id, rx, conversation).await;
});
}
EventMsg::ExecApprovalRequest(ExecApprovalRequestEvent {
call_id,
turn_id,
command,
cwd,
reason,
risk,
parsed_cmd,
}) => match api_version {
ApiVersion::V1 => {
let params = ExecCommandApprovalParams {
conversation_id,
call_id,
command,
cwd,
reason,
risk,
parsed_cmd,
};
let rx = outgoing
.send_request(ServerRequestPayload::ExecCommandApproval(params))
.await;
tokio::spawn(async move {
on_exec_approval_response(event_id, rx, conversation).await;
});
}
ApiVersion::V2 => {
let params = CommandExecutionRequestApprovalParams {
thread_id: conversation_id.to_string(),
turn_id: turn_id.clone(),
// Until we migrate the core to be aware of a first class CommandExecutionItem
// and emit the corresponding EventMsg, we repurpose the call_id as the item_id.
item_id: call_id.clone(),
reason,
risk: risk.map(V2SandboxCommandAssessment::from),
};
let rx = outgoing
.send_request(ServerRequestPayload::CommandExecutionRequestApproval(
params,
))
.await;
tokio::spawn(async move {
on_command_execution_request_approval_response(event_id, rx, conversation)
.await;
});
}
},
// TODO(celia): properly construct McpToolCall TurnItem in core.
EventMsg::McpToolCallBegin(begin_event) => {
let notification = construct_mcp_tool_call_notification(begin_event).await;
outgoing
.send_server_notification(ServerNotification::ItemStarted(notification))
.await;
}
EventMsg::McpToolCallEnd(end_event) => {
let notification = construct_mcp_tool_call_end_notification(end_event).await;
outgoing
.send_server_notification(ServerNotification::ItemCompleted(notification))
.await;
}
EventMsg::AgentMessageContentDelta(event) => {
let notification = AgentMessageDeltaNotification {
item_id: event.item_id,
delta: event.delta,
};
outgoing
.send_server_notification(ServerNotification::AgentMessageDelta(notification))
.await;
}
EventMsg::ReasoningContentDelta(event) => {
let notification = ReasoningSummaryTextDeltaNotification {
item_id: event.item_id,
delta: event.delta,
summary_index: event.summary_index,
};
outgoing
.send_server_notification(ServerNotification::ReasoningSummaryTextDelta(
notification,
))
.await;
}
EventMsg::ReasoningRawContentDelta(event) => {
let notification = ReasoningTextDeltaNotification {
item_id: event.item_id,
delta: event.delta,
content_index: event.content_index,
};
outgoing
.send_server_notification(ServerNotification::ReasoningTextDelta(notification))
.await;
}
EventMsg::AgentReasoningSectionBreak(event) => {
let notification = ReasoningSummaryPartAddedNotification {
item_id: event.item_id,
summary_index: event.summary_index,
};
outgoing
.send_server_notification(ServerNotification::ReasoningSummaryPartAdded(
notification,
))
.await;
}
EventMsg::TokenCount(token_count_event) => {
if let Some(rate_limits) = token_count_event.rate_limits {
outgoing
.send_server_notification(ServerNotification::AccountRateLimitsUpdated(
AccountRateLimitsUpdatedNotification {
rate_limits: rate_limits.into(),
},
))
.await;
}
}
EventMsg::ItemStarted(item_started_event) => {
let item: ThreadItem = item_started_event.item.clone().into();
let notification = ItemStartedNotification { item };
outgoing
.send_server_notification(ServerNotification::ItemStarted(notification))
.await;
}
EventMsg::ItemCompleted(item_completed_event) => {
let item: ThreadItem = item_completed_event.item.clone().into();
let notification = ItemCompletedNotification { item };
outgoing
.send_server_notification(ServerNotification::ItemCompleted(notification))
.await;
}
EventMsg::ExecCommandBegin(exec_command_begin_event) => {
let item = ThreadItem::CommandExecution {
id: exec_command_begin_event.call_id.clone(),
command: shlex_join(&exec_command_begin_event.command),
cwd: exec_command_begin_event.cwd,
status: CommandExecutionStatus::InProgress,
command_actions: exec_command_begin_event
.parsed_cmd
.into_iter()
.map(V2ParsedCommand::from)
.collect(),
aggregated_output: None,
exit_code: None,
duration_ms: None,
};
let notification = ItemStartedNotification { item };
outgoing
.send_server_notification(ServerNotification::ItemStarted(notification))
.await;
}
EventMsg::ExecCommandOutputDelta(exec_command_output_delta_event) => {
let notification = CommandExecutionOutputDeltaNotification {
item_id: exec_command_output_delta_event.call_id.clone(),
delta: String::from_utf8_lossy(&exec_command_output_delta_event.chunk).to_string(),
};
outgoing
.send_server_notification(ServerNotification::CommandExecutionOutputDelta(
notification,
))
.await;
}
EventMsg::ExecCommandEnd(exec_command_end_event) => {
let ExecCommandEndEvent {
call_id,
command,
cwd,
parsed_cmd,
aggregated_output,
exit_code,
duration,
..
} = exec_command_end_event;
let status = if exit_code == 0 {
CommandExecutionStatus::Completed
} else {
CommandExecutionStatus::Failed
};
let aggregated_output = if aggregated_output.is_empty() {
None
} else {
Some(aggregated_output)
};
let duration_ms = i64::try_from(duration.as_millis()).unwrap_or(i64::MAX);
let item = ThreadItem::CommandExecution {
id: call_id,
command: shlex_join(&command),
cwd,
status,
command_actions: parsed_cmd.into_iter().map(V2ParsedCommand::from).collect(),
aggregated_output,
exit_code: Some(exit_code),
duration_ms: Some(duration_ms),
};
let notification = ItemCompletedNotification { item };
outgoing
.send_server_notification(ServerNotification::ItemCompleted(notification))
.await;
}
// If this is a TurnAborted, reply to any pending interrupt requests.
EventMsg::TurnAborted(turn_aborted_event) => {
let pending = {
let mut map = pending_interrupts.lock().await;
map.remove(&conversation_id).unwrap_or_default()
};
if !pending.is_empty() {
for (rid, ver) in pending {
match ver {
ApiVersion::V1 => {
let response = InterruptConversationResponse {
abort_reason: turn_aborted_event.reason.clone(),
};
outgoing.send_response(rid, response).await;
}
ApiVersion::V2 => {
let response = TurnInterruptResponse {};
outgoing.send_response(rid, response).await;
}
}
}
}
}
_ => {}
}
}
async fn on_patch_approval_response(
event_id: String,
receiver: oneshot::Receiver<JsonValue>,
codex: Arc<CodexConversation>,
) {
let response = receiver.await;
let value = match response {
Ok(value) => value,
Err(err) => {
error!("request failed: {err:?}");
if let Err(submit_err) = codex
.submit(Op::PatchApproval {
id: event_id.clone(),
decision: ReviewDecision::Denied,
})
.await
{
error!("failed to submit denied PatchApproval after request failure: {submit_err}");
}
return;
}
};
let response =
serde_json::from_value::<ApplyPatchApprovalResponse>(value).unwrap_or_else(|err| {
error!("failed to deserialize ApplyPatchApprovalResponse: {err}");
ApplyPatchApprovalResponse {
decision: ReviewDecision::Denied,
}
});
if let Err(err) = codex
.submit(Op::PatchApproval {
id: event_id,
decision: response.decision,
})
.await
{
error!("failed to submit PatchApproval: {err}");
}
}
async fn on_exec_approval_response(
event_id: String,
receiver: oneshot::Receiver<JsonValue>,
conversation: Arc<CodexConversation>,
) {
let response = receiver.await;
let value = match response {
Ok(value) => value,
Err(err) => {
error!("request failed: {err:?}");
return;
}
};
// Try to deserialize `value` and then make the appropriate call to `codex`.
let response =
serde_json::from_value::<ExecCommandApprovalResponse>(value).unwrap_or_else(|err| {
error!("failed to deserialize ExecCommandApprovalResponse: {err}");
// If we cannot deserialize the response, we deny the request to be
// conservative.
ExecCommandApprovalResponse {
decision: ReviewDecision::Denied,
}
});
if let Err(err) = conversation
.submit(Op::ExecApproval {
id: event_id,
decision: response.decision,
})
.await
{
error!("failed to submit ExecApproval: {err}");
}
}
async fn on_command_execution_request_approval_response(
event_id: String,
receiver: oneshot::Receiver<JsonValue>,
conversation: Arc<CodexConversation>,
) {
let response = receiver.await;
let value = match response {
Ok(value) => value,
Err(err) => {
error!("request failed: {err:?}");
return;
}
};
let response = serde_json::from_value::<CommandExecutionRequestApprovalResponse>(value)
.unwrap_or_else(|err| {
error!("failed to deserialize CommandExecutionRequestApprovalResponse: {err}");
CommandExecutionRequestApprovalResponse {
decision: ApprovalDecision::Decline,
accept_settings: None,
}
});
let CommandExecutionRequestApprovalResponse {
decision,
accept_settings,
} = response;
let decision = match (decision, accept_settings) {
(ApprovalDecision::Accept, Some(settings)) if settings.for_session => {
ReviewDecision::ApprovedForSession
}
(ApprovalDecision::Accept, _) => ReviewDecision::Approved,
(ApprovalDecision::Decline, _) => ReviewDecision::Denied,
(ApprovalDecision::Cancel, _) => ReviewDecision::Abort,
};
if let Err(err) = conversation
.submit(Op::ExecApproval {
id: event_id,
decision,
})
.await
{
error!("failed to submit ExecApproval: {err}");
}
}
/// similar to handle_mcp_tool_call_begin in exec
async fn construct_mcp_tool_call_notification(
begin_event: McpToolCallBeginEvent,
) -> ItemStartedNotification {
let item = ThreadItem::McpToolCall {
id: begin_event.call_id,
server: begin_event.invocation.server,
tool: begin_event.invocation.tool,
status: McpToolCallStatus::InProgress,
arguments: begin_event.invocation.arguments.unwrap_or(JsonValue::Null),
result: None,
error: None,
};
ItemStartedNotification { item }
}
/// simiilar to handle_mcp_tool_call_end in exec
async fn construct_mcp_tool_call_end_notification(
end_event: McpToolCallEndEvent,
) -> ItemCompletedNotification {
let status = if end_event.is_success() {
McpToolCallStatus::Completed
} else {
McpToolCallStatus::Failed
};
let (result, error) = match &end_event.result {
Ok(value) => (
Some(McpToolCallResult {
content: value.content.clone(),
structured_content: value.structured_content.clone(),
}),
None,
),
Err(message) => (
None,
Some(McpToolCallError {
message: message.clone(),
}),
),
};
let item = ThreadItem::McpToolCall {
id: end_event.call_id,
server: end_event.invocation.server,
tool: end_event.invocation.tool,
status,
arguments: end_event.invocation.arguments.unwrap_or(JsonValue::Null),
result,
error,
};
ItemCompletedNotification { item }
}
#[cfg(test)]
mod tests {
use super::*;
use codex_core::protocol::McpInvocation;
use mcp_types::CallToolResult;
use mcp_types::ContentBlock;
use mcp_types::TextContent;
use pretty_assertions::assert_eq;
use serde_json::Value as JsonValue;
use std::time::Duration;
#[tokio::test]
async fn test_construct_mcp_tool_call_begin_notification_with_args() {
let begin_event = McpToolCallBeginEvent {
call_id: "call_123".to_string(),
invocation: McpInvocation {
server: "codex".to_string(),
tool: "list_mcp_resources".to_string(),
arguments: Some(serde_json::json!({"server": ""})),
},
};
let notification = construct_mcp_tool_call_notification(begin_event.clone()).await;
let expected = ItemStartedNotification {
item: ThreadItem::McpToolCall {
id: begin_event.call_id,
server: begin_event.invocation.server,
tool: begin_event.invocation.tool,
status: McpToolCallStatus::InProgress,
arguments: serde_json::json!({"server": ""}),
result: None,
error: None,
},
};
assert_eq!(notification, expected);
}
#[tokio::test]
async fn test_construct_mcp_tool_call_begin_notification_without_args() {
let begin_event = McpToolCallBeginEvent {
call_id: "call_456".to_string(),
invocation: McpInvocation {
server: "codex".to_string(),
tool: "list_mcp_resources".to_string(),
arguments: None,
},
};
let notification = construct_mcp_tool_call_notification(begin_event.clone()).await;
let expected = ItemStartedNotification {
item: ThreadItem::McpToolCall {
id: begin_event.call_id,
server: begin_event.invocation.server,
tool: begin_event.invocation.tool,
status: McpToolCallStatus::InProgress,
arguments: JsonValue::Null,
result: None,
error: None,
},
};
assert_eq!(notification, expected);
}
#[tokio::test]
async fn test_construct_mcp_tool_call_end_notification_success() {
let content = vec![ContentBlock::TextContent(TextContent {
annotations: None,
text: "{\"resources\":[]}".to_string(),
r#type: "text".to_string(),
})];
let result = CallToolResult {
content: content.clone(),
is_error: Some(false),
structured_content: None,
};
let end_event = McpToolCallEndEvent {
call_id: "call_789".to_string(),
invocation: McpInvocation {
server: "codex".to_string(),
tool: "list_mcp_resources".to_string(),
arguments: Some(serde_json::json!({"server": ""})),
},
duration: Duration::from_nanos(92708),
result: Ok(result),
};
let notification = construct_mcp_tool_call_end_notification(end_event.clone()).await;
let expected = ItemCompletedNotification {
item: ThreadItem::McpToolCall {
id: end_event.call_id,
server: end_event.invocation.server,
tool: end_event.invocation.tool,
status: McpToolCallStatus::Completed,
arguments: serde_json::json!({"server": ""}),
result: Some(McpToolCallResult {
content,
structured_content: None,
}),
error: None,
},
};
assert_eq!(notification, expected);
}
#[tokio::test]
async fn test_construct_mcp_tool_call_end_notification_error() {
let end_event = McpToolCallEndEvent {
call_id: "call_err".to_string(),
invocation: McpInvocation {
server: "codex".to_string(),
tool: "list_mcp_resources".to_string(),
arguments: None,
},
duration: Duration::from_millis(1),
result: Err("boom".to_string()),
};
let notification = construct_mcp_tool_call_end_notification(end_event.clone()).await;
let expected = ItemCompletedNotification {
item: ThreadItem::McpToolCall {
id: end_event.call_id,
server: end_event.invocation.server,
tool: end_event.invocation.tool,
status: McpToolCallStatus::Failed,
arguments: JsonValue::Null,
result: None,
error: Some(McpToolCallError {
message: "boom".to_string(),
}),
},
};
assert_eq!(notification, expected);
}
}
codex-rs/app-server/src/codex_message_processor.rs
+9 -207
View File
@@ -1,3 +1,4 @@
use crate::bespoke_event_handling::apply_bespoke_event_handling;
use crate::error_code::INTERNAL_ERROR_CODE;
use crate::error_code::INVALID_REQUEST_ERROR_CODE;
use crate::fuzzy_file_search::run_fuzzy_file_search;
@@ -8,12 +9,9 @@ use chrono::DateTime;
use chrono::Utc;
use codex_app_server_protocol::Account;
use codex_app_server_protocol::AccountLoginCompletedNotification;
use codex_app_server_protocol::AccountRateLimitsUpdatedNotification;
use codex_app_server_protocol::AccountUpdatedNotification;
use codex_app_server_protocol::AddConversationListenerParams;
use codex_app_server_protocol::AddConversationSubscriptionResponse;
use codex_app_server_protocol::ApplyPatchApprovalParams;
use codex_app_server_protocol::ApplyPatchApprovalResponse;
use codex_app_server_protocol::ArchiveConversationParams;
use codex_app_server_protocol::ArchiveConversationResponse;
use codex_app_server_protocol::AskForApproval;
@@ -25,8 +23,6 @@ use codex_app_server_protocol::CancelLoginChatGptResponse;
use codex_app_server_protocol::ClientRequest;
use codex_app_server_protocol::ConversationGitInfo;
use codex_app_server_protocol::ConversationSummary;
use codex_app_server_protocol::ExecCommandApprovalParams;
use codex_app_server_protocol::ExecCommandApprovalResponse;
use codex_app_server_protocol::ExecOneOffCommandParams;
use codex_app_server_protocol::ExecOneOffCommandResponse;
use codex_app_server_protocol::FeedbackUploadParams;
@@ -45,9 +41,6 @@ use codex_app_server_protocol::GetUserSavedConfigResponse;
use codex_app_server_protocol::GitDiffToRemoteResponse;
use codex_app_server_protocol::InputItem as WireInputItem;
use codex_app_server_protocol::InterruptConversationParams;
use codex_app_server_protocol::InterruptConversationResponse;
use codex_app_server_protocol::ItemCompletedNotification;
use codex_app_server_protocol::ItemStartedNotification;
use codex_app_server_protocol::JSONRPCErrorError;
use codex_app_server_protocol::ListConversationsParams;
use codex_app_server_protocol::ListConversationsResponse;
@@ -65,7 +58,6 @@ use codex_app_server_protocol::NewConversationResponse;
use codex_app_server_protocol::RemoveConversationListenerParams;
use codex_app_server_protocol::RemoveConversationSubscriptionResponse;
use codex_app_server_protocol::RequestId;
use codex_app_server_protocol::Result as JsonRpcResult;
use codex_app_server_protocol::ResumeConversationParams;
use codex_app_server_protocol::ResumeConversationResponse;
use codex_app_server_protocol::SandboxMode;
@@ -74,7 +66,6 @@ use codex_app_server_protocol::SendUserMessageResponse;
use codex_app_server_protocol::SendUserTurnParams;
use codex_app_server_protocol::SendUserTurnResponse;
use codex_app_server_protocol::ServerNotification;
use codex_app_server_protocol::ServerRequestPayload;
use codex_app_server_protocol::SessionConfiguredNotification;
use codex_app_server_protocol::SetDefaultModelParams;
use codex_app_server_protocol::SetDefaultModelResponse;
@@ -91,7 +82,6 @@ use codex_app_server_protocol::ThreadStartResponse;
use codex_app_server_protocol::ThreadStartedNotification;
use codex_app_server_protocol::Turn;
use codex_app_server_protocol::TurnInterruptParams;
use codex_app_server_protocol::TurnInterruptResponse;
use codex_app_server_protocol::TurnStartParams;
use codex_app_server_protocol::TurnStartResponse;
use codex_app_server_protocol::TurnStartedNotification;
@@ -123,12 +113,8 @@ use codex_core::find_conversation_path_by_id_str;
use codex_core::get_platform_sandbox;
use codex_core::git_info::git_diff_to_remote;
use codex_core::parse_cursor;
use codex_core::protocol::ApplyPatchApprovalRequestEvent;
use codex_core::protocol::Event;
use codex_core::protocol::EventMsg;
use codex_core::protocol::ExecApprovalRequestEvent;
use codex_core::protocol::Op;
use codex_core::protocol::ReviewDecision;
use codex_core::read_head_for_summary;
use codex_feedback::CodexFeedback;
use codex_login::ServerOptions as LoginServerOptions;
@@ -163,7 +149,7 @@ use tracing::warn;
use uuid::Uuid;
type PendingInterruptQueue = Vec<(RequestId, ApiVersion)>;
type PendingInterrupts = Arc<Mutex<HashMap<ConversationId, PendingInterruptQueue>>>;
pub(crate) type PendingInterrupts = Arc<Mutex<HashMap<ConversationId, PendingInterruptQueue>>>;
// Duration before a ChatGPT login attempt is abandoned.
const LOGIN_CHATGPT_TIMEOUT: Duration = Duration::from_secs(10 * 60);
@@ -194,7 +180,7 @@ pub(crate) struct CodexMessageProcessor {
}
#[derive(Clone, Copy, Debug)]
enum ApiVersion {
pub(crate) enum ApiVersion {
V1,
V2,
}
@@ -1259,7 +1245,7 @@ impl CodexMessageProcessor {
// Auto-attach a conversation listener when starting a thread.
// Use the same behavior as the v1 API with experimental_raw_events=false.
if let Err(err) = self
.attach_conversation_listener(conversation_id, false)
.attach_conversation_listener(conversation_id, false, ApiVersion::V2)
.await
{
tracing::warn!(
@@ -1537,7 +1523,7 @@ impl CodexMessageProcessor {
}) => {
// Auto-attach a conversation listener when resuming a thread.
if let Err(err) = self
.attach_conversation_listener(conversation_id, false)
.attach_conversation_listener(conversation_id, false, ApiVersion::V2)
.await
{
tracing::warn!(
@@ -2390,7 +2376,7 @@ impl CodexMessageProcessor {
experimental_raw_events,
} = params;
match self
.attach_conversation_listener(conversation_id, experimental_raw_events)
.attach_conversation_listener(conversation_id, experimental_raw_events, ApiVersion::V1)
.await
{
Ok(subscription_id) => {
@@ -2431,6 +2417,7 @@ impl CodexMessageProcessor {
&mut self,
conversation_id: ConversationId,
experimental_raw_events: bool,
api_version: ApiVersion,
) -> Result<Uuid, JSONRPCErrorError> {
let conversation = match self
.conversation_manager
@@ -2454,6 +2441,7 @@ impl CodexMessageProcessor {
let outgoing_for_task = self.outgoing.clone();
let pending_interrupts = self.pending_interrupts.clone();
let api_version_for_task = api_version;
tokio::spawn(async move {
loop {
tokio::select! {
@@ -2509,6 +2497,7 @@ impl CodexMessageProcessor {
conversation.clone(),
outgoing_for_task.clone(),
pending_interrupts.clone(),
api_version_for_task,
)
.await;
}
@@ -2651,115 +2640,6 @@ impl CodexMessageProcessor {
}
}
async fn apply_bespoke_event_handling(
event: Event,
conversation_id: ConversationId,
conversation: Arc<CodexConversation>,
outgoing: Arc<OutgoingMessageSender>,
pending_interrupts: PendingInterrupts,
) {
let Event { id: event_id, msg } = event;
match msg {
EventMsg::ApplyPatchApprovalRequest(ApplyPatchApprovalRequestEvent {
call_id,
changes,
reason,
grant_root,
}) => {
let params = ApplyPatchApprovalParams {
conversation_id,
call_id,
file_changes: changes,
reason,
grant_root,
};
let rx = outgoing
.send_request(ServerRequestPayload::ApplyPatchApproval(params))
.await;
// TODO(mbolin): Enforce a timeout so this task does not live indefinitely?
tokio::spawn(async move {
on_patch_approval_response(event_id, rx, conversation).await;
});
}
EventMsg::ExecApprovalRequest(ExecApprovalRequestEvent {
call_id,
command,
cwd,
reason,
risk,
parsed_cmd,
}) => {
let params = ExecCommandApprovalParams {
conversation_id,
call_id,
command,
cwd,
reason,
risk,
parsed_cmd,
};
let rx = outgoing
.send_request(ServerRequestPayload::ExecCommandApproval(params))
.await;
// TODO(mbolin): Enforce a timeout so this task does not live indefinitely?
tokio::spawn(async move {
on_exec_approval_response(event_id, rx, conversation).await;
});
}
EventMsg::TokenCount(token_count_event) => {
if let Some(rate_limits) = token_count_event.rate_limits {
outgoing
.send_server_notification(ServerNotification::AccountRateLimitsUpdated(
AccountRateLimitsUpdatedNotification {
rate_limits: rate_limits.into(),
},
))
.await;
}
}
EventMsg::ItemStarted(item_started_event) => {
let item: ThreadItem = item_started_event.item.clone().into();
let notification = ItemStartedNotification { item };
outgoing
.send_server_notification(ServerNotification::ItemStarted(notification))
.await;
}
EventMsg::ItemCompleted(item_completed_event) => {
let item: ThreadItem = item_completed_event.item.clone().into();
let notification = ItemCompletedNotification { item };
outgoing
.send_server_notification(ServerNotification::ItemCompleted(notification))
.await;
}
// If this is a TurnAborted, reply to any pending interrupt requests.
EventMsg::TurnAborted(turn_aborted_event) => {
let pending = {
let mut map = pending_interrupts.lock().await;
map.remove(&conversation_id).unwrap_or_default()
};
if !pending.is_empty() {
for (rid, ver) in pending {
match ver {
ApiVersion::V1 => {
let response = InterruptConversationResponse {
abort_reason: turn_aborted_event.reason.clone(),
};
outgoing.send_response(rid, response).await;
}
ApiVersion::V2 => {
let response = TurnInterruptResponse {};
outgoing.send_response(rid, response).await;
}
}
}
}
}
_ => {}
}
}
async fn derive_config_from_params(
overrides: ConfigOverrides,
cli_overrides: Option<std::collections::HashMap<String, serde_json::Value>>,
@@ -2773,84 +2653,6 @@ async fn derive_config_from_params(
Config::load_with_cli_overrides(cli_overrides, overrides).await
}
async fn on_patch_approval_response(
event_id: String,
receiver: oneshot::Receiver<JsonRpcResult>,
codex: Arc<CodexConversation>,
) {
let response = receiver.await;
let value = match response {
Ok(value) => value,
Err(err) => {
error!("request failed: {err:?}");
if let Err(submit_err) = codex
.submit(Op::PatchApproval {
id: event_id.clone(),
decision: ReviewDecision::Denied,
})
.await
{
error!("failed to submit denied PatchApproval after request failure: {submit_err}");
}
return;
}
};
let response =
serde_json::from_value::<ApplyPatchApprovalResponse>(value).unwrap_or_else(|err| {
error!("failed to deserialize ApplyPatchApprovalResponse: {err}");
ApplyPatchApprovalResponse {
decision: ReviewDecision::Denied,
}
});
if let Err(err) = codex
.submit(Op::PatchApproval {
id: event_id,
decision: response.decision,
})
.await
{
error!("failed to submit PatchApproval: {err}");
}
}
async fn on_exec_approval_response(
event_id: String,
receiver: oneshot::Receiver<JsonRpcResult>,
conversation: Arc<CodexConversation>,
) {
let response = receiver.await;
let value = match response {
Ok(value) => value,
Err(err) => {
error!("request failed: {err:?}");
return;
}
};
// Try to deserialize `value` and then make the appropriate call to `codex`.
let response =
serde_json::from_value::<ExecCommandApprovalResponse>(value).unwrap_or_else(|err| {
error!("failed to deserialize ExecCommandApprovalResponse: {err}");
// If we cannot deserialize the response, we deny the request to be
// conservative.
ExecCommandApprovalResponse {
decision: ReviewDecision::Denied,
}
});
if let Err(err) = conversation
.submit(Op::ExecApproval {
id: event_id,
decision: response.decision,
})
.await
{
error!("failed to submit ExecApproval: {err}");
}
}
async fn read_summary_from_rollout(
path: &Path,
fallback_provider: &str,
codex-rs/app-server/src/lib.rs
+1
View File
@@ -28,6 +28,7 @@ use tracing_subscriber::filter::Targets;
use tracing_subscriber::layer::SubscriberExt;
use tracing_subscriber::util::SubscriberInitExt;
mod bespoke_event_handling;
mod codex_message_processor;
mod error_code;
mod fuzzy_file_search;
codex-rs/app-server/tests/suite/config.rs
+2 -2
View File
@@ -27,7 +27,7 @@ fn create_config_toml(codex_home: &Path) -> std::io::Result<()> {
std::fs::write(
config_toml,
r#"
model = "gpt-5-codex"
model = "gpt-5.1-codex"
approval_policy = "on-request"
sandbox_mode = "workspace-write"
model_reasoning_summary = "detailed"
@@ -87,7 +87,7 @@ async fn get_config_toml_parses_all_fields() -> Result<()> {
}),
forced_chatgpt_workspace_id: Some("12345678-0000-0000-0000-000000000000".into()),
forced_login_method: Some(ForcedLoginMethod::Chatgpt),
model: Some("gpt-5-codex".into()),
model: Some("gpt-5.1-codex".into()),
model_reasoning_effort: Some(ReasoningEffort::High),
model_reasoning_summary: Some(ReasoningSummary::Detailed),
model_verbosity: Some(Verbosity::Medium),
codex-rs/app-server/tests/suite/set_default_model.rs
+1 -1
View File
@@ -57,7 +57,7 @@ fn create_config_toml(codex_home: &Path) -> std::io::Result<()> {
std::fs::write(
config_toml,
r#"
model = "gpt-5-codex"
model = "gpt-5.1-codex"
model_reasoning_effort = "medium"
"#,
)
codex-rs/app-server/tests/suite/v2/thread_list.rs
+7 -42
View File
@@ -6,10 +6,8 @@ use codex_app_server_protocol::JSONRPCResponse;
use codex_app_server_protocol::RequestId;
use codex_app_server_protocol::ThreadListParams;
use codex_app_server_protocol::ThreadListResponse;
use serde_json::json;
use tempfile::TempDir;
use tokio::time::timeout;
use uuid::Uuid;
const DEFAULT_READ_TIMEOUT: std::time::Duration = std::time::Duration::from_secs(10);
@@ -150,46 +148,13 @@ async fn thread_list_respects_provider_filter() -> Result<()> {
"X",
Some("mock_provider"),
)?; // mock_provider
// one with a different provider
let uuid = Uuid::new_v4();
let dir = codex_home
.path()
.join("sessions")
.join("2025")
.join("01")
.join("02");
std::fs::create_dir_all(&dir)?;
let file_path = dir.join(format!("rollout-2025-01-02T11-00-00-{uuid}.jsonl"));
let lines = [
json!({
"timestamp": "2025-01-02T11:00:00Z",
"type": "session_meta",
"payload": {
"id": uuid,
"timestamp": "2025-01-02T11:00:00Z",
"cwd": "/",
"originator": "codex",
"cli_version": "0.0.0",
"instructions": null,
"source": "vscode",
"model_provider": "other_provider"
}
})
.to_string(),
json!({
"timestamp": "2025-01-02T11:00:00Z",
"type":"response_item",
"payload": {"type":"message","role":"user","content":[{"type":"input_text","text":"X"}]}
})
.to_string(),
json!({
"timestamp": "2025-01-02T11:00:00Z",
"type":"event_msg",
"payload": {"type":"user_message","message":"X","kind":"plain"}
})
.to_string(),
];
std::fs::write(file_path, lines.join("\n") + "\n")?;
let _b = create_fake_rollout(
codex_home.path(),
"2025-01-02T11-00-00",
"2025-01-02T11:00:00Z",
"X",
Some("other_provider"),
)?;
let mut mcp = McpProcess::new(codex_home.path()).await?;
timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
codex-rs/app-server/tests/suite/v2/thread_resume.rs
+3 -3
View File
@@ -27,7 +27,7 @@ async fn thread_resume_returns_original_thread() -> Result<()> {
// Start a thread.
let start_id = mcp
.send_thread_start_request(ThreadStartParams {
model: Some("gpt-5-codex".to_string()),
model: Some("gpt-5.1-codex".to_string()),
..Default::default()
})
.await?;
@@ -68,7 +68,7 @@ async fn thread_resume_prefers_path_over_thread_id() -> Result<()> {
let start_id = mcp
.send_thread_start_request(ThreadStartParams {
model: Some("gpt-5-codex".to_string()),
model: Some("gpt-5.1-codex".to_string()),
..Default::default()
})
.await?;
@@ -112,7 +112,7 @@ async fn thread_resume_supports_history_and_overrides() -> Result<()> {
// Start a thread.
let start_id = mcp
.send_thread_start_request(ThreadStartParams {
model: Some("gpt-5-codex".to_string()),
model: Some("gpt-5.1-codex".to_string()),
..Default::default()
})
.await?;
codex-rs/app-server/tests/suite/v2/thread_start.rs
+1 -1
View File
@@ -29,7 +29,7 @@ async fn thread_start_creates_thread_and_emits_started() -> Result<()> {
// Start a v2 thread with an explicit model override.
let req_id = mcp
.send_thread_start_request(ThreadStartParams {
model: Some("gpt-5".to_string()),
model: Some("gpt-5.1".to_string()),
..Default::default()
})
.await?;
codex-rs/app-server/tests/suite/v2/turn_start.rs
+35 -37
View File
@@ -5,10 +5,13 @@ use app_test_support::create_mock_chat_completions_server;
use app_test_support::create_mock_chat_completions_server_unchecked;
use app_test_support::create_shell_sse_response;
use app_test_support::to_response;
use codex_app_server_protocol::CommandExecutionStatus;
use codex_app_server_protocol::ItemStartedNotification;
use codex_app_server_protocol::JSONRPCNotification;
use codex_app_server_protocol::JSONRPCResponse;
use codex_app_server_protocol::RequestId;
use codex_app_server_protocol::ServerRequest;
use codex_app_server_protocol::ThreadItem;
use codex_app_server_protocol::ThreadStartParams;
use codex_app_server_protocol::ThreadStartResponse;
use codex_app_server_protocol::TurnStartParams;
@@ -17,9 +20,6 @@ use codex_app_server_protocol::TurnStartedNotification;
use codex_app_server_protocol::UserInput as V2UserInput;
use codex_core::protocol_config_types::ReasoningEffort;
use codex_core::protocol_config_types::ReasoningSummary;
use codex_protocol::parse_command::ParsedCommand;
use codex_protocol::protocol::Event;
use codex_protocol::protocol::EventMsg;
use core_test_support::skip_if_no_network;
use pretty_assertions::assert_eq;
use std::path::Path;
@@ -235,7 +235,7 @@ async fn turn_start_exec_approval_toggle_v2() -> Result<()> {
.await??;
let ThreadStartResponse { thread } = to_response::<ThreadStartResponse>(start_resp)?;
// turn/start — expect ExecCommandApproval request from server
// turn/start — expect CommandExecutionRequestApproval request from server
let first_turn_id = mcp
.send_turn_start_request(TurnStartParams {
thread_id: thread.id.clone(),
@@ -258,16 +258,10 @@ async fn turn_start_exec_approval_toggle_v2() -> Result<()> {
mcp.read_stream_until_request_message(),
)
.await??;
let ServerRequest::ExecCommandApproval { request_id, params } = server_req else {
panic!("expected ExecCommandApproval request");
let ServerRequest::CommandExecutionRequestApproval { request_id, params } = server_req else {
panic!("expected CommandExecutionRequestApproval request");
};
assert_eq!(params.call_id, "call1");
assert_eq!(
params.parsed_cmd,
vec![ParsedCommand::Unknown {
cmd: "python3 -c 'print(42)'".to_string()
}]
);
assert_eq!(params.item_id, "call1");
// Approve and wait for task completion
mcp.send_response(
@@ -302,7 +296,7 @@ async fn turn_start_exec_approval_toggle_v2() -> Result<()> {
)
.await??;
// Ensure we do NOT receive an ExecCommandApproval request before task completes
// Ensure we do NOT receive a CommandExecutionRequestApproval request before task completes
timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_notification_message("codex/event/task_complete"),
@@ -314,8 +308,6 @@ async fn turn_start_exec_approval_toggle_v2() -> Result<()> {
#[tokio::test]
async fn turn_start_updates_sandbox_and_cwd_between_turns_v2() -> Result<()> {
// When returning Result from a test, pass an Ok(()) to the skip macro
// so the early return type matches. The no-arg form returns unit.
skip_if_no_network!(Ok(()));
let tmp = TempDir::new()?;
@@ -424,29 +416,35 @@ async fn turn_start_updates_sandbox_and_cwd_between_turns_v2() -> Result<()> {
)
.await??;
let exec_begin_notification = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_notification_message("codex/event/exec_command_begin"),
)
let command_exec_item = timeout(DEFAULT_READ_TIMEOUT, async {
loop {
let item_started_notification = mcp
.read_stream_until_notification_message("item/started")
.await?;
let params = item_started_notification
.params
.clone()
.expect("item/started params");
let item_started: ItemStartedNotification =
serde_json::from_value(params).expect("deserialize item/started notification");
if matches!(item_started.item, ThreadItem::CommandExecution { .. }) {
return Ok::<ThreadItem, anyhow::Error>(item_started.item);
}
}
})
.await??;
let params = exec_begin_notification
.params
.clone()
.expect("exec_command_begin params");
let event: Event = serde_json::from_value(params).expect("deserialize exec begin event");
let exec_begin = match event.msg {
EventMsg::ExecCommandBegin(exec_begin) => exec_begin,
other => panic!("expected ExecCommandBegin event, got {other:?}"),
let ThreadItem::CommandExecution {
cwd,
command,
status,
..
} = command_exec_item
else {
unreachable!("loop ensures we break on command execution items");
};
assert_eq!(exec_begin.cwd, second_cwd);
assert_eq!(
exec_begin.command,
vec![
"bash".to_string(),
"-lc".to_string(),
"echo second turn".to_string()
]
);
assert_eq!(cwd, second_cwd);
assert_eq!(command, "bash -lc 'echo second turn'");
assert_eq!(status, CommandExecutionStatus::InProgress);
timeout(
DEFAULT_READ_TIMEOUT,
codex-rs/cli/src/debug_sandbox.rs
+1 -1
View File
@@ -155,11 +155,11 @@ async fn run_command_under_sandbox(
run_windows_sandbox_capture(
policy_str,
&sandbox_cwd,
base_dir.as_path(),
command_vec,
&cwd_clone,
env_map,
None,
Some(base_dir.as_path()),
)
})
.await;
codex-rs/cli/src/main.rs
+4 -4
View File
@@ -761,9 +761,9 @@ mod tests {
#[test]
fn resume_model_flag_applies_when_no_root_flags() {
let interactive = finalize_from_args(["codex", "resume", "-m", "gpt-5-test"].as_ref());
let interactive = finalize_from_args(["codex", "resume", "-m", "gpt-5.1-test"].as_ref());
assert_eq!(interactive.model.as_deref(), Some("gpt-5-test"));
assert_eq!(interactive.model.as_deref(), Some("gpt-5.1-test"));
assert!(interactive.resume_picker);
assert!(!interactive.resume_last);
assert_eq!(interactive.resume_session_id, None);
@@ -808,7 +808,7 @@ mod tests {
"--ask-for-approval",
"on-request",
"-m",
"gpt-5-test",
"gpt-5.1-test",
"-p",
"my-profile",
"-C",
@@ -819,7 +819,7 @@ mod tests {
.as_ref(),
);
assert_eq!(interactive.model.as_deref(), Some("gpt-5-test"));
assert_eq!(interactive.model.as_deref(), Some("gpt-5.1-test"));
assert!(interactive.oss);
assert_eq!(interactive.config_profile.as_deref(), Some("my-profile"));
assert_matches!(
codex-rs/cloud-tasks/src/lib.rs
+1
View File
@@ -1721,6 +1721,7 @@ mod tests {
use ratatui::layout::Rect;
#[test]
#[ignore = "very slow"]
fn composer_input_renders_typed_characters() {
let mut composer = ComposerInput::new();
let key = KeyEvent::new(KeyCode::Char('a'), KeyModifiers::NONE);
codex-rs/common/Cargo.toml
+2
View File
@@ -10,6 +10,8 @@ workspace = true
clap = { workspace = true, features = ["derive", "wrap_help"], optional = true }
codex-app-server-protocol = { workspace = true }
codex-core = { workspace = true }
codex-lmstudio = { workspace = true }
codex-ollama = { workspace = true }
codex-protocol = { workspace = true }
once_cell = { workspace = true }
serde = { workspace = true, optional = true }
codex-rs/common/src/lib.rs
+2
View File
@@ -37,3 +37,5 @@ pub mod model_presets;
// Shared approval presets (AskForApproval + Sandbox) used by TUI and MCP server
// Not to be confused with AskForApproval, which we should probably rename to EscalationPolicy.
pub mod approval_presets;
// Shared OSS provider utilities used by TUI and exec
pub mod oss;
codex-rs/common/src/oss.rs
+60
View File
@@ -0,0 +1,60 @@
//! OSS provider utilities shared between TUI and exec.
use codex_core::LMSTUDIO_OSS_PROVIDER_ID;
use codex_core::OLLAMA_OSS_PROVIDER_ID;
use codex_core::config::Config;
/// Returns the default model for a given OSS provider.
pub fn get_default_model_for_oss_provider(provider_id: &str) -> Option<&'static str> {
match provider_id {
LMSTUDIO_OSS_PROVIDER_ID => Some(codex_lmstudio::DEFAULT_OSS_MODEL),
OLLAMA_OSS_PROVIDER_ID => Some(codex_ollama::DEFAULT_OSS_MODEL),
_ => None,
}
}
/// Ensures the specified OSS provider is ready (models downloaded, service reachable).
pub async fn ensure_oss_provider_ready(
provider_id: &str,
config: &Config,
) -> Result<(), std::io::Error> {
match provider_id {
LMSTUDIO_OSS_PROVIDER_ID => {
codex_lmstudio::ensure_oss_ready(config)
.await
.map_err(|e| std::io::Error::other(format!("OSS setup failed: {e}")))?;
}
OLLAMA_OSS_PROVIDER_ID => {
codex_ollama::ensure_oss_ready(config)
.await
.map_err(|e| std::io::Error::other(format!("OSS setup failed: {e}")))?;
}
_ => {
// Unknown provider, skip setup
}
}
Ok(())
}
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn test_get_default_model_for_provider_lmstudio() {
let result = get_default_model_for_oss_provider(LMSTUDIO_OSS_PROVIDER_ID);
assert_eq!(result, Some(codex_lmstudio::DEFAULT_OSS_MODEL));
}
#[test]
fn test_get_default_model_for_provider_ollama() {
let result = get_default_model_for_oss_provider(OLLAMA_OSS_PROVIDER_ID);
assert_eq!(result, Some(codex_ollama::DEFAULT_OSS_MODEL));
}
#[test]
fn test_get_default_model_for_provider_unknown() {
let result = get_default_model_for_oss_provider("unknown-provider");
assert_eq!(result, None);
}
}
codex-rs/core/Cargo.toml
+10 -6
View File
@@ -40,12 +40,7 @@ eventsource-stream = { workspace = true }
futures = { workspace = true }
http = { workspace = true }
indexmap = { workspace = true }
keyring = { workspace = true, features = [
"apple-native",
"crypto-rust",
"linux-native-async-persistent",
"windows-native",
] }
keyring = { workspace = true, features = ["crypto-rust"] }
libc = { workspace = true }
mcp-types = { workspace = true }
os_info = { workspace = true }
@@ -60,6 +55,7 @@ shlex = { workspace = true }
similar = { workspace = true }
strum_macros = { workspace = true }
tempfile = { workspace = true }
test-case = "3.3.1"
test-log = { workspace = true }
thiserror = { workspace = true }
time = { workspace = true, features = [
@@ -89,9 +85,11 @@ wildmatch = { workspace = true }
[target.'cfg(target_os = "linux")'.dependencies]
landlock = { workspace = true }
seccompiler = { workspace = true }
keyring = { workspace = true, features = ["linux-native-async-persistent"] }
[target.'cfg(target_os = "macos")'.dependencies]
core-foundation = "0.9"
keyring = { workspace = true, features = ["apple-native"] }
# Build OpenSSL from source for musl builds.
[target.x86_64-unknown-linux-musl.dependencies]
@@ -101,6 +99,12 @@ openssl-sys = { workspace = true, features = ["vendored"] }
[target.aarch64-unknown-linux-musl.dependencies]
openssl-sys = { workspace = true, features = ["vendored"] }
[target.'cfg(target_os = "windows")'.dependencies]
keyring = { workspace = true, features = ["windows-native"] }
[target.'cfg(any(target_os = "freebsd", target_os = "openbsd"))'.dependencies]
keyring = { workspace = true, features = ["sync-secret-service"] }
[dev-dependencies]
assert_cmd = { workspace = true }
assert_matches = { workspace = true }
codex-rs/core/gpt_5_1_prompt.md
+94 -36
View File
@@ -1,4 +1,4 @@
You are a coding agent running in the Codex CLI, a terminal-based coding assistant. Codex CLI is an open source project led by OpenAI. You are expected to be precise, safe, and helpful.
You are GPT-5.1 running in the Codex CLI, a terminal-based coding assistant. Codex CLI is an open source project led by OpenAI. You are expected to be precise, safe, and helpful.
Your capabilities:
@@ -26,17 +26,28 @@ Your default personality and tone is concise, direct, and friendly. You communic
- Direct system/developer/user instructions (as part of a prompt) take precedence over AGENTS.md instructions.
- The contents of the AGENTS.md file at the root of the repo and any directories from the CWD up to the root are included with the developer message and don't need to be re-read. When working in a subdirectory of CWD, or a directory outside the CWD, check for any AGENTS.md files that may be applicable.
## Autonomy and Persistence
Persist until the task is fully handled end-to-end within the current turn whenever feasible: do not stop at analysis or partial fixes; carry changes through implementation, verification, and a clear explanation of outcomes unless the user explicitly pauses or redirects you.
Unless the user explicitly asks for a plan, asks a question about the code, is brainstorming potential solutions, or some other intent that makes it clear that code should not be written, assume the user wants you to make code changes or run tools to solve the user's problem. In these cases, it's bad to output your proposed solution in a message, you should go ahead and actually implement the change. If you encounter challenges or blockers, you should attempt to resolve them yourself.
## Responsiveness
### Preamble messages
### User Updates Spec
You'll work for stretches with tool calls — it's critical to keep the user updated as you work.
Before making tool calls, send a brief preamble to the user explaining what you’re about to do. When sending preamble messages, follow these principles and examples:
Frequency & Length:
- Send short updates (1–2 sentences) whenever there is a meaningful, important insight you need to share with the user to keep them informed.
- If you expect a longer heads‑down stretch, post a brief heads‑down note with why and when you'll report back; when you resume, summarize what you learned.
- Only the initial plan, plan updates, and final recap can be longer, with multiple bullets and paragraphs
- **Logically group related actions**: if you’re about to run several related commands, describe them together in one preamble rather than sending a separate note for each.
- **Keep it concise**: be no more than 1-2 sentences, focused on immediate, tangible next steps. (8–12 words for quick updates).
- **Build on prior context**: if this is not your first tool call, use the preamble message to connect the dots with what’s been done so far and create a sense of momentum and clarity for the user to understand your next actions.
- **Keep your tone light, friendly and curious**: add small touches of personality in preambles feel collaborative and engaging.
- **Exception**: Avoid adding a preamble for every trivial read (e.g., `cat` a single file) unless it’s part of a larger grouped action.
Tone:
- Friendly, confident, senior-engineer energy. Positive, collaborative, humble; fix mistakes quickly.
Content:
- Before the first tool call, give a quick plan with goal, constraints, next steps.
- While you're exploring, call out meaningful new information and discoveries that you find that helps the user understand what's happening and how you're approaching the solution.
- If you change the plan (e.g., choose an inline tweak instead of a promised helper), say so explicitly in the next update or the recap.
**Examples:**
@@ -59,6 +70,8 @@ Do not repeat the full contents of the plan after an `update_plan` call — the
Before running a command, consider whether or not you have completed the previous step, and make sure to mark it as completed before moving on to the next step. It may be the case that you complete all steps in your plan after a single pass of implementation. If this is the case, you can simply mark all the planned steps as completed. Sometimes, you may need to change plans in the middle of a task: call `update_plan` with the updated plan and make sure to provide an `explanation` of the rationale when doing so.
Maintain statuses in the tool: exactly one item in_progress at a time; mark items complete when done; post timely status transitions. Do not jump an item from pending to completed: always set it to in_progress first. Do not batch-complete multiple items after the fact. Finish with all items completed or explicitly canceled/deferred before ending the turn. Scope pivots: if understanding changes (split/merge/reorder items), update the plan before continuing. Do not let the plan go stale while coding.
Use a plan when:
- The task is non-trivial and will require multiple actions over a long time horizon.
@@ -122,14 +135,14 @@ If you need to write a plan, only write high quality plans, not low quality ones
## Task execution
You are a coding agent. Please keep going until the query is completely resolved, before ending your turn and yielding back to the user. Only terminate your turn when you are sure that the problem is solved. Autonomously resolve the query to the best of your ability, using the tools available to you, before coming back to the user. Do NOT guess or make up an answer.
You are a coding agent. You must keep going until the query or task is completely resolved, before ending your turn and yielding back to the user. Persist until the task is fully handled end-to-end within the current turn whenever feasible and persevere even when function calls fail. Only terminate your turn when you are sure that the problem is solved. Autonomously resolve the query to the best of your ability, using the tools available to you, before coming back to the user. Do NOT guess or make up an answer.
You MUST adhere to the following criteria when solving queries:
- Working on the repo(s) in the current environment is allowed, even if they are proprietary.
- Analyzing code for vulnerabilities is allowed.
- Showing user code and tool call details is allowed.
- Use the `apply_patch` tool to edit files (NEVER try `applypatch` or `apply-patch`, only `apply_patch`): {"command":["apply_patch","*** Begin Patch\\n*** Update File: path/to/file.py\\n@@ def example():\\n- pass\\n+ return 123\\n*** End Patch"]}
- Use the `apply_patch` tool to edit files (NEVER try `applypatch` or `apply-patch`, only `apply_patch`). This is a FREEFORM tool, so do not wrap the patch in JSON.
If completing the user's task requires writing or modifying files, your code and final answer should follow these coding guidelines, though user instructions (i.e. AGENTS.md) may override these guidelines:
@@ -146,44 +159,46 @@ If completing the user's task requires writing or modifying files, your code and
- Do not use one-letter variable names unless explicitly requested.
- NEVER output inline citations like "【F:README.md†L5-L14】" in your outputs. The CLI is not able to render these so they will just be broken in the UI. Instead, if you output valid filepaths, users will be able to click on them to open the files in their editor.
## Sandbox and approvals
## Codex CLI harness, sandboxing, and approvals
The Codex CLI harness supports several different sandboxing, and approval configurations that the user can choose from.
The Codex CLI harness supports several different configurations for sandboxing and escalation approvals that the user can choose from.
Filesystem sandboxing prevents you from editing files without user approval. The options are:
Filesystem sandboxing defines which files can be read or written. The options for `sandbox_mode` are:
- **read-only**: The sandbox only permits reading files.
- **workspace-write**: The sandbox permits reading files, and editing files in `cwd` and `writable_roots`. Editing files in other directories requires approval.
- **danger-full-access**: No filesystem sandboxing - all commands are permitted.
- **read-only**: You can only read files.
- **workspace-write**: You can read files. You can write to files in your workspace folder, but not outside it.
- **danger-full-access**: No filesystem sandboxing.
Network sandboxing prevents you from accessing network without approval. Options are
- **restricted**
- **enabled**
Approvals are your mechanism to get user consent to perform more privileged actions. Although they introduce friction to the user because your work is paused until the user responds, you should leverage them to accomplish your important work. Do not let these settings or the sandbox deter you from attempting to accomplish the user's task. Approval options are
Network sandboxing defines whether network can be accessed without approval. Options for `network_access` are:
- **restricted**: Requires approval
- **enabled**: No approval needed
Approvals are your mechanism to get user consent to run shell commands without the sandbox. Possible configuration options for `approval_policy` are
- **untrusted**: The harness will escalate most commands for user approval, apart from a limited allowlist of safe "read" commands.
- **on-failure**: The harness will allow all commands to run in the sandbox (if enabled), and failures will be escalated to the user for approval to run again without the sandbox.
- **on-request**: Commands will be run in the sandbox by default, and you can specify in your tool call if you want to escalate a command to run without sandboxing. (Note that this mode is not always available. If it is, you'll see parameters for it in the `shell` command description.)
- **never**: This is a non-interactive mode where you may NEVER ask the user for approval to run commands. Instead, you must always persist and work around constraints to solve the task for the user. You MUST do your utmost best to finish the task and validate your work before yielding. If this mode is pared with `danger-full-access`, take advantage of it to deliver the best outcome for the user. Further, in this mode, your default testing philosophy is overridden: Even if you don't see local patterns for testing, you may add tests and scripts to validate your work. Just remove them before yielding.
- **on-request**: Commands will be run in the sandbox by default, and you can specify in your tool call if you want to escalate a command to run without sandboxing. (Note that this mode is not always available. If it is, you'll see parameters for escalating in the tool definition.)
- **never**: This is a non-interactive mode where you may NEVER ask the user for approval to run commands. Instead, you must always persist and work around constraints to solve the task for the user. You MUST do your utmost best to finish the task and validate your work before yielding. If this mode is paired with `danger-full-access`, take advantage of it to deliver the best outcome for the user. Further, in this mode, your default testing philosophy is overridden: Even if you don't see local patterns for testing, you may add tests and scripts to validate your work. Just remove them before yielding.
When you are running with approvals `on-request`, and sandboxing enabled, here are scenarios where you'll need to request approval:
- You need to run a command that writes to a directory that requires it (e.g. running tests that write to /tmp)
When you are running with `approval_policy == on-request`, and sandboxing enabled, here are scenarios where you'll need to request approval:
- You need to run a command that writes to a directory that requires it (e.g. running tests that write to /var)
- You need to run a GUI app (e.g., open/xdg-open/osascript) to open browsers or files.
- You are running sandboxed and need to run a command that requires network access (e.g. installing packages)
- If you run a command that is important to solving the user's query, but it fails because of sandboxing, rerun the command with approval.
- If you run a command that is important to solving the user's query, but it fails because of sandboxing, rerun the command with approval. ALWAYS proceed to use the `with_escalated_permissions` and `justification` parameters. Within this harness, prefer requesting approval via the tool over asking in natural language.
- You are about to take a potentially destructive action such as an `rm` or `git reset` that the user did not explicitly ask for
- (For all of these, you should weigh alternative paths that do not require approval.)
- (for all of these, you should weigh alternative paths that do not require approval)
Note that when sandboxing is set to read-only, you'll need to request approval for any command that isn't a read.
When `sandbox_mode` is set to read-only, you'll need to request approval for any command that isn't a read.
You will be told what filesystem sandboxing, network sandboxing, and approval mode are active in a developer or user message. If you are not told about this, assume that you are running with workspace-write, network sandboxing ON, and approval on-failure.
You will be told what filesystem sandboxing, network sandboxing, and approval mode are active in a developer or user message. If you are not told about this, assume that you are running with workspace-write, network sandboxing enabled, and approval on-failure.
Although they introduce friction to the user because your work is paused until the user responds, you should leverage them when necessary to accomplish important work. If the completing the task requires escalated permissions, Do not let these settings or the sandbox deter you from attempting to accomplish the user's task unless it is set to "never", in which case never ask for approvals.
When requesting approval to execute a command that will require escalated privileges:
- Provide the `with_escalated_permissions` parameter with the boolean value true
- Include a short, 1 sentence explanation for why you need to enable `with_escalated_permissions` in the justification parameter
## Validating your work
If the codebase has tests or the ability to build or run, consider using them to verify that your work is complete.
If the codebase has tests or the ability to build or run, consider using them to verify changes once your work is complete.
When testing, your philosophy should be to start as specific as possible to the code you changed so that you can catch issues efficiently, then make your way to broader tests as you build confidence. If there's no test for the code you changed, and if the adjacent patterns in the codebases show that there's a logical place for you to add a test, you may do so. However, do not add tests to codebases with no tests.
@@ -193,7 +208,7 @@ For all of testing, running, building, and formatting, do not attempt to fix unr
Be mindful of whether to run validation commands proactively. In the absence of behavioral guidance:
- When running in non-interactive approval modes like **never** or **on-failure**, proactively run tests, lint and do whatever you need to ensure you've completed the task.
- When running in non-interactive approval modes like **never** or **on-failure**, you can proactively run tests, lint and do whatever you need to ensure you've completed the task. If you are unable to run tests, you must still do your utmost best to complete the task.
- When working in interactive approval modes like **untrusted**, or **on-request**, hold off on running tests or lint commands until the user is ready for you to finalize your output, because these commands take time to run and slow down iteration. Instead suggest what you want to do next, and let the user confirm first.
- When working on test-related tasks, such as adding tests, fixing tests, or reproducing a bug to verify behavior, you may proactively run tests regardless of approval mode. Use your judgement to decide whether this is a test-related task.
@@ -219,7 +234,7 @@ Your final message should read naturally, like an update from a concise teammate
You can skip heavy formatting for single, simple actions or confirmations. In these cases, respond in plain sentences with any relevant next step or quick option. Reserve multi-section structured responses for results that need grouping or explanation.
The user is working on the same computer as you, and has access to your work. As such there's no need to show the full contents of large files you have already written unless the user explicitly asks for them. Similarly, if you've created or modified files using `apply_patch`, there's no need to tell users to "save the file" or "copy the code into a file"—just reference the file path.
The user is working on the same computer as you, and has access to your work. As such there's no need to show the contents of files you have already written unless the user explicitly asks for them. Similarly, if you've created or modified files using `apply_patch`, there's no need to tell users to "save the file" or "copy the code into a file"—just reference the file path.
If there's something that you think you could help with as a logical next step, concisely ask the user if they want you to do so. Good examples of this are running tests, committing changes, or building out the next logical component. If there’s something that you couldn't do (even with approval) but that the user might want to do (such as verifying changes by running the app), include those instructions succinctly.
@@ -247,7 +262,7 @@ You are producing plain text that will later be styled by the CLI. Follow these
**Monospace**
- Wrap all commands, file paths, env vars, and code identifiers in backticks (`` `...` ``).
- Wrap all commands, file paths, env vars, code identifiers, and code samples in backticks (`` `...` ``).
- Apply to inline examples and to bullet keywords if the keyword itself is a literal file/command.
- Never mix monospace and bold markers; choose one based on whether it’s a keyword (`**`) or inline code/path (`` ` ``).
@@ -278,6 +293,13 @@ When referencing files in your response, make sure to include the relevant start
- Keep descriptions self-contained; don’t refer to “above” or “below”.
- Use parallel structure in lists for consistency.
**Verbosity**
- Final answer compactness rules (enforced):
- Tiny/small single-file change (≤ ~10 lines): 2–5 sentences or ≤3 bullets. No headings. 0–1 short snippet (≤3 lines) only if essential.
- Medium change (single area or a few files): ≤6 bullets or 6–10 sentences. At most 1–2 short snippets total (≤8 lines each).
- Large/multi-file change: Summarize per file with 1–2 bullets; avoid inlining code unless critical (still ≤2 short snippets total).
- Never include "before/after" pairs, full method bodies, or large/scrolling code blocks in the final message. Prefer referencing file/symbol names instead.
**Don’t**
- Don’t use literal words “bold” or “monospace” in the content.
@@ -299,6 +321,42 @@ When using the shell, you must adhere to the following guidelines:
- When searching for text or files, prefer using `rg` or `rg --files` respectively because `rg` is much faster than alternatives like `grep`. (If the `rg` command is not found, then use alternatives.)
- Read files in chunks with a max chunk size of 250 lines. Do not use python scripts to attempt to output larger chunks of a file. Command line output will be truncated after 10 kilobytes or 256 lines of output, regardless of the command used.
## apply_patch
Use the `apply_patch` tool to edit files. Your patch language is a stripped‑down, file‑oriented diff format designed to be easy to parse and safe to apply. You can think of it as a high‑level envelope:
*** Begin Patch
[ one or more file sections ]
*** End Patch
Within that envelope, you get a sequence of file operations.
You MUST include a header to specify the action you are taking.
Each operation starts with one of three headers:
*** Add File: <path> - create a new file. Every following line is a + line (the initial contents).
*** Delete File: <path> - remove an existing file. Nothing follows.
*** Update File: <path> - patch an existing file in place (optionally with a rename).
Example patch:
```
*** Begin Patch
*** Add File: hello.txt
+Hello world
*** Update File: src/app.py
*** Move to: src/main.py
@@ def greet():
-print("Hi")
+print("Hello, world!")
*** Delete File: obsolete.txt
*** End Patch
```
It is important to remember:
- You must include a header with your intended action (Add/Delete/Update)
- You must prefix new lines with `+` even when creating a new file
## `update_plan`
A tool named `update_plan` is available to you. You can use it to keep an up‑to‑date, step‑by‑step plan for the task.
codex-rs/core/gpt_5_codex_prompt.md
-2
View File
@@ -2,8 +2,6 @@ You are Codex, based on GPT-5. You are running as a coding agent in the Codex CL
## General
- The arguments to `shell` will be passed to execvp(). Most terminal commands should be prefixed with ["bash", "-lc"].
- Always set the `workdir` param when using the shell function. Do not use `cd` unless absolutely necessary.
- When searching for text or files, prefer using `rg` or `rg --files` respectively because `rg` is much faster than alternatives like `grep`. (If the `rg` command is not found, then use alternatives.)
## Editing constraints
codex-rs/core/src/bash.rs
+11 -13
View File
@@ -1,8 +1,13 @@
use std::path::PathBuf;
use tree_sitter::Node;
use tree_sitter::Parser;
use tree_sitter::Tree;
use tree_sitter_bash::LANGUAGE as BASH;
use crate::shell::ShellType;
use crate::shell::detect_shell_type;
/// Parse the provided bash source using tree-sitter-bash, returning a Tree on
/// success or None if parsing failed.
pub fn try_parse_shell(shell_lc_arg: &str) -> Option<Tree> {
@@ -88,23 +93,16 @@ pub fn try_parse_word_only_commands_sequence(tree: &Tree, src: &str) -> Option<V
Some(commands)
}
pub fn is_well_known_sh_shell(shell: &str) -> bool {
if shell == "bash" || shell == "zsh" {
return true;
}
let shell_name = std::path::Path::new(shell)
.file_name()
.and_then(|s| s.to_str())
.unwrap_or(shell);
matches!(shell_name, "bash" | "zsh")
}
pub fn extract_bash_command(command: &[String]) -> Option<(&str, &str)> {
let [shell, flag, script] = command else {
return None;
};
if !matches!(flag.as_str(), "-lc" | "-c") || !is_well_known_sh_shell(shell) {
if !matches!(flag.as_str(), "-lc" | "-c")
|| !matches!(
detect_shell_type(&PathBuf::from(shell)),
Some(ShellType::Zsh) | Some(ShellType::Bash)
)
{
return None;
}
Some((shell, script))
codex-rs/core/src/chat_completions.rs
+18 -6
View File
@@ -477,10 +477,14 @@ async fn append_reasoning_text(
..
}) = reasoning_item
{
let content_index = content.len() as i64;
content.push(ReasoningItemContent::ReasoningText { text: text.clone() });
let _ = tx_event
.send(Ok(ResponseEvent::ReasoningContentDelta(text.clone())))
.send(Ok(ResponseEvent::ReasoningContentDelta {
delta: text.clone(),
content_index,
}))
.await;
}
}
@@ -669,7 +673,9 @@ async fn process_chat_sse<S>(
}
// Emit end-of-turn when finish_reason signals completion.
if let Some(finish_reason) = choice.get("finish_reason").and_then(|v| v.as_str()) {
if let Some(finish_reason) = choice.get("finish_reason").and_then(|v| v.as_str())
&& !finish_reason.is_empty()
{
match finish_reason {
"tool_calls" if fn_call_state.active => {
// First, flush the terminal raw reasoning so UIs can finalize
@@ -898,20 +904,26 @@ where
continue;
}
}
Poll::Ready(Some(Ok(ResponseEvent::ReasoningContentDelta(delta)))) => {
Poll::Ready(Some(Ok(ResponseEvent::ReasoningContentDelta {
delta,
content_index,
}))) => {
// Always accumulate reasoning deltas so we can emit a final Reasoning item at Completed.
this.cumulative_reasoning.push_str(&delta);
if matches!(this.mode, AggregateMode::Streaming) {
// In streaming mode, also forward the delta immediately.
return Poll::Ready(Some(Ok(ResponseEvent::ReasoningContentDelta(delta))));
return Poll::Ready(Some(Ok(ResponseEvent::ReasoningContentDelta {
delta,
content_index,
})));
} else {
continue;
}
}
Poll::Ready(Some(Ok(ResponseEvent::ReasoningSummaryDelta(_)))) => {
Poll::Ready(Some(Ok(ResponseEvent::ReasoningSummaryDelta { .. }))) => {
continue;
}
Poll::Ready(Some(Ok(ResponseEvent::ReasoningSummaryPartAdded))) => {
Poll::Ready(Some(Ok(ResponseEvent::ReasoningSummaryPartAdded { .. }))) => {
continue;
}
Poll::Ready(Some(Ok(ResponseEvent::OutputItemAdded(item)))) => {
codex-rs/core/src/client.rs
+22 -12
View File
@@ -560,6 +560,8 @@ struct SseEvent {
response: Option<Value>,
item: Option<Value>,
delta: Option<String>,
summary_index: Option<i64>,
content_index: Option<i64>,
}
#[derive(Debug, Deserialize)]
@@ -819,16 +821,22 @@ async fn process_sse<S>(
}
}
"response.reasoning_summary_text.delta" => {
if let Some(delta) = event.delta {
let event = ResponseEvent::ReasoningSummaryDelta(delta);
if let (Some(delta), Some(summary_index)) = (event.delta, event.summary_index) {
let event = ResponseEvent::ReasoningSummaryDelta {
delta,
summary_index,
};
if tx_event.send(Ok(event)).await.is_err() {
return;
}
}
}
"response.reasoning_text.delta" => {
if let Some(delta) = event.delta {
let event = ResponseEvent::ReasoningContentDelta(delta);
if let (Some(delta), Some(content_index)) = (event.delta, event.content_index) {
let event = ResponseEvent::ReasoningContentDelta {
delta,
content_index,
};
if tx_event.send(Ok(event)).await.is_err() {
return;
}
@@ -905,10 +913,12 @@ async fn process_sse<S>(
}
}
"response.reasoning_summary_part.added" => {
// Boundary between reasoning summary sections (e.g., titles).
let event = ResponseEvent::ReasoningSummaryPartAdded;
if tx_event.send(Ok(event)).await.is_err() {
return;
if let Some(summary_index) = event.summary_index {
// Boundary between reasoning summary sections (e.g., titles).
let event = ResponseEvent::ReasoningSummaryPartAdded { summary_index };
if tx_event.send(Ok(event)).await.is_err() {
return;
}
}
}
"response.reasoning_summary_text.done" => {}
@@ -1215,7 +1225,7 @@ mod tests {
#[tokio::test]
async fn error_when_error_event() {
let raw_error = r#"{"type":"response.failed","sequence_number":3,"response":{"id":"resp_689bcf18d7f08194bf3440ba62fe05d803fee0cdac429894","object":"response","created_at":1755041560,"status":"failed","background":false,"error":{"code":"rate_limit_exceeded","message":"Rate limit reached for gpt-5 in organization org-AAA on tokens per min (TPM): Limit 30000, Used 22999, Requested 12528. Please try again in 11.054s. Visit https://platform.openai.com/account/rate-limits to learn more."}, "usage":null,"user":null,"metadata":{}}}"#;
let raw_error = r#"{"type":"response.failed","sequence_number":3,"response":{"id":"resp_689bcf18d7f08194bf3440ba62fe05d803fee0cdac429894","object":"response","created_at":1755041560,"status":"failed","background":false,"error":{"code":"rate_limit_exceeded","message":"Rate limit reached for gpt-5.1 in organization org-AAA on tokens per min (TPM): Limit 30000, Used 22999, Requested 12528. Please try again in 11.054s. Visit https://platform.openai.com/account/rate-limits to learn more."}, "usage":null,"user":null,"metadata":{}}}"#;
let sse1 = format!("event: response.failed\ndata: {raw_error}\n\n");
let provider = ModelProviderInfo {
@@ -1244,7 +1254,7 @@ mod tests {
Err(CodexErr::Stream(msg, delay)) => {
assert_eq!(
msg,
"Rate limit reached for gpt-5 in organization org-AAA on tokens per min (TPM): Limit 30000, Used 22999, Requested 12528. Please try again in 11.054s. Visit https://platform.openai.com/account/rate-limits to learn more."
"Rate limit reached for gpt-5.1 in organization org-AAA on tokens per min (TPM): Limit 30000, Used 22999, Requested 12528. Please try again in 11.054s. Visit https://platform.openai.com/account/rate-limits to learn more."
);
assert_eq!(*delay, Some(Duration::from_secs_f64(11.054)));
}
@@ -1463,7 +1473,7 @@ mod tests {
fn test_try_parse_retry_after() {
let err = Error {
r#type: None,
message: Some("Rate limit reached for gpt-5 in organization org- on tokens per min (TPM): Limit 1, Used 1, Requested 19304. Please try again in 28ms. Visit https://platform.openai.com/account/rate-limits to learn more.".to_string()),
message: Some("Rate limit reached for gpt-5.1 in organization org- on tokens per min (TPM): Limit 1, Used 1, Requested 19304. Please try again in 28ms. Visit https://platform.openai.com/account/rate-limits to learn more.".to_string()),
code: Some("rate_limit_exceeded".to_string()),
plan_type: None,
resets_at: None
@@ -1477,7 +1487,7 @@ mod tests {
fn test_try_parse_retry_after_no_delay() {
let err = Error {
r#type: None,
message: Some("Rate limit reached for gpt-5 in organization <ORG> on tokens per min (TPM): Limit 30000, Used 6899, Requested 24050. Please try again in 1.898s. Visit https://platform.openai.com/account/rate-limits to learn more.".to_string()),
message: Some("Rate limit reached for gpt-5.1 in organization <ORG> on tokens per min (TPM): Limit 30000, Used 6899, Requested 24050. Please try again in 1.898s. Visit https://platform.openai.com/account/rate-limits to learn more.".to_string()),
code: Some("rate_limit_exceeded".to_string()),
plan_type: None,
resets_at: None
codex-rs/core/src/client_common.rs
+16 -8
View File
@@ -136,7 +136,7 @@ fn reserialize_shell_outputs(items: &mut [ResponseItem]) {
}
fn is_shell_tool_name(name: &str) -> bool {
matches!(name, "shell" | "container.exec")
matches!(name, "shell" | "container.exec" | "shell_command")
}
#[derive(Deserialize)]
@@ -203,9 +203,17 @@ pub enum ResponseEvent {
token_usage: Option<TokenUsage>,
},
OutputTextDelta(String),
ReasoningSummaryDelta(String),
ReasoningContentDelta(String),
ReasoningSummaryPartAdded,
ReasoningSummaryDelta {
delta: String,
summary_index: i64,
},
ReasoningContentDelta {
delta: String,
content_index: i64,
},
ReasoningSummaryPartAdded {
summary_index: i64,
},
RateLimits(RateLimitSnapshot),
}
@@ -419,7 +427,7 @@ mod tests {
expects_apply_patch_instructions: false,
},
InstructionsTestCase {
slug: "gpt-5-codex",
slug: "gpt-5.1-codex",
expects_apply_patch_instructions: false,
},
InstructionsTestCase {
@@ -449,7 +457,7 @@ mod tests {
let input: Vec<ResponseItem> = vec![];
let tools: Vec<serde_json::Value> = vec![];
let req = ResponsesApiRequest {
model: "gpt-5",
model: "gpt-5.1",
instructions: "i",
input: &input,
tools: &tools,
@@ -490,7 +498,7 @@ mod tests {
create_text_param_for_request(None, &Some(schema.clone())).expect("text controls");
let req = ResponsesApiRequest {
model: "gpt-5",
model: "gpt-5.1",
instructions: "i",
input: &input,
tools: &tools,
@@ -526,7 +534,7 @@ mod tests {
let input: Vec<ResponseItem> = vec![];
let tools: Vec<serde_json::Value> = vec![];
let req = ResponsesApiRequest {
model: "gpt-5",
model: "gpt-5.1",
instructions: "i",
input: &input,
tools: &tools,
codex-rs/core/src/codex.rs
+143 -232
View File
@@ -9,8 +9,6 @@ use crate::client_common::REVIEW_PROMPT;
use crate::compact;
use crate::features::Feature;
use crate::function_tool::FunctionCallError;
use crate::mcp::auth::McpAuthStatusEntry;
use crate::mcp_connection_manager::DEFAULT_STARTUP_TIMEOUT;
use crate::parse_command::parse_command;
use crate::parse_turn_item;
use crate::response_processing::process_items;
@@ -45,6 +43,7 @@ use mcp_types::ReadResourceResult;
use serde_json;
use serde_json::Value;
use tokio::sync::Mutex;
use tokio::sync::RwLock;
use tokio::sync::oneshot;
use tokio_util::sync::CancellationToken;
use tracing::debug;
@@ -57,7 +56,6 @@ use crate::client::ModelClient;
use crate::client_common::Prompt;
use crate::client_common::ResponseEvent;
use crate::config::Config;
use crate::config::types::McpServerTransportConfig;
use crate::config::types::ShellEnvironmentPolicy;
use crate::context_manager::ContextManager;
use crate::environment_context::EnvironmentContext;
@@ -96,6 +94,7 @@ use crate::protocol::StreamErrorEvent;
use crate::protocol::Submission;
use crate::protocol::TokenCountEvent;
use crate::protocol::TokenUsage;
use crate::protocol::TokenUsageInfo;
use crate::protocol::TurnDiffEvent;
use crate::protocol::WarningEvent;
use crate::rollout::RolloutRecorder;
@@ -132,6 +131,7 @@ use codex_protocol::protocol::InitialHistory;
use codex_protocol::user_input::UserInput;
use codex_utils_readiness::Readiness;
use codex_utils_readiness::ReadinessFlag;
use codex_utils_tokenizer::warm_model_cache;
/// The high-level interface to the Codex system.
/// It operates as a queue pair where you send submissions and receive events.
@@ -474,21 +474,13 @@ impl Session {
),
};
// Error messages to dispatch after SessionConfigured is sent.
let mut post_session_configured_events = Vec::<Event>::new();
// Kick off independent async setup tasks in parallel to reduce startup latency.
//
// - initialize RolloutRecorder with new or resumed session info
// - spin up MCP connection manager
// - perform default shell discovery
// - load history metadata
let rollout_fut = RolloutRecorder::new(&config, rollout_params);
let mcp_fut = McpConnectionManager::new(
config.mcp_servers.clone(),
config.mcp_oauth_credentials_store_mode,
);
let default_shell_fut = shell::default_user_shell();
let history_meta_fut = crate::message_history::history_metadata(&config);
let auth_statuses_fut = compute_auth_statuses(
@@ -497,15 +489,8 @@ impl Session {
);
// Join all independent futures.
let (
rollout_recorder,
mcp_res,
default_shell,
(history_log_id, history_entry_count),
auth_statuses,
) = tokio::join!(
let (rollout_recorder, default_shell, (history_log_id, history_entry_count), auth_statuses) = tokio::join!(
rollout_fut,
mcp_fut,
default_shell_fut,
history_meta_fut,
auth_statuses_fut
@@ -517,34 +502,7 @@ impl Session {
})?;
let rollout_path = rollout_recorder.rollout_path.clone();
// Handle MCP manager result and record any startup failures.
let (mcp_connection_manager, failed_clients) = match mcp_res {
Ok((mgr, failures)) => (mgr, failures),
Err(e) => {
let message = format!("Failed to create MCP connection manager: {e:#}");
error!("{message}");
post_session_configured_events.push(Event {
id: INITIAL_SUBMIT_ID.to_owned(),
msg: EventMsg::Error(ErrorEvent { message }),
});
(McpConnectionManager::default(), Default::default())
}
};
// Surface individual client start-up failures to the user.
if !failed_clients.is_empty() {
for (server_name, err) in failed_clients {
let auth_entry = auth_statuses.get(&server_name);
let display_message = mcp_init_error_display(&server_name, auth_entry, &err);
warn!("MCP client for `{server_name}` failed to start: {err:#}");
post_session_configured_events.push(Event {
id: INITIAL_SUBMIT_ID.to_owned(),
msg: EventMsg::Error(ErrorEvent {
message: display_message,
}),
});
}
}
let mut post_session_configured_events = Vec::<Event>::new();
for (alias, feature) in session_configuration.features.legacy_feature_usages() {
let canonical = feature.key();
@@ -589,8 +547,12 @@ impl Session {
// Create the mutable state for the Session.
let state = SessionState::new(session_configuration.clone());
// Warm the tokenizer cache for the session model without blocking startup.
warm_model_cache(&session_configuration.model);
let services = SessionServices {
mcp_connection_manager,
mcp_connection_manager: Arc::new(RwLock::new(McpConnectionManager::default())),
mcp_startup_cancellation_token: CancellationToken::new(),
unified_exec_manager: UnifiedExecSessionManager::default(),
notifier: UserNotifier::new(config.notify.clone()),
rollout: Mutex::new(Some(rollout_recorder)),
@@ -630,6 +592,18 @@ impl Session {
for event in events {
sess.send_event_raw(event).await;
}
sess.services
.mcp_connection_manager
.write()
.await
.initialize(
config.mcp_servers.clone(),
config.mcp_oauth_credentials_store_mode,
auth_statuses.clone(),
tx_event.clone(),
sess.services.mcp_startup_cancellation_token.clone(),
)
.await;
// record_initial_history can emit events. We record only after the SessionConfiguredEvent is emitted.
sess.record_initial_history(initial_history).await;
@@ -889,6 +863,7 @@ impl Session {
let parsed_cmd = parse_command(&command);
let event = EventMsg::ExecApprovalRequest(ExecApprovalRequestEvent {
call_id,
turn_id: turn_context.sub_id.clone(),
command,
cwd,
reason,
@@ -1082,6 +1057,36 @@ impl Session {
self.send_token_count_event(turn_context).await;
}
pub(crate) async fn override_last_token_usage_estimate(
&self,
turn_context: &TurnContext,
estimated_total_tokens: i64,
) {
{
let mut state = self.state.lock().await;
let mut info = state.token_info().unwrap_or(TokenUsageInfo {
total_token_usage: TokenUsage::default(),
last_token_usage: TokenUsage::default(),
model_context_window: None,
});
info.last_token_usage = TokenUsage {
input_tokens: 0,
cached_input_tokens: 0,
output_tokens: 0,
reasoning_output_tokens: 0,
total_tokens: estimated_total_tokens.max(0),
};
if info.model_context_window.is_none() {
info.model_context_window = turn_context.client.get_model_context_window();
}
state.set_token_info(Some(info));
}
self.send_token_count_event(turn_context).await;
}
pub(crate) async fn update_rate_limits(
&self,
turn_context: &TurnContext,
@@ -1223,6 +1228,8 @@ impl Session {
) -> anyhow::Result<ListResourcesResult> {
self.services
.mcp_connection_manager
.read()
.await
.list_resources(server, params)
.await
}
@@ -1234,6 +1241,8 @@ impl Session {
) -> anyhow::Result<ListResourceTemplatesResult> {
self.services
.mcp_connection_manager
.read()
.await
.list_resource_templates(server, params)
.await
}
@@ -1245,6 +1254,8 @@ impl Session {
) -> anyhow::Result<ReadResourceResult> {
self.services
.mcp_connection_manager
.read()
.await
.read_resource(server, params)
.await
}
@@ -1257,19 +1268,29 @@ impl Session {
) -> anyhow::Result<CallToolResult> {
self.services
.mcp_connection_manager
.read()
.await
.call_tool(server, tool, arguments)
.await
}
pub(crate) fn parse_mcp_tool_name(&self, tool_name: &str) -> Option<(String, String)> {
pub(crate) async fn parse_mcp_tool_name(&self, tool_name: &str) -> Option<(String, String)> {
self.services
.mcp_connection_manager
.read()
.await
.parse_tool_name(tool_name)
.await
}
pub async fn interrupt_task(self: &Arc<Self>) {
info!("interrupt received: abort current task, if any");
self.abort_all_tasks(TurnAbortReason::Interrupted).await;
let has_active_turn = { self.active_turn.lock().await.is_some() };
if has_active_turn {
self.abort_all_tasks(TurnAbortReason::Interrupted).await;
} else {
self.cancel_mcp_startup().await;
}
}
pub(crate) fn notifier(&self) -> &UserNotifier {
@@ -1283,6 +1304,10 @@ impl Session {
fn show_raw_agent_reasoning(&self) -> bool {
self.services.show_raw_agent_reasoning
}
async fn cancel_mcp_startup(&self) {
self.services.mcp_startup_cancellation_token.cancel();
}
}
async fn submission_loop(sess: Arc<Session>, config: Arc<Config>, rx_sub: Receiver<Submission>) {
@@ -1540,17 +1565,15 @@ mod handlers {
}
pub async fn list_mcp_tools(sess: &Session, config: &Arc<Config>, sub_id: String) {
// This is a cheap lookup from the connection manager's cache.
let tools = sess.services.mcp_connection_manager.list_all_tools();
let (auth_status_entries, resources, resource_templates) = tokio::join!(
let mcp_connection_manager = sess.services.mcp_connection_manager.read().await;
let (tools, auth_status_entries, resources, resource_templates) = tokio::join!(
mcp_connection_manager.list_all_tools(),
compute_auth_statuses(
config.mcp_servers.iter(),
config.mcp_oauth_credentials_store_mode,
),
sess.services.mcp_connection_manager.list_all_resources(),
sess.services
.mcp_connection_manager
.list_all_resource_templates()
mcp_connection_manager.list_all_resources(),
mcp_connection_manager.list_all_resource_templates(),
);
let auth_statuses = auth_status_entries
.iter()
@@ -1559,7 +1582,10 @@ mod handlers {
let event = Event {
id: sub_id,
msg: EventMsg::McpListToolsResponse(crate::protocol::McpListToolsResponseEvent {
tools,
tools: tools
.into_iter()
.map(|(name, tool)| (name, tool.tool))
.collect(),
resources,
resource_templates,
auth_statuses,
@@ -1784,7 +1810,6 @@ pub(crate) async fn run_task(
// Although from the perspective of codex.rs, TurnDiffTracker has the lifecycle of a Task which contains
// many turns, from the perspective of the user, it is a single turn.
let turn_diff_tracker = Arc::new(tokio::sync::Mutex::new(TurnDiffTracker::new()));
let mut auto_compact_recently_attempted = false;
loop {
// Note that pending_input would be something like a message the user
@@ -1839,27 +1864,12 @@ pub(crate) async fn run_task(
let (responses, items_to_record_in_conversation_history) =
process_items(processed_items, &sess, &turn_context).await;
// as long as compaction works well in getting us way below the token limit, we shouldn't worry about being in an infinite loop.
if token_limit_reached {
if auto_compact_recently_attempted {
let limit_str = limit.to_string();
let current_tokens = total_usage_tokens
.map(|tokens| tokens.to_string())
.unwrap_or_else(|| "unknown".to_string());
let event = EventMsg::Error(ErrorEvent {
message: format!(
"Conversation is still above the token limit after automatic summarization (limit {limit_str}, current {current_tokens}). Please start a new session or trim your input."
),
});
sess.send_event(&turn_context, event).await;
break;
}
auto_compact_recently_attempted = true;
compact::run_inline_auto_compact_task(sess.clone(), turn_context.clone()).await;
continue;
}
auto_compact_recently_attempted = false;
if responses.is_empty() {
last_agent_message = get_last_assistant_message_from_turn(
&items_to_record_in_conversation_history,
@@ -1905,10 +1915,22 @@ async fn run_turn(
input: Vec<ResponseItem>,
cancellation_token: CancellationToken,
) -> CodexResult<TurnRunResult> {
let mcp_tools = sess.services.mcp_connection_manager.list_all_tools();
let mcp_tools = sess
.services
.mcp_connection_manager
.read()
.await
.list_all_tools()
.or_cancel(&cancellation_token)
.await?;
let router = Arc::new(ToolRouter::from_config(
&turn_context.tools_config,
Some(mcp_tools),
Some(
mcp_tools
.into_iter()
.map(|(name, tool)| (name, tool.tool))
.collect(),
),
));
let model_supports_parallel = turn_context
@@ -2077,7 +2099,7 @@ async fn try_run_turn(
ResponseEvent::Created => {}
ResponseEvent::OutputItemDone(item) => {
let previously_active_item = active_item.take();
match ToolRouter::build_tool_call(sess.as_ref(), item.clone()) {
match ToolRouter::build_tool_call(sess.as_ref(), item.clone()).await {
Ok(Some(call)) => {
let payload_preview = call.payload.log_payload().into_owned();
tracing::info!("ToolCall: {} {}", call.tool_name, payload_preview);
@@ -2201,13 +2223,17 @@ async fn try_run_turn(
error_or_panic("ReasoningSummaryDelta without active item".to_string());
}
}
ResponseEvent::ReasoningSummaryDelta(delta) => {
ResponseEvent::ReasoningSummaryDelta {
delta,
summary_index,
} => {
if let Some(active) = active_item.as_ref() {
let event = ReasoningContentDeltaEvent {
thread_id: sess.conversation_id.to_string(),
turn_id: turn_context.sub_id.clone(),
item_id: active.id(),
delta: delta.clone(),
delta,
summary_index,
};
sess.send_event(&turn_context, EventMsg::ReasoningContentDelta(event))
.await;
@@ -2215,18 +2241,29 @@ async fn try_run_turn(
error_or_panic("ReasoningSummaryDelta without active item".to_string());
}
}
ResponseEvent::ReasoningSummaryPartAdded => {
let event =
EventMsg::AgentReasoningSectionBreak(AgentReasoningSectionBreakEvent {});
sess.send_event(&turn_context, event).await;
ResponseEvent::ReasoningSummaryPartAdded { summary_index } => {
if let Some(active) = active_item.as_ref() {
let event =
EventMsg::AgentReasoningSectionBreak(AgentReasoningSectionBreakEvent {
item_id: active.id(),
summary_index,
});
sess.send_event(&turn_context, event).await;
} else {
error_or_panic("ReasoningSummaryPartAdded without active item".to_string());
}
}
ResponseEvent::ReasoningContentDelta(delta) => {
ResponseEvent::ReasoningContentDelta {
delta,
content_index,
} => {
if let Some(active) = active_item.as_ref() {
let event = ReasoningRawContentDeltaEvent {
thread_id: sess.conversation_id.to_string(),
turn_id: turn_context.sub_id.clone(),
item_id: active.id(),
delta: delta.clone(),
delta,
content_index,
};
sess.send_event(&turn_context, EventMsg::ReasoningRawContentDelta(event))
.await;
@@ -2273,59 +2310,6 @@ pub(super) fn get_last_assistant_message_from_turn(responses: &[ResponseItem]) -
})
}
fn mcp_init_error_display(
server_name: &str,
entry: Option<&McpAuthStatusEntry>,
err: &anyhow::Error,
) -> String {
if let Some(McpServerTransportConfig::StreamableHttp {
url,
bearer_token_env_var,
http_headers,
..
}) = &entry.map(|entry| &entry.config.transport)
&& url == "https://api.githubcopilot.com/mcp/"
&& bearer_token_env_var.is_none()
&& http_headers.as_ref().map(HashMap::is_empty).unwrap_or(true)
{
// GitHub only supports OAUth for first party MCP clients.
// That means that the user has to specify a personal access token either via bearer_token_env_var or http_headers.
// https://github.com/github/github-mcp-server/issues/921#issuecomment-3221026448
format!(
"GitHub MCP does not support OAuth. Log in by adding a personal access token (https://github.com/settings/personal-access-tokens) to your environment and config.toml:\n[mcp_servers.{server_name}]\nbearer_token_env_var = CODEX_GITHUB_PERSONAL_ACCESS_TOKEN"
)
} else if is_mcp_client_auth_required_error(err) {
format!(
"The {server_name} MCP server is not logged in. Run `codex mcp login {server_name}`."
)
} else if is_mcp_client_startup_timeout_error(err) {
let startup_timeout_secs = match entry {
Some(entry) => match entry.config.startup_timeout_sec {
Some(timeout) => timeout,
None => DEFAULT_STARTUP_TIMEOUT,
},
None => DEFAULT_STARTUP_TIMEOUT,
}
.as_secs();
format!(
"MCP client for `{server_name}` timed out after {startup_timeout_secs} seconds. Add or adjust `startup_timeout_sec` in your config.toml:\n[mcp_servers.{server_name}]\nstartup_timeout_sec = XX"
)
} else {
format!("MCP client for `{server_name}` failed to start: {err:#}")
}
}
fn is_mcp_client_auth_required_error(error: &anyhow::Error) -> bool {
// StreamableHttpError::AuthRequired from the MCP SDK.
error.to_string().contains("Auth required")
}
fn is_mcp_client_startup_timeout_error(error: &anyhow::Error) -> bool {
let error_message = error.to_string();
error_message.contains("request timed out")
|| error_message.contains("timed out handshaking with MCP server")
}
use crate::features::Features;
#[cfg(test)]
pub(crate) use tests::make_session_and_context;
@@ -2335,10 +2319,7 @@ mod tests {
use super::*;
use crate::config::ConfigOverrides;
use crate::config::ConfigToml;
use crate::config::types::McpServerConfig;
use crate::config::types::McpServerTransportConfig;
use crate::exec::ExecToolCallOutput;
use crate::mcp::auth::McpAuthStatusEntry;
use crate::tools::format_exec_output_str;
use crate::protocol::CompactedItem;
@@ -2358,7 +2339,6 @@ mod tests {
use codex_app_server_protocol::AuthMode;
use codex_protocol::models::ContentItem;
use codex_protocol::models::ResponseItem;
use codex_protocol::protocol::McpAuthStatus;
use std::time::Duration;
use tokio::time::sleep;
@@ -2572,7 +2552,8 @@ mod tests {
let state = SessionState::new(session_configuration.clone());
let services = SessionServices {
mcp_connection_manager: McpConnectionManager::default(),
mcp_connection_manager: Arc::new(RwLock::new(McpConnectionManager::default())),
mcp_startup_cancellation_token: CancellationToken::new(),
unified_exec_manager: UnifiedExecSessionManager::default(),
notifier: UserNotifier::new(None),
rollout: Mutex::new(None),
@@ -2648,7 +2629,8 @@ mod tests {
let state = SessionState::new(session_configuration.clone());
let services = SessionServices {
mcp_connection_manager: McpConnectionManager::default(),
mcp_connection_manager: Arc::new(RwLock::new(McpConnectionManager::default())),
mcp_startup_cancellation_token: CancellationToken::new(),
unified_exec_manager: UnifiedExecSessionManager::default(),
notifier: UserNotifier::new(None),
rollout: Mutex::new(None),
@@ -2829,9 +2811,23 @@ mod tests {
#[tokio::test]
async fn fatal_tool_error_stops_turn_and_reports_error() {
let (session, turn_context, _rx) = make_session_and_context_with_rx();
let tools = {
session
.services
.mcp_connection_manager
.read()
.await
.list_all_tools()
.await
};
let router = ToolRouter::from_config(
&turn_context.tools_config,
Some(session.services.mcp_connection_manager.list_all_tools()),
Some(
tools
.into_iter()
.map(|(name, tool)| (name, tool.tool))
.collect(),
),
);
let item = ResponseItem::CustomToolCall {
id: None,
@@ -2842,6 +2838,7 @@ mod tests {
};
let call = ToolRouter::build_tool_call(session.as_ref(), item.clone())
.await
.expect("build tool call")
.expect("tool call present");
let tracker = Arc::new(tokio::sync::Mutex::new(TurnDiffTracker::new()));
@@ -3091,7 +3088,6 @@ mod tests {
pretty_assertions::assert_eq!(exec_output.metadata, ResponseExecMetadata { exit_code: 0 });
assert!(exec_output.output.contains("hi"));
}
#[tokio::test]
async fn unified_exec_rejects_escalated_permissions_when_policy_not_on_request() {
use crate::protocol::AskForApproval;
@@ -3133,89 +3129,4 @@ mod tests {
pretty_assertions::assert_eq!(output, expected);
}
#[test]
fn mcp_init_error_display_prompts_for_github_pat() {
let server_name = "github";
let entry = McpAuthStatusEntry {
config: McpServerConfig {
transport: McpServerTransportConfig::StreamableHttp {
url: "https://api.githubcopilot.com/mcp/".to_string(),
bearer_token_env_var: None,
http_headers: None,
env_http_headers: None,
},
enabled: true,
startup_timeout_sec: None,
tool_timeout_sec: None,
enabled_tools: None,
disabled_tools: None,
},
auth_status: McpAuthStatus::Unsupported,
};
let err = anyhow::anyhow!("OAuth is unsupported");
let display = mcp_init_error_display(server_name, Some(&entry), &err);
let expected = format!(
"GitHub MCP does not support OAuth. Log in by adding a personal access token (https://github.com/settings/personal-access-tokens) to your environment and config.toml:\n[mcp_servers.{server_name}]\nbearer_token_env_var = CODEX_GITHUB_PERSONAL_ACCESS_TOKEN"
);
assert_eq!(expected, display);
}
#[test]
fn mcp_init_error_display_prompts_for_login_when_auth_required() {
let server_name = "example";
let err = anyhow::anyhow!("Auth required for server");
let display = mcp_init_error_display(server_name, None, &err);
let expected = format!(
"The {server_name} MCP server is not logged in. Run `codex mcp login {server_name}`."
);
assert_eq!(expected, display);
}
#[test]
fn mcp_init_error_display_reports_generic_errors() {
let server_name = "custom";
let entry = McpAuthStatusEntry {
config: McpServerConfig {
transport: McpServerTransportConfig::StreamableHttp {
url: "https://example.com".to_string(),
bearer_token_env_var: Some("TOKEN".to_string()),
http_headers: None,
env_http_headers: None,
},
enabled: true,
startup_timeout_sec: None,
tool_timeout_sec: None,
enabled_tools: None,
disabled_tools: None,
},
auth_status: McpAuthStatus::Unsupported,
};
let err = anyhow::anyhow!("boom");
let display = mcp_init_error_display(server_name, Some(&entry), &err);
let expected = format!("MCP client for `{server_name}` failed to start: {err:#}");
assert_eq!(expected, display);
}
#[test]
fn mcp_init_error_display_includes_startup_timeout_hint() {
let server_name = "slow";
let err = anyhow::anyhow!("request timed out");
let display = mcp_init_error_display(server_name, None, &err);
assert_eq!(
"MCP client for `slow` timed out after 10 seconds. Add or adjust `startup_timeout_sec` in your config.toml:\n[mcp_servers.slow]\nstartup_timeout_sec = XX",
display
);
}
}
codex-rs/core/src/command_safety/is_safe_command.rs
+4 -6
View File
@@ -1,4 +1,5 @@
use crate::bash::parse_shell_lc_plain_commands;
use crate::command_safety::windows_safe_commands::is_safe_command_windows;
pub fn is_known_safe_command(command: &[String]) -> bool {
let command: Vec<String> = command
@@ -11,12 +12,9 @@ pub fn is_known_safe_command(command: &[String]) -> bool {
}
})
.collect();
#[cfg(target_os = "windows")]
{
use super::windows_safe_commands::is_safe_command_windows;
if is_safe_command_windows(&command) {
return true;
}
if is_safe_command_windows(&command) {
return true;
}
if is_safe_to_call_with_exec(&command) {
codex-rs/core/src/command_safety/mod.rs
-1
View File
@@ -1,4 +1,3 @@
pub mod is_dangerous_command;
pub mod is_safe_command;
#[cfg(target_os = "windows")]
pub mod windows_safe_commands;
codex-rs/core/src/compact.rs
+24 -2
View File
@@ -26,6 +26,7 @@ use futures::prelude::*;
use tracing::error;
pub const SUMMARIZATION_PROMPT: &str = include_str!("../templates/compact/prompt.md");
pub const SUMMARY_PREFIX: &str = include_str!("../templates/compact/summary_prefix.md");
const COMPACT_USER_MESSAGE_MAX_TOKENS: usize = 20_000;
pub(crate) async fn run_inline_auto_compact_task(
@@ -140,7 +141,9 @@ async fn run_compact_task_inner(
}
let history_snapshot = sess.clone_history().await.get_history();
let summary_text = get_last_assistant_message_from_turn(&history_snapshot).unwrap_or_default();
let summary_suffix =
get_last_assistant_message_from_turn(&history_snapshot).unwrap_or_default();
let summary_text = format!("{SUMMARY_PREFIX}\n{summary_suffix}");
let user_messages = collect_user_messages(&history_snapshot);
let initial_context = sess.build_initial_context(turn_context.as_ref());
@@ -153,6 +156,15 @@ async fn run_compact_task_inner(
new_history.extend(ghost_snapshots);
sess.replace_history(new_history).await;
if let Some(estimated_tokens) = sess
.clone_history()
.await
.estimate_token_count(&turn_context)
{
sess.override_last_token_usage_estimate(&turn_context, estimated_tokens)
.await;
}
let rollout_item = RolloutItem::Compacted(CompactedItem {
message: summary_text.clone(),
});
@@ -192,12 +204,22 @@ pub(crate) fn collect_user_messages(items: &[ResponseItem]) -> Vec<String> {
items
.iter()
.filter_map(|item| match crate::event_mapping::parse_turn_item(item) {
Some(TurnItem::UserMessage(user)) => Some(user.message()),
Some(TurnItem::UserMessage(user)) => {
if is_summary_message(&user.message()) {
None
} else {
Some(user.message())
}
}
_ => None,
})
.collect()
}
pub(crate) fn is_summary_message(message: &str) -> bool {
message.starts_with(format!("{SUMMARY_PREFIX}\n").as_str())
}
pub(crate) fn build_compacted_history(
initial_context: Vec<ResponseItem>,
user_messages: &[String],
codex-rs/core/src/config/edit.rs
+29 -14
View File
@@ -3,6 +3,8 @@ use crate::config::types::McpServerConfig;
use crate::config::types::Notice;
use anyhow::Context;
use codex_protocol::config_types::ReasoningEffort;
use codex_protocol::config_types::TrustLevel;
use codex_utils_tokenizer::warm_model_cache;
use std::collections::BTreeMap;
use std::path::Path;
use std::path::PathBuf;
@@ -33,9 +35,9 @@ pub enum ConfigEdit {
SetNoticeHideModelMigrationPrompt(String, bool),
/// Replace the entire `[mcp_servers]` table.
ReplaceMcpServers(BTreeMap<String, McpServerConfig>),
/// Set trust_level = "trusted" under `[projects."<path>"]`,
/// Set trust_level under `[projects."<path>"]`,
/// migrating inline tables to explicit tables.
SetProjectTrusted(PathBuf),
SetProjectTrustLevel { path: PathBuf, level: TrustLevel },
/// Set the value stored at the exact dotted path.
SetPath {
segments: Vec<String>,
@@ -229,6 +231,9 @@ impl ConfigDocument {
fn apply(&mut self, edit: &ConfigEdit) -> anyhow::Result<bool> {
match edit {
ConfigEdit::SetModel { model, effort } => Ok({
if let Some(model) = &model {
warm_model_cache(model)
}
let mut mutated = false;
mutated |= self.write_profile_value(
&["model"],
@@ -270,10 +275,14 @@ impl ConfigDocument {
ConfigEdit::ReplaceMcpServers(servers) => Ok(self.replace_mcp_servers(servers)),
ConfigEdit::SetPath { segments, value } => Ok(self.insert(segments, value.clone())),
ConfigEdit::ClearPath { segments } => Ok(self.clear_owned(segments)),
ConfigEdit::SetProjectTrusted(project_path) => {
ConfigEdit::SetProjectTrustLevel { path, level } => {
// Delegate to the existing, tested logic in config.rs to
// ensure tables are explicit and migration is preserved.
crate::config::set_project_trusted_inner(&mut self.doc, project_path.as_path())?;
crate::config::set_project_trust_level_inner(
&mut self.doc,
path.as_path(),
*level,
)?;
Ok(true)
}
}
@@ -529,9 +538,15 @@ impl ConfigEditsBuilder {
self
}
pub fn set_project_trusted<P: Into<PathBuf>>(mut self, project_path: P) -> Self {
self.edits
.push(ConfigEdit::SetProjectTrusted(project_path.into()));
pub fn set_project_trust_level<P: Into<PathBuf>>(
mut self,
project_path: P,
trust_level: TrustLevel,
) -> Self {
self.edits.push(ConfigEdit::SetProjectTrustLevel {
path: project_path.into(),
level: trust_level,
});
self
}
@@ -569,7 +584,7 @@ mod tests {
codex_home,
None,
&[ConfigEdit::SetModel {
model: Some("gpt-5-codex".to_string()),
model: Some("gpt-5.1-codex".to_string()),
effort: Some(ReasoningEffort::High),
}],
)
@@ -577,7 +592,7 @@ mod tests {
let contents =
std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
let expected = r#"model = "gpt-5-codex"
let expected = r#"model = "gpt-5.1-codex"
model_reasoning_effort = "high"
"#;
assert_eq!(contents, expected);
@@ -707,7 +722,7 @@ model = "o5-preview"
std::fs::write(
codex_home.join(CONFIG_TOML_FILE),
r#"[profiles."team a"]
model = "gpt-5-codex"
model = "gpt-5.1-codex"
"#,
)
.expect("seed");
@@ -957,14 +972,14 @@ B = \"2\"
let codex_home = tmp.path().to_path_buf();
ConfigEditsBuilder::new(&codex_home)
.set_model(Some("gpt-5-codex"), Some(ReasoningEffort::High))
.set_model(Some("gpt-5.1-codex"), Some(ReasoningEffort::High))
.apply()
.await
.expect("persist");
let contents =
std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
let expected = r#"model = "gpt-5-codex"
let expected = r#"model = "gpt-5.1-codex"
model_reasoning_effort = "high"
"#;
assert_eq!(contents, expected);
@@ -986,11 +1001,11 @@ model_reasoning_effort = "low"
std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
assert_eq!(contents, initial_expected);
let updated_expected = r#"model = "gpt-5-codex"
let updated_expected = r#"model = "gpt-5.1-codex"
model_reasoning_effort = "high"
"#;
ConfigEditsBuilder::new(codex_home)
.set_model(Some("gpt-5-codex"), Some(ReasoningEffort::High))
.set_model(Some("gpt-5.1-codex"), Some(ReasoningEffort::High))
.apply_blocking()
.expect("persist update");
contents = std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
codex-rs/core/src/config/mod.rs
+312 -33
View File
@@ -25,7 +25,9 @@ use crate::git_info::resolve_root_git_project_for_trust;
use crate::model_family::ModelFamily;
use crate::model_family::derive_default_model_family;
use crate::model_family::find_family_for_model;
use crate::model_provider_info::LMSTUDIO_OSS_PROVIDER_ID;
use crate::model_provider_info::ModelProviderInfo;
use crate::model_provider_info::OLLAMA_OSS_PROVIDER_ID;
use crate::model_provider_info::built_in_model_providers;
use crate::openai_model_info::get_model_info;
use crate::project_doc::DEFAULT_PROJECT_DOC_FILENAME;
@@ -38,6 +40,7 @@ use codex_protocol::config_types::ForcedLoginMethod;
use codex_protocol::config_types::ReasoningEffort;
use codex_protocol::config_types::ReasoningSummary;
use codex_protocol::config_types::SandboxMode;
use codex_protocol::config_types::TrustLevel;
use codex_protocol::config_types::Verbosity;
use codex_rmcp_client::OAuthCredentialsStoreMode;
use dirs::home_dir;
@@ -59,11 +62,11 @@ pub mod profile;
pub mod types;
#[cfg(target_os = "windows")]
pub const OPENAI_DEFAULT_MODEL: &str = "gpt-5";
pub const OPENAI_DEFAULT_MODEL: &str = "gpt-5.1";
#[cfg(not(target_os = "windows"))]
pub const OPENAI_DEFAULT_MODEL: &str = "gpt-5-codex";
const OPENAI_DEFAULT_REVIEW_MODEL: &str = "gpt-5-codex";
pub const GPT_5_CODEX_MEDIUM_MODEL: &str = "gpt-5-codex";
pub const OPENAI_DEFAULT_MODEL: &str = "gpt-5.1-codex";
const OPENAI_DEFAULT_REVIEW_MODEL: &str = "gpt-5.1-codex";
pub const GPT_5_CODEX_MEDIUM_MODEL: &str = "gpt-5.1-codex";
/// Maximum number of bytes of the documentation that will be embedded. Larger
/// files are *silently truncated* to this size so we do not take up too much of
@@ -78,7 +81,7 @@ pub struct Config {
/// Optional override of model selection.
pub model: String,
/// Model used specifically for review sessions. Defaults to "gpt-5-codex".
/// Model used specifically for review sessions. Defaults to "gpt-5.1-codex".
pub review_model: String,
pub model_family: ModelFamily,
@@ -382,15 +385,16 @@ fn ensure_no_inline_bearer_tokens(value: &TomlValue) -> std::io::Result<()> {
Ok(())
}
pub(crate) fn set_project_trusted_inner(
pub(crate) fn set_project_trust_level_inner(
doc: &mut DocumentMut,
project_path: &Path,
trust_level: TrustLevel,
) -> anyhow::Result<()> {
// Ensure we render a human-friendly structure:
//
// [projects]
// [projects."/path/to/project"]
// trust_level = "trusted"
// trust_level = "trusted" or "untrusted"
//
// rather than inline tables like:
//
@@ -446,20 +450,66 @@ pub(crate) fn set_project_trusted_inner(
return Err(anyhow::anyhow!("project table missing for {project_key}"));
};
proj_tbl.set_implicit(false);
proj_tbl["trust_level"] = toml_edit::value("trusted");
proj_tbl["trust_level"] = toml_edit::value(trust_level.to_string());
Ok(())
}
/// Patch `CODEX_HOME/config.toml` project state.
/// Patch `CODEX_HOME/config.toml` project state to set trust level.
/// Use with caution.
pub fn set_project_trusted(codex_home: &Path, project_path: &Path) -> anyhow::Result<()> {
pub fn set_project_trust_level(
codex_home: &Path,
project_path: &Path,
trust_level: TrustLevel,
) -> anyhow::Result<()> {
use crate::config::edit::ConfigEditsBuilder;
ConfigEditsBuilder::new(codex_home)
.set_project_trusted(project_path)
.set_project_trust_level(project_path, trust_level)
.apply_blocking()
}
/// Save the default OSS provider preference to config.toml
pub fn set_default_oss_provider(codex_home: &Path, provider: &str) -> std::io::Result<()> {
// Validate that the provider is one of the known OSS providers
match provider {
LMSTUDIO_OSS_PROVIDER_ID | OLLAMA_OSS_PROVIDER_ID => {
// Valid provider, continue
}
_ => {
return Err(std::io::Error::new(
std::io::ErrorKind::InvalidInput,
format!(
"Invalid OSS provider '{provider}'. Must be one of: {LMSTUDIO_OSS_PROVIDER_ID}, {OLLAMA_OSS_PROVIDER_ID}"
),
));
}
}
let config_path = codex_home.join(CONFIG_TOML_FILE);
// Read existing config or create empty string if file doesn't exist
let content = match std::fs::read_to_string(&config_path) {
Ok(content) => content,
Err(e) if e.kind() == std::io::ErrorKind::NotFound => String::new(),
Err(e) => return Err(e),
};
// Parse as DocumentMut for editing while preserving structure
let mut doc = content.parse::<DocumentMut>().map_err(|e| {
std::io::Error::new(
std::io::ErrorKind::InvalidData,
format!("failed to parse config.toml: {e}"),
)
})?;
// Set the default_oss_provider at root level
use toml_edit::value;
doc["oss_provider"] = value(provider);
// Write the modified document back
std::fs::write(&config_path, doc.to_string())?;
Ok(())
}
/// Apply a single dotted-path override onto a TOML value.
fn apply_toml_override(root: &mut TomlValue, path: &str, value: TomlValue) {
use toml::value::Table;
@@ -657,6 +707,8 @@ pub struct ConfigToml {
pub experimental_use_rmcp_client: Option<bool>,
pub experimental_use_freeform_apply_patch: Option<bool>,
pub experimental_sandbox_command_assessment: Option<bool>,
/// Preferred OSS provider for local models, e.g. "lmstudio" or "ollama".
pub oss_provider: Option<String>,
}
impl From<ConfigToml> for UserSavedConfig {
@@ -686,15 +738,16 @@ impl From<ConfigToml> for UserSavedConfig {
#[derive(Deserialize, Debug, Clone, PartialEq, Eq)]
pub struct ProjectConfig {
pub trust_level: Option<String>,
pub trust_level: Option<TrustLevel>,
}
impl ProjectConfig {
pub fn is_trusted(&self) -> bool {
match &self.trust_level {
Some(trust_level) => trust_level == "trusted",
None => false,
}
matches!(self.trust_level, Some(TrustLevel::Trusted))
}
pub fn is_untrusted(&self) -> bool {
matches!(self.trust_level, Some(TrustLevel::Untrusted))
}
}
@@ -735,9 +788,9 @@ impl ConfigToml {
.or(profile_sandbox_mode)
.or(self.sandbox_mode)
.or_else(|| {
// if no sandbox_mode is set, but user has marked directory as trusted, use WorkspaceWrite
// if no sandbox_mode is set, but user has marked directory as trusted or untrusted, use WorkspaceWrite
self.get_active_project(resolved_cwd).and_then(|p| {
if p.is_trusted() {
if p.is_trusted() || p.is_untrusted() {
Some(SandboxMode::WorkspaceWrite)
} else {
None
@@ -844,6 +897,34 @@ pub struct ConfigOverrides {
pub additional_writable_roots: Vec<PathBuf>,
}
/// Resolves the OSS provider from CLI override, profile config, or global config.
/// Returns `None` if no provider is configured at any level.
pub fn resolve_oss_provider(
explicit_provider: Option<&str>,
config_toml: &ConfigToml,
config_profile: Option<String>,
) -> Option<String> {
if let Some(provider) = explicit_provider {
// Explicit provider specified (e.g., via --local-provider)
Some(provider.to_string())
} else {
// Check profile config first, then global config
let profile = config_toml.get_config_profile(config_profile).ok();
if let Some(profile) = &profile {
// Check if profile has an oss provider
if let Some(profile_oss_provider) = &profile.oss_provider {
Some(profile_oss_provider.clone())
}
// If not then check if the toml has an oss provider
else {
config_toml.oss_provider.clone()
}
} else {
config_toml.oss_provider.clone()
}
}
}
impl Config {
/// Meant to be used exclusively for tests: `load_with_overrides()` should
/// be used in all other cases.
@@ -958,6 +1039,9 @@ impl Config {
if active_project.is_trusted() {
// If no explicit approval policy is set, but we trust cwd, default to OnRequest
AskForApproval::OnRequest
} else if active_project.is_untrusted() {
// If project is explicitly marked untrusted, require approval for non-safe commands
AskForApproval::UnlessTrusted
} else {
AskForApproval::default()
}
@@ -1323,7 +1407,7 @@ persistence = "none"
}
#[test]
fn tui_config_missing_notifications_field_defaults_to_disabled() {
fn tui_config_missing_notifications_field_defaults_to_enabled() {
let cfg = r#"
[tui]
"#;
@@ -1332,7 +1416,7 @@ persistence = "none"
.expect("TUI config without notifications should succeed");
let tui = parsed.tui.expect("config should include tui section");
assert_eq!(tui.notifications, Notifications::Enabled(false));
assert_eq!(tui.notifications, Notifications::Enabled(true));
}
#[test]
@@ -2532,7 +2616,7 @@ url = "https://example.com/mcp"
let codex_home = TempDir::new()?;
ConfigEditsBuilder::new(codex_home.path())
.set_model(Some("gpt-5-codex"), Some(ReasoningEffort::High))
.set_model(Some("gpt-5.1-codex"), Some(ReasoningEffort::High))
.apply()
.await?;
@@ -2540,7 +2624,7 @@ url = "https://example.com/mcp"
tokio::fs::read_to_string(codex_home.path().join(CONFIG_TOML_FILE)).await?;
let parsed: ConfigToml = toml::from_str(&serialized)?;
assert_eq!(parsed.model.as_deref(), Some("gpt-5-codex"));
assert_eq!(parsed.model.as_deref(), Some("gpt-5.1-codex"));
assert_eq!(parsed.model_reasoning_effort, Some(ReasoningEffort::High));
Ok(())
@@ -2554,7 +2638,7 @@ url = "https://example.com/mcp"
tokio::fs::write(
&config_path,
r#"
model = "gpt-5-codex"
model = "gpt-5.1-codex"
model_reasoning_effort = "medium"
[profiles.dev]
@@ -2590,7 +2674,7 @@ model = "gpt-4.1"
ConfigEditsBuilder::new(codex_home.path())
.with_profile(Some("dev"))
.set_model(Some("gpt-5-codex"), Some(ReasoningEffort::Medium))
.set_model(Some("gpt-5.1-codex"), Some(ReasoningEffort::Medium))
.apply()
.await?;
@@ -2602,7 +2686,7 @@ model = "gpt-4.1"
.get("dev")
.expect("profile should be created");
assert_eq!(profile.model.as_deref(), Some("gpt-5-codex"));
assert_eq!(profile.model.as_deref(), Some("gpt-5.1-codex"));
assert_eq!(
profile.model_reasoning_effort,
Some(ReasoningEffort::Medium)
@@ -2624,7 +2708,7 @@ model = "gpt-4"
model_reasoning_effort = "medium"
[profiles.prod]
model = "gpt-5-codex"
model = "gpt-5.1-codex"
"#,
)
.await?;
@@ -2653,7 +2737,7 @@ model = "gpt-5-codex"
.profiles
.get("prod")
.and_then(|profile| profile.model.as_deref()),
Some("gpt-5-codex"),
Some("gpt-5.1-codex"),
);
Ok(())
@@ -2768,7 +2852,7 @@ model_provider = "openai"
approval_policy = "on-failure"
[profiles.gpt5]
model = "gpt-5"
model = "gpt-5.1"
model_provider = "openai"
approval_policy = "on-failure"
model_reasoning_effort = "high"
@@ -3084,9 +3168,9 @@ model_verbosity = "high"
fixture.codex_home(),
)?;
let expected_gpt5_profile_config = Config {
model: "gpt-5".to_string(),
model: "gpt-5.1".to_string(),
review_model: OPENAI_DEFAULT_REVIEW_MODEL.to_string(),
model_family: find_family_for_model("gpt-5").expect("known model slug"),
model_family: find_family_for_model("gpt-5.1").expect("known model slug"),
model_context_window: Some(272_000),
model_max_output_tokens: Some(128_000),
model_auto_compact_token_limit: Some(244_800),
@@ -3164,7 +3248,7 @@ model_verbosity = "high"
let project_dir = Path::new("/some/path");
let mut doc = DocumentMut::new();
set_project_trusted_inner(&mut doc, project_dir)?;
set_project_trust_level_inner(&mut doc, project_dir, TrustLevel::Trusted)?;
let contents = doc.to_string();
@@ -3204,7 +3288,7 @@ trust_level = "trusted"
let mut doc = initial.parse::<DocumentMut>()?;
// Run the function; it should convert to explicit tables and set trusted
set_project_trusted_inner(&mut doc, project_dir)?;
set_project_trust_level_inner(&mut doc, project_dir, TrustLevel::Trusted)?;
let contents = doc.to_string();
@@ -3231,7 +3315,7 @@ model = "foo""#;
// Approve a new directory
let new_project = Path::new("/Users/mbolin/code/codex2");
set_project_trusted_inner(&mut doc, new_project)?;
set_project_trust_level_inner(&mut doc, new_project, TrustLevel::Trusted)?;
let contents = doc.to_string();
@@ -3254,6 +3338,201 @@ trust_level = "trusted"
Ok(())
}
#[test]
fn test_set_default_oss_provider() -> std::io::Result<()> {
let temp_dir = TempDir::new()?;
let codex_home = temp_dir.path();
let config_path = codex_home.join(CONFIG_TOML_FILE);
// Test setting valid provider on empty config
set_default_oss_provider(codex_home, OLLAMA_OSS_PROVIDER_ID)?;
let content = std::fs::read_to_string(&config_path)?;
assert!(content.contains("oss_provider = \"ollama\""));
// Test updating existing config
std::fs::write(&config_path, "model = \"gpt-4\"\n")?;
set_default_oss_provider(codex_home, LMSTUDIO_OSS_PROVIDER_ID)?;
let content = std::fs::read_to_string(&config_path)?;
assert!(content.contains("oss_provider = \"lmstudio\""));
assert!(content.contains("model = \"gpt-4\""));
// Test overwriting existing oss_provider
set_default_oss_provider(codex_home, OLLAMA_OSS_PROVIDER_ID)?;
let content = std::fs::read_to_string(&config_path)?;
assert!(content.contains("oss_provider = \"ollama\""));
assert!(!content.contains("oss_provider = \"lmstudio\""));
// Test invalid provider
let result = set_default_oss_provider(codex_home, "invalid_provider");
assert!(result.is_err());
let error = result.unwrap_err();
assert_eq!(error.kind(), std::io::ErrorKind::InvalidInput);
assert!(error.to_string().contains("Invalid OSS provider"));
assert!(error.to_string().contains("invalid_provider"));
Ok(())
}
#[test]
fn test_untrusted_project_gets_workspace_write_sandbox() -> anyhow::Result<()> {
let config_with_untrusted = r#"
[projects."/tmp/test"]
trust_level = "untrusted"
"#;
let cfg = toml::from_str::<ConfigToml>(config_with_untrusted)
.expect("TOML deserialization should succeed");
let resolution = cfg.derive_sandbox_policy(None, None, &PathBuf::from("/tmp/test"));
// Verify that untrusted projects get WorkspaceWrite (or ReadOnly on Windows due to downgrade)
if cfg!(target_os = "windows") {
assert!(
matches!(resolution.policy, SandboxPolicy::ReadOnly),
"Expected ReadOnly on Windows, got {:?}",
resolution.policy
);
} else {
assert!(
matches!(resolution.policy, SandboxPolicy::WorkspaceWrite { .. }),
"Expected WorkspaceWrite for untrusted project, got {:?}",
resolution.policy
);
}
Ok(())
}
#[test]
fn test_resolve_oss_provider_explicit_override() {
let config_toml = ConfigToml::default();
let result = resolve_oss_provider(Some("custom-provider"), &config_toml, None);
assert_eq!(result, Some("custom-provider".to_string()));
}
#[test]
fn test_resolve_oss_provider_from_profile() {
let mut profiles = std::collections::HashMap::new();
let profile = ConfigProfile {
oss_provider: Some("profile-provider".to_string()),
..Default::default()
};
profiles.insert("test-profile".to_string(), profile);
let config_toml = ConfigToml {
profiles,
..Default::default()
};
let result = resolve_oss_provider(None, &config_toml, Some("test-profile".to_string()));
assert_eq!(result, Some("profile-provider".to_string()));
}
#[test]
fn test_resolve_oss_provider_from_global_config() {
let config_toml = ConfigToml {
oss_provider: Some("global-provider".to_string()),
..Default::default()
};
let result = resolve_oss_provider(None, &config_toml, None);
assert_eq!(result, Some("global-provider".to_string()));
}
#[test]
fn test_resolve_oss_provider_profile_fallback_to_global() {
let mut profiles = std::collections::HashMap::new();
let profile = ConfigProfile::default(); // No oss_provider set
profiles.insert("test-profile".to_string(), profile);
let config_toml = ConfigToml {
oss_provider: Some("global-provider".to_string()),
profiles,
..Default::default()
};
let result = resolve_oss_provider(None, &config_toml, Some("test-profile".to_string()));
assert_eq!(result, Some("global-provider".to_string()));
}
#[test]
fn test_resolve_oss_provider_none_when_not_configured() {
let config_toml = ConfigToml::default();
let result = resolve_oss_provider(None, &config_toml, None);
assert_eq!(result, None);
}
#[test]
fn test_resolve_oss_provider_explicit_overrides_all() {
let mut profiles = std::collections::HashMap::new();
let profile = ConfigProfile {
oss_provider: Some("profile-provider".to_string()),
..Default::default()
};
profiles.insert("test-profile".to_string(), profile);
let config_toml = ConfigToml {
oss_provider: Some("global-provider".to_string()),
profiles,
..Default::default()
};
let result = resolve_oss_provider(
Some("explicit-provider"),
&config_toml,
Some("test-profile".to_string()),
);
assert_eq!(result, Some("explicit-provider".to_string()));
}
#[test]
fn test_untrusted_project_gets_unless_trusted_approval_policy() -> std::io::Result<()> {
let codex_home = TempDir::new()?;
let test_project_dir = TempDir::new()?;
let test_path = test_project_dir.path();
let mut projects = std::collections::HashMap::new();
projects.insert(
test_path.to_string_lossy().to_string(),
ProjectConfig {
trust_level: Some(TrustLevel::Untrusted),
},
);
let cfg = ConfigToml {
projects: Some(projects),
..Default::default()
};
let config = Config::load_from_base_config_with_overrides(
cfg,
ConfigOverrides {
cwd: Some(test_path.to_path_buf()),
..Default::default()
},
codex_home.path().to_path_buf(),
)?;
// Verify that untrusted projects get UnlessTrusted approval policy
assert_eq!(
config.approval_policy,
AskForApproval::UnlessTrusted,
"Expected UnlessTrusted approval policy for untrusted project"
);
// Verify that untrusted projects still get WorkspaceWrite sandbox (or ReadOnly on Windows)
if cfg!(target_os = "windows") {
assert!(
matches!(config.sandbox_policy, SandboxPolicy::ReadOnly),
"Expected ReadOnly on Windows"
);
} else {
assert!(
matches!(config.sandbox_policy, SandboxPolicy::WorkspaceWrite { .. }),
"Expected WorkspaceWrite sandbox for untrusted project"
);
}
Ok(())
}
}
#[cfg(test)]
codex-rs/core/src/config/profile.rs
+1
View File
@@ -33,6 +33,7 @@ pub struct ConfigProfile {
/// Optional feature toggles scoped to this profile.
#[serde(default)]
pub features: Option<crate::features::FeaturesToml>,
pub oss_provider: Option<String>,
}
impl From<ConfigProfile> for codex_app_server_protocol::Profile {
codex-rs/core/src/config/types.rs
+2 -2
View File
@@ -338,7 +338,7 @@ pub enum Notifications {
impl Default for Notifications {
fn default() -> Self {
Self::Enabled(false)
Self::Enabled(true)
}
}
@@ -346,7 +346,7 @@ impl Default for Notifications {
#[derive(Deserialize, Debug, Clone, PartialEq, Default)]
pub struct Tui {
/// Enable desktop notifications from the TUI when the terminal is unfocused.
/// Defaults to `false`.
/// Defaults to `true`.
#[serde(default)]
pub notifications: Notifications,
}
codex-rs/core/src/context_manager/history.rs
+47 -5
View File
@@ -1,12 +1,20 @@
use crate::codex::TurnContext;
use crate::context_manager::normalize;
use crate::truncate;
use crate::truncate::format_output_for_model_body;
use crate::truncate::globally_truncate_function_output_items;
use codex_protocol::models::FunctionCallOutputPayload;
use codex_protocol::models::ResponseItem;
use codex_protocol::protocol::TokenUsage;
use codex_protocol::protocol::TokenUsageInfo;
use codex_utils_tokenizer::Tokenizer;
use std::ops::Deref;
use crate::context_manager::normalize;
use crate::context_manager::truncate::format_output_for_model_body;
use crate::context_manager::truncate::globally_truncate_function_output_items;
const CONTEXT_WINDOW_HARD_LIMIT_FACTOR: f64 = 1.1;
const CONTEXT_WINDOW_HARD_LIMIT_BYTES: usize =
(truncate::MODEL_FORMAT_MAX_BYTES as f64 * CONTEXT_WINDOW_HARD_LIMIT_FACTOR) as usize;
const CONTEXT_WINDOW_HARD_LIMIT_LINES: usize =
(truncate::MODEL_FORMAT_MAX_LINES as f64 * CONTEXT_WINDOW_HARD_LIMIT_FACTOR) as usize;
/// Transcript of conversation history
#[derive(Debug, Clone, Default)]
@@ -28,6 +36,10 @@ impl ContextManager {
self.token_info.clone()
}
pub(crate) fn set_token_info(&mut self, info: Option<TokenUsageInfo>) {
self.token_info = info;
}
pub(crate) fn set_token_usage_full(&mut self, context_window: i64) {
match &mut self.token_info {
Some(info) => info.fill_to_context_window(context_window),
@@ -68,6 +80,28 @@ impl ContextManager {
history
}
// Estimate the number of tokens in the history. Return None if no tokenizer
// is available. This does not consider the reasoning traces.
// /!\ The value is a lower bound estimate and does not represent the exact
// context length.
pub(crate) fn estimate_token_count(&self, turn_context: &TurnContext) -> Option<i64> {
let model = turn_context.client.get_model();
let tokenizer = Tokenizer::for_model(model.as_str()).ok()?;
let model_family = turn_context.client.get_model_family();
Some(
self.items
.iter()
.map(|item| {
serde_json::to_string(&item)
.map(|item| tokenizer.count(&item))
.unwrap_or_default()
})
.sum::<i64>()
+ tokenizer.count(model_family.base_instructions.as_str()),
)
}
pub(crate) fn remove_first_item(&mut self) {
if !self.items.is_empty() {
// Remove the oldest item (front of the list). Items are ordered from
@@ -119,7 +153,11 @@ impl ContextManager {
fn process_item(item: &ResponseItem) -> ResponseItem {
match item {
ResponseItem::FunctionCallOutput { call_id, output } => {
let truncated = format_output_for_model_body(output.content.as_str());
let truncated = format_output_for_model_body(
output.content.as_str(),
CONTEXT_WINDOW_HARD_LIMIT_BYTES,
CONTEXT_WINDOW_HARD_LIMIT_LINES,
);
let truncated_items = output
.content_items
.as_ref()
@@ -134,7 +172,11 @@ impl ContextManager {
}
}
ResponseItem::CustomToolCallOutput { call_id, output } => {
let truncated = format_output_for_model_body(output);
let truncated = format_output_for_model_body(
output,
CONTEXT_WINDOW_HARD_LIMIT_BYTES,
CONTEXT_WINDOW_HARD_LIMIT_LINES,
);
ResponseItem::CustomToolCallOutput {
call_id: call_id.clone(),
output: truncated,
codex-rs/core/src/context_manager/history_tests.rs
+34 -8
View File
@@ -1,5 +1,6 @@
use super::*;
use crate::context_manager::truncate;
use crate::context_manager::MODEL_FORMAT_MAX_LINES;
use crate::truncate;
use codex_git::GhostCommit;
use codex_protocol::models::ContentItem;
use codex_protocol::models::FunctionCallOutputContentItem;
@@ -308,8 +309,10 @@ fn assert_truncated_message_matches(message: &str, line: &str, total_lines: usiz
}
fn truncated_message_pattern(line: &str, total_lines: usize) -> String {
let head_take = truncate::MODEL_FORMAT_HEAD_LINES.min(total_lines);
let tail_take = truncate::MODEL_FORMAT_TAIL_LINES.min(total_lines.saturating_sub(head_take));
let head_lines = MODEL_FORMAT_MAX_LINES / 2;
let tail_lines = MODEL_FORMAT_MAX_LINES - head_lines;
let head_take = head_lines.min(total_lines);
let tail_take = tail_lines.min(total_lines.saturating_sub(head_take));
let omitted = total_lines.saturating_sub(head_take + tail_take);
let escaped_line = regex_lite::escape(line);
if omitted == 0 {
@@ -328,7 +331,11 @@ fn format_exec_output_truncates_large_error() {
let line = "very long execution error line that should trigger truncation\n";
let large_error = line.repeat(2_500); // way beyond both byte and line limits
let truncated = truncate::format_output_for_model_body(&large_error);
let truncated = truncate::format_output_for_model_body(
&large_error,
truncate::MODEL_FORMAT_MAX_BYTES,
truncate::MODEL_FORMAT_MAX_LINES,
);
let total_lines = large_error.lines().count();
assert_truncated_message_matches(&truncated, line, total_lines);
@@ -338,7 +345,11 @@ fn format_exec_output_truncates_large_error() {
#[test]
fn format_exec_output_marks_byte_truncation_without_omitted_lines() {
let long_line = "a".repeat(truncate::MODEL_FORMAT_MAX_BYTES + 50);
let truncated = truncate::format_output_for_model_body(&long_line);
let truncated = truncate::format_output_for_model_body(
&long_line,
truncate::MODEL_FORMAT_MAX_BYTES,
truncate::MODEL_FORMAT_MAX_LINES,
);
assert_ne!(truncated, long_line);
let marker_line = format!(
@@ -359,7 +370,14 @@ fn format_exec_output_marks_byte_truncation_without_omitted_lines() {
fn format_exec_output_returns_original_when_within_limits() {
let content = "example output\n".repeat(10);
assert_eq!(truncate::format_output_for_model_body(&content), content);
assert_eq!(
truncate::format_output_for_model_body(
&content,
truncate::MODEL_FORMAT_MAX_BYTES,
truncate::MODEL_FORMAT_MAX_LINES
),
content
);
}
#[test]
@@ -369,7 +387,11 @@ fn format_exec_output_reports_omitted_lines_and_keeps_head_and_tail() {
.map(|idx| format!("line-{idx}\n"))
.collect();
let truncated = truncate::format_output_for_model_body(&content);
let truncated = truncate::format_output_for_model_body(
&content,
truncate::MODEL_FORMAT_MAX_BYTES,
truncate::MODEL_FORMAT_MAX_LINES,
);
let omitted = total_lines - truncate::MODEL_FORMAT_MAX_LINES;
let expected_marker = format!("[... omitted {omitted} of {total_lines} lines ...]");
@@ -397,7 +419,11 @@ fn format_exec_output_prefers_line_marker_when_both_limits_exceeded() {
.map(|idx| format!("line-{idx}-{long_line}\n"))
.collect();
let truncated = truncate::format_output_for_model_body(&content);
let truncated = truncate::format_output_for_model_body(
&content,
truncate::MODEL_FORMAT_MAX_BYTES,
truncate::MODEL_FORMAT_MAX_LINES,
);
assert!(
truncated.contains("[... omitted 42 of 298 lines ...]"),
codex-rs/core/src/context_manager/mod.rs
+3 -2
View File
@@ -1,6 +1,7 @@
mod history;
mod normalize;
mod truncate;
pub(crate) use crate::truncate::MODEL_FORMAT_MAX_BYTES;
pub(crate) use crate::truncate::MODEL_FORMAT_MAX_LINES;
pub(crate) use crate::truncate::format_output_for_model_body;
pub(crate) use history::ContextManager;
pub(crate) use truncate::format_output_for_model_body;
codex-rs/core/src/context_manager/truncate.rs
-128
View File
@@ -1,128 +0,0 @@
use codex_protocol::models::FunctionCallOutputContentItem;
use codex_utils_string::take_bytes_at_char_boundary;
use codex_utils_string::take_last_bytes_at_char_boundary;
// Model-formatting limits: clients get full streams; only content sent to the model is truncated.
pub(crate) const MODEL_FORMAT_MAX_BYTES: usize = 10 * 1024; // 10 KiB
pub(crate) const MODEL_FORMAT_MAX_LINES: usize = 256; // lines
pub(crate) const MODEL_FORMAT_HEAD_LINES: usize = MODEL_FORMAT_MAX_LINES / 2;
pub(crate) const MODEL_FORMAT_TAIL_LINES: usize = MODEL_FORMAT_MAX_LINES - MODEL_FORMAT_HEAD_LINES; // 128
pub(crate) const MODEL_FORMAT_HEAD_BYTES: usize = MODEL_FORMAT_MAX_BYTES / 2;
pub(crate) fn globally_truncate_function_output_items(
items: &[FunctionCallOutputContentItem],
) -> Vec<FunctionCallOutputContentItem> {
let mut out: Vec<FunctionCallOutputContentItem> = Vec::with_capacity(items.len());
let mut remaining = MODEL_FORMAT_MAX_BYTES;
let mut omitted_text_items = 0usize;
for it in items {
match it {
FunctionCallOutputContentItem::InputText { text } => {
if remaining == 0 {
omitted_text_items += 1;
continue;
}
let len = text.len();
if len <= remaining {
out.push(FunctionCallOutputContentItem::InputText { text: text.clone() });
remaining -= len;
} else {
let slice = take_bytes_at_char_boundary(text, remaining);
if !slice.is_empty() {
out.push(FunctionCallOutputContentItem::InputText {
text: slice.to_string(),
});
}
remaining = 0;
}
}
// todo(aibrahim): handle input images; resize
FunctionCallOutputContentItem::InputImage { image_url } => {
out.push(FunctionCallOutputContentItem::InputImage {
image_url: image_url.clone(),
});
}
}
}
if omitted_text_items > 0 {
out.push(FunctionCallOutputContentItem::InputText {
text: format!("[omitted {omitted_text_items} text items ...]"),
});
}
out
}
pub(crate) fn format_output_for_model_body(content: &str) -> String {
// Head+tail truncation for the model: show the beginning and end with an elision.
// Clients still receive full streams; only this formatted summary is capped.
let total_lines = content.lines().count();
if content.len() <= MODEL_FORMAT_MAX_BYTES && total_lines <= MODEL_FORMAT_MAX_LINES {
return content.to_string();
}
let output = truncate_formatted_exec_output(content, total_lines);
format!("Total output lines: {total_lines}\n\n{output}")
}
fn truncate_formatted_exec_output(content: &str, total_lines: usize) -> String {
let segments: Vec<&str> = content.split_inclusive('\n').collect();
let head_take = MODEL_FORMAT_HEAD_LINES.min(segments.len());
let tail_take = MODEL_FORMAT_TAIL_LINES.min(segments.len().saturating_sub(head_take));
let omitted = segments.len().saturating_sub(head_take + tail_take);
let head_slice_end: usize = segments
.iter()
.take(head_take)
.map(|segment| segment.len())
.sum();
let tail_slice_start: usize = if tail_take == 0 {
content.len()
} else {
content.len()
- segments
.iter()
.rev()
.take(tail_take)
.map(|segment| segment.len())
.sum::<usize>()
};
let head_slice = &content[..head_slice_end];
let tail_slice = &content[tail_slice_start..];
let truncated_by_bytes = content.len() > MODEL_FORMAT_MAX_BYTES;
// this is a bit wrong. We are counting metadata lines and not just shell output lines.
let marker = if omitted > 0 {
Some(format!(
"\n[... omitted {omitted} of {total_lines} lines ...]\n\n"
))
} else if truncated_by_bytes {
Some(format!(
"\n[... output truncated to fit {MODEL_FORMAT_MAX_BYTES} bytes ...]\n\n"
))
} else {
None
};
let marker_len = marker.as_ref().map_or(0, String::len);
let base_head_budget = MODEL_FORMAT_HEAD_BYTES.min(MODEL_FORMAT_MAX_BYTES);
let head_budget = base_head_budget.min(MODEL_FORMAT_MAX_BYTES.saturating_sub(marker_len));
let head_part = take_bytes_at_char_boundary(head_slice, head_budget);
let mut result = String::with_capacity(MODEL_FORMAT_MAX_BYTES.min(content.len()));
result.push_str(head_part);
if let Some(marker_text) = marker.as_ref() {
result.push_str(marker_text);
}
let remaining = MODEL_FORMAT_MAX_BYTES.saturating_sub(result.len());
if remaining == 0 {
return result;
}
let tail_part = take_last_bytes_at_char_boundary(tail_slice, remaining);
result.push_str(tail_part);
result
}
codex-rs/core/src/environment_context.rs
-2
View File
@@ -329,7 +329,6 @@ mod tests {
Some(workspace_write_policy(vec!["/repo"], false)),
Some(Shell::Bash(BashShell {
shell_path: "/bin/bash".into(),
bashrc_path: "/home/user/.bashrc".into(),
})),
);
let context2 = EnvironmentContext::new(
@@ -338,7 +337,6 @@ mod tests {
Some(workspace_write_policy(vec!["/repo"], false)),
Some(Shell::Zsh(ZshShell {
shell_path: "/bin/zsh".into(),
zshrc_path: "/home/user/.zshrc".into(),
})),
);
codex-rs/core/src/exec.rs
+6 -2
View File
@@ -189,16 +189,20 @@ async fn exec_windows_sandbox(
};
let sandbox_cwd = cwd.clone();
let logs_base_dir = find_codex_home().ok();
let codex_home = find_codex_home().map_err(|err| {
CodexErr::Io(io::Error::other(format!(
"windows sandbox: failed to resolve codex_home: {err}"
)))
})?;
let spawn_res = tokio::task::spawn_blocking(move || {
run_windows_sandbox_capture(
policy_str,
&sandbox_cwd,
codex_home.as_ref(),
command,
&cwd,
env,
timeout_ms,
logs_base_dir.as_deref(),
)
})
.await;
codex-rs/core/src/features.rs
+8
View File
@@ -46,6 +46,8 @@ pub enum Feature {
GhostCommit,
/// Enable Windows sandbox (restricted token) on Windows.
WindowsSandbox,
/// Enable the default shell tool.
ShellTool,
}
impl Feature {
@@ -301,4 +303,10 @@ pub const FEATURES: &[FeatureSpec] = &[
stage: Stage::Experimental,
default_enabled: false,
},
FeatureSpec {
id: Feature::ShellTool,
key: "shell_tool",
stage: Stage::Stable,
default_enabled: true,
},
];
codex-rs/core/src/lib.rs
+5 -1
View File
@@ -13,6 +13,7 @@ mod client;
mod client_common;
pub mod codex;
mod codex_conversation;
pub mod powershell;
pub use codex_conversation::CodexConversation;
mod codex_delegate;
mod command_safety;
@@ -40,8 +41,11 @@ pub mod token_data;
mod truncate;
mod unified_exec;
mod user_instructions;
pub use model_provider_info::BUILT_IN_OSS_MODEL_PROVIDER_ID;
pub use model_provider_info::DEFAULT_LMSTUDIO_PORT;
pub use model_provider_info::DEFAULT_OLLAMA_PORT;
pub use model_provider_info::LMSTUDIO_OSS_PROVIDER_ID;
pub use model_provider_info::ModelProviderInfo;
pub use model_provider_info::OLLAMA_OSS_PROVIDER_ID;
pub use model_provider_info::WireApi;
pub use model_provider_info::built_in_model_providers;
pub use model_provider_info::create_oss_provider_with_base_url;
codex-rs/core/src/mcp_connection_manager.rs
+532 -307
View File
File diff suppressed because it is too large Load Diff
codex-rs/core/src/model_family.rs
+3 -1
View File
@@ -132,7 +132,7 @@ pub fn find_family_for_model(slug: &str) -> Option<ModelFamily> {
model_family!(slug, "gpt-4o", needs_special_apply_patch_instructions: true)
} else if slug.starts_with("gpt-3.5") {
model_family!(slug, "gpt-3.5", needs_special_apply_patch_instructions: true)
} else if slug.starts_with("test-gpt-5-codex") {
} else if slug.starts_with("test-gpt-5") {
model_family!(
slug, slug,
supports_reasoning_summaries: true,
@@ -161,6 +161,7 @@ pub fn find_family_for_model(slug: &str) -> Option<ModelFamily> {
"list_dir".to_string(),
"read_file".to_string(),
],
shell_type: if cfg!(windows) { ConfigShellToolType::ShellCommand } else { ConfigShellToolType::Default },
supports_parallel_tool_calls: true,
support_verbosity: true,
)
@@ -176,6 +177,7 @@ pub fn find_family_for_model(slug: &str) -> Option<ModelFamily> {
reasoning_summary_format: ReasoningSummaryFormat::Experimental,
base_instructions: GPT_5_CODEX_INSTRUCTIONS.to_string(),
apply_patch_tool_type: Some(ApplyPatchToolType::Freeform),
shell_type: if cfg!(windows) { ConfigShellToolType::ShellCommand } else { ConfigShellToolType::Default },
support_verbosity: false,
)
} else if slug.starts_with("gpt-5.1") {
codex-rs/core/src/model_provider_info.rs
+18 -10
View File
@@ -258,9 +258,11 @@ impl ModelProviderInfo {
}
}
const DEFAULT_OLLAMA_PORT: u32 = 11434;
pub const DEFAULT_LMSTUDIO_PORT: u16 = 1234;
pub const DEFAULT_OLLAMA_PORT: u16 = 11434;
pub const BUILT_IN_OSS_MODEL_PROVIDER_ID: &str = "oss";
pub const LMSTUDIO_OSS_PROVIDER_ID: &str = "lmstudio";
pub const OLLAMA_OSS_PROVIDER_ID: &str = "ollama";
/// Built-in default provider list.
pub fn built_in_model_providers() -> HashMap<String, ModelProviderInfo> {
@@ -311,14 +313,21 @@ pub fn built_in_model_providers() -> HashMap<String, ModelProviderInfo> {
requires_openai_auth: true,
},
),
(BUILT_IN_OSS_MODEL_PROVIDER_ID, create_oss_provider()),
(
OLLAMA_OSS_PROVIDER_ID,
create_oss_provider(DEFAULT_OLLAMA_PORT, WireApi::Chat),
),
(
LMSTUDIO_OSS_PROVIDER_ID,
create_oss_provider(DEFAULT_LMSTUDIO_PORT, WireApi::Responses),
),
]
.into_iter()
.map(|(k, v)| (k.to_string(), v))
.collect()
}
pub fn create_oss_provider() -> ModelProviderInfo {
pub fn create_oss_provider(default_provider_port: u16, wire_api: WireApi) -> ModelProviderInfo {
// These CODEX_OSS_ environment variables are experimental: we may
// switch to reading values from config.toml instead.
let codex_oss_base_url = match std::env::var("CODEX_OSS_BASE_URL")
@@ -331,22 +340,21 @@ pub fn create_oss_provider() -> ModelProviderInfo {
port = std::env::var("CODEX_OSS_PORT")
.ok()
.filter(|v| !v.trim().is_empty())
.and_then(|v| v.parse::<u32>().ok())
.unwrap_or(DEFAULT_OLLAMA_PORT)
.and_then(|v| v.parse::<u16>().ok())
.unwrap_or(default_provider_port)
),
};
create_oss_provider_with_base_url(&codex_oss_base_url)
create_oss_provider_with_base_url(&codex_oss_base_url, wire_api)
}
pub fn create_oss_provider_with_base_url(base_url: &str) -> ModelProviderInfo {
pub fn create_oss_provider_with_base_url(base_url: &str, wire_api: WireApi) -> ModelProviderInfo {
ModelProviderInfo {
name: "gpt-oss".into(),
base_url: Some(base_url.into()),
env_key: None,
env_key_instructions: None,
experimental_bearer_token: None,
wire_api: WireApi::Chat,
wire_api,
query_params: None,
http_headers: None,
env_http_headers: None,
codex-rs/core/src/parse_command.rs
+50 -1
View File
@@ -1,16 +1,22 @@
use crate::bash::extract_bash_command;
use crate::bash::try_parse_shell;
use crate::bash::try_parse_word_only_commands_sequence;
use crate::powershell::extract_powershell_command;
use codex_protocol::parse_command::ParsedCommand;
use shlex::split as shlex_split;
use shlex::try_join as shlex_try_join;
use std::path::PathBuf;
fn shlex_join(tokens: &[String]) -> String {
pub fn shlex_join(tokens: &[String]) -> String {
shlex_try_join(tokens.iter().map(String::as_str))
.unwrap_or_else(|_| "<command included NUL byte>".to_string())
}
/// Extracts the shell and script from a command, regardless of platform
pub fn extract_shell_command(command: &[String]) -> Option<(&str, &str)> {
extract_bash_command(command).or_else(|| extract_powershell_command(command))
}
/// DO NOT REVIEW THIS CODE BY HAND
/// This parsing code is quite complex and not easy to hand-modify.
/// The easiest way to iterate is to add unit tests and have Codex fix the implementation.
@@ -877,6 +883,42 @@ mod tests {
}],
);
}
#[test]
fn powershell_command_is_stripped() {
assert_parsed(
&vec_str(&["powershell", "-Command", "Get-ChildItem"]),
vec![ParsedCommand::Unknown {
cmd: "Get-ChildItem".to_string(),
}],
);
}
#[test]
fn pwsh_with_noprofile_and_c_alias_is_stripped() {
assert_parsed(
&vec_str(&["pwsh", "-NoProfile", "-c", "Write-Host hi"]),
vec![ParsedCommand::Unknown {
cmd: "Write-Host hi".to_string(),
}],
);
}
#[test]
fn powershell_with_path_is_stripped() {
let command = if cfg!(windows) {
"C:\\windows\\System32\\WindowsPowerShell\\v1.0\\powershell.exe"
} else {
"/usr/local/bin/powershell.exe"
};
assert_parsed(
&vec_str(&[command, "-NoProfile", "-c", "Write-Host hi"]),
vec![ParsedCommand::Unknown {
cmd: "Write-Host hi".to_string(),
}],
);
}
}
pub fn parse_command_impl(command: &[String]) -> Vec<ParsedCommand> {
@@ -884,6 +926,12 @@ pub fn parse_command_impl(command: &[String]) -> Vec<ParsedCommand> {
return commands;
}
if let Some((_, script)) = extract_powershell_command(command) {
return vec![ParsedCommand::Unknown {
cmd: script.to_string(),
}];
}
let normalized = normalize_tokens(command);
let parts = if contains_connectors(&normalized) {
@@ -1190,6 +1238,7 @@ fn parse_find_query_and_path(tail: &[String]) -> (Option<String>, Option<String>
}
fn parse_shell_lc_commands(original: &[String]) -> Option<Vec<ParsedCommand>> {
// Only handle bash/zsh here; PowerShell is stripped separately without bash parsing.
let (_, script) = extract_bash_command(original)?;
if let Some(tree) = try_parse_shell(script)
codex-rs/core/src/powershell.rs
+93
View File
@@ -0,0 +1,93 @@
use std::path::PathBuf;
use crate::shell::ShellType;
use crate::shell::detect_shell_type;
const POWERSHELL_FLAGS: &[&str] = &["-nologo", "-noprofile", "-command", "-c"];
/// Extract the PowerShell script body from an invocation such as:
///
/// - ["pwsh", "-NoProfile", "-Command", "Get-ChildItem -Recurse | Select-String foo"]
/// - ["powershell.exe", "-Command", "Write-Host hi"]
/// - ["powershell", "-NoLogo", "-NoProfile", "-Command", "...script..."]
///
/// Returns (`shell`, `script`) when the first arg is a PowerShell executable and a
/// `-Command` (or `-c`) flag is present followed by a script string.
pub fn extract_powershell_command(command: &[String]) -> Option<(&str, &str)> {
if command.len() < 3 {
return None;
}
let shell = &command[0];
if detect_shell_type(&PathBuf::from(shell)) != Some(ShellType::PowerShell) {
return None;
}
// Find the first occurrence of -Command (accept common short alias -c as well)
let mut i = 1usize;
while i + 1 < command.len() {
let flag = &command[i];
// Reject unknown flags
if !POWERSHELL_FLAGS.contains(&flag.to_ascii_lowercase().as_str()) {
return None;
}
if flag.eq_ignore_ascii_case("-Command") || flag.eq_ignore_ascii_case("-c") {
let script = &command[i + 1];
return Some((shell, script.as_str()));
}
i += 1;
}
None
}
#[cfg(test)]
mod tests {
use super::extract_powershell_command;
#[test]
fn extracts_basic_powershell_command() {
let cmd = vec![
"powershell".to_string(),
"-Command".to_string(),
"Write-Host hi".to_string(),
];
let (_shell, script) = extract_powershell_command(&cmd).expect("extract");
assert_eq!(script, "Write-Host hi");
}
#[test]
fn extracts_lowercase_flags() {
let cmd = vec![
"powershell".to_string(),
"-nologo".to_string(),
"-command".to_string(),
"Write-Host hi".to_string(),
];
let (_shell, script) = extract_powershell_command(&cmd).expect("extract");
assert_eq!(script, "Write-Host hi");
}
#[test]
fn extracts_full_path_powershell_command() {
let command = if cfg!(windows) {
"C:\\windows\\System32\\WindowsPowerShell\\v1.0\\powershell.exe".to_string()
} else {
"/usr/local/bin/powershell.exe".to_string()
};
let cmd = vec![command, "-Command".to_string(), "Write-Host hi".to_string()];
let (_shell, script) = extract_powershell_command(&cmd).expect("extract");
assert_eq!(script, "Write-Host hi");
}
#[test]
fn extracts_with_noprofile_and_alias() {
let cmd = vec![
"pwsh".to_string(),
"-NoProfile".to_string(),
"-c".to_string(),
"Get-ChildItem | Select-String foo".to_string(),
];
let (_shell, script) = extract_powershell_command(&cmd).expect("extract");
assert_eq!(script, "Get-ChildItem | Select-String foo");
}
}
codex-rs/core/src/response_processing.rs
+23 -57
View File
@@ -13,92 +13,58 @@ pub(crate) async fn process_items(
sess: &Session,
turn_context: &TurnContext,
) -> (Vec<ResponseInputItem>, Vec<ResponseItem>) {
let mut items_to_record_in_conversation_history = Vec::<ResponseItem>::new();
let mut outputs_to_record = Vec::<ResponseItem>::new();
let mut new_inputs_to_record = Vec::<ResponseItem>::new();
let mut responses = Vec::<ResponseInputItem>::new();
for processed_response_item in processed_items {
let crate::codex::ProcessedResponseItem { item, response } = processed_response_item;
match (&item, &response) {
(ResponseItem::Message { role, .. }, None) if role == "assistant" => {
// If the model returned a message, we need to record it.
items_to_record_in_conversation_history.push(item);
}
(
ResponseItem::LocalShellCall { .. },
Some(ResponseInputItem::FunctionCallOutput { call_id, output }),
) => {
items_to_record_in_conversation_history.push(item);
items_to_record_in_conversation_history.push(ResponseItem::FunctionCallOutput {
if let Some(response) = &response {
responses.push(response.clone());
}
match response {
Some(ResponseInputItem::FunctionCallOutput { call_id, output }) => {
new_inputs_to_record.push(ResponseItem::FunctionCallOutput {
call_id: call_id.clone(),
output: output.clone(),
});
}
(
ResponseItem::FunctionCall { .. },
Some(ResponseInputItem::FunctionCallOutput { call_id, output }),
) => {
items_to_record_in_conversation_history.push(item);
items_to_record_in_conversation_history.push(ResponseItem::FunctionCallOutput {
Some(ResponseInputItem::CustomToolCallOutput { call_id, output }) => {
new_inputs_to_record.push(ResponseItem::CustomToolCallOutput {
call_id: call_id.clone(),
output: output.clone(),
});
}
(
ResponseItem::CustomToolCall { .. },
Some(ResponseInputItem::CustomToolCallOutput { call_id, output }),
) => {
items_to_record_in_conversation_history.push(item);
items_to_record_in_conversation_history.push(ResponseItem::CustomToolCallOutput {
call_id: call_id.clone(),
output: output.clone(),
});
}
(
ResponseItem::FunctionCall { .. },
Some(ResponseInputItem::McpToolCallOutput { call_id, result }),
) => {
items_to_record_in_conversation_history.push(item);
Some(ResponseInputItem::McpToolCallOutput { call_id, result }) => {
let output = match result {
Ok(call_tool_result) => FunctionCallOutputPayload::from(call_tool_result),
Ok(call_tool_result) => FunctionCallOutputPayload::from(&call_tool_result),
Err(err) => FunctionCallOutputPayload {
content: err.clone(),
success: Some(false),
..Default::default()
},
};
items_to_record_in_conversation_history.push(ResponseItem::FunctionCallOutput {
new_inputs_to_record.push(ResponseItem::FunctionCallOutput {
call_id: call_id.clone(),
output,
});
}
(
ResponseItem::Reasoning {
id,
summary,
content,
encrypted_content,
},
None,
) => {
items_to_record_in_conversation_history.push(ResponseItem::Reasoning {
id: id.clone(),
summary: summary.clone(),
content: content.clone(),
encrypted_content: encrypted_content.clone(),
});
}
None => {}
_ => {
warn!("Unexpected response item: {item:?} with response: {response:?}");
}
};
if let Some(response) = response {
responses.push(response);
}
outputs_to_record.push(item);
}
let all_items_to_record = [outputs_to_record, new_inputs_to_record].concat();
// Only attempt to take the lock if there is something to record.
if !items_to_record_in_conversation_history.is_empty() {
sess.record_conversation_items(turn_context, &items_to_record_in_conversation_history)
if !all_items_to_record.is_empty() {
sess.record_conversation_items(turn_context, &all_items_to_record)
.await;
}
(responses, items_to_record_in_conversation_history)
(responses, all_items_to_record)
}
codex-rs/core/src/rollout/policy.rs
+2
View File
@@ -72,6 +72,8 @@ pub(crate) fn should_persist_event_msg(ev: &EventMsg) -> bool {
| EventMsg::GetHistoryEntryResponse(_)
| EventMsg::UndoStarted(_)
| EventMsg::McpListToolsResponse(_)
| EventMsg::McpStartupUpdate(_)
| EventMsg::McpStartupComplete(_)
| EventMsg::ListCustomPromptsResponse(_)
| EventMsg::PlanUpdate(_)
| EventMsg::ShutdownComplete
codex-rs/core/src/shell.rs
+235 -338
View File
@@ -2,22 +2,26 @@ use serde::Deserialize;
use serde::Serialize;
use std::path::PathBuf;
#[derive(Debug, PartialEq, Eq, Clone, Serialize, Deserialize)]
pub enum ShellType {
Zsh,
Bash,
PowerShell,
}
#[derive(Debug, PartialEq, Eq, Clone, Serialize, Deserialize)]
pub struct ZshShell {
pub(crate) shell_path: String,
pub(crate) zshrc_path: String,
pub(crate) shell_path: PathBuf,
}
#[derive(Debug, PartialEq, Eq, Clone, Serialize, Deserialize)]
pub struct BashShell {
pub(crate) shell_path: String,
pub(crate) bashrc_path: String,
pub(crate) shell_path: PathBuf,
}
#[derive(Debug, PartialEq, Eq, Clone, Serialize, Deserialize)]
pub struct PowerShellConfig {
pub(crate) exe: String, // Executable name or path, e.g. "pwsh" or "powershell.exe".
pub(crate) bash_exe_fallback: Option<PathBuf>, // In case the model generates a bash command.
pub(crate) shell_path: PathBuf, // Executable name or path, e.g. "pwsh" or "powershell.exe".
}
#[derive(Debug, PartialEq, Eq, Clone, Serialize, Deserialize)]
@@ -36,7 +40,10 @@ impl Shell {
.file_name()
.map(|s| s.to_string_lossy().to_string())
}
Shell::PowerShell(ps) => Some(ps.exe.clone()),
Shell::PowerShell(ps) => ps
.shell_path
.file_stem()
.map(|s| s.to_string_lossy().to_string()),
Shell::Unknown => None,
}
}
@@ -47,10 +54,14 @@ impl Shell {
match self {
Shell::Zsh(ZshShell { shell_path, .. }) | Shell::Bash(BashShell { shell_path, .. }) => {
let arg = if use_login_shell { "-lc" } else { "-c" };
vec![shell_path.clone(), arg.to_string(), command.to_string()]
vec![
shell_path.to_string_lossy().to_string(),
arg.to_string(),
command.to_string(),
]
}
Shell::PowerShell(ps) => {
let mut args = vec![ps.exe.clone(), "-NoLogo".to_string()];
let mut args = vec![ps.shell_path.to_string_lossy().to_string()];
if !use_login_shell {
args.push("-NoProfile".to_string());
}
@@ -65,7 +76,7 @@ impl Shell {
}
#[cfg(unix)]
fn detect_default_user_shell() -> Shell {
fn get_user_shell_path() -> Option<PathBuf> {
use libc::getpwuid;
use libc::getuid;
use std::ffi::CStr;
@@ -78,75 +89,181 @@ fn detect_default_user_shell() -> Shell {
let shell_path = CStr::from_ptr((*pw).pw_shell)
.to_string_lossy()
.into_owned();
let home_path = CStr::from_ptr((*pw).pw_dir).to_string_lossy().into_owned();
Some(PathBuf::from(shell_path))
} else {
None
}
}
}
if shell_path.ends_with("/zsh") {
return Shell::Zsh(ZshShell {
shell_path,
zshrc_path: format!("{home_path}/.zshrc"),
});
}
#[cfg(not(unix))]
fn get_user_shell_path() -> Option<PathBuf> {
None
}
if shell_path.ends_with("/bash") {
return Shell::Bash(BashShell {
shell_path,
bashrc_path: format!("{home_path}/.bashrc"),
});
fn file_exists(path: &PathBuf) -> Option<PathBuf> {
if std::fs::metadata(path).is_ok_and(|metadata| metadata.is_file()) {
Some(PathBuf::from(path))
} else {
None
}
}
fn get_shell_path(
shell_type: ShellType,
provided_path: Option<&PathBuf>,
binary_name: &str,
fallback_paths: Vec<&str>,
) -> Option<PathBuf> {
// If exact provided path exists, use it
if provided_path.and_then(file_exists).is_some() {
return provided_path.cloned();
}
// Check if the shell we are trying to load is user's default shell
// if just use it
let default_shell_path = get_user_shell_path();
if let Some(default_shell_path) = default_shell_path
&& detect_shell_type(&default_shell_path) == Some(shell_type)
{
return Some(default_shell_path);
}
if let Ok(path) = which::which(binary_name) {
return Some(path);
}
for path in fallback_paths {
//check exists
if let Some(path) = file_exists(&PathBuf::from(path)) {
return Some(path);
}
}
None
}
fn get_zsh_shell(path: Option<&PathBuf>) -> Option<ZshShell> {
let shell_path = get_shell_path(ShellType::Zsh, path, "zsh", vec!["/bin/zsh"]);
shell_path.map(|shell_path| ZshShell { shell_path })
}
fn get_bash_shell(path: Option<&PathBuf>) -> Option<BashShell> {
let shell_path = get_shell_path(ShellType::Bash, path, "bash", vec!["/bin/bash"]);
shell_path.map(|shell_path| BashShell { shell_path })
}
fn get_powershell_shell(path: Option<&PathBuf>) -> Option<PowerShellConfig> {
let shell_path = get_shell_path(
ShellType::PowerShell,
path,
"pwsh",
vec!["/usr/local/bin/pwsh"],
)
.or_else(|| get_shell_path(ShellType::PowerShell, path, "powershell", vec![]));
shell_path.map(|shell_path| PowerShellConfig { shell_path })
}
pub fn get_shell_by_model_provided_path(shell_path: &PathBuf) -> Shell {
detect_shell_type(shell_path)
.and_then(|shell_type| get_shell(shell_type, Some(shell_path)))
.unwrap_or(Shell::Unknown)
}
pub fn get_shell(shell_type: ShellType, path: Option<&PathBuf>) -> Option<Shell> {
match shell_type {
ShellType::Zsh => get_zsh_shell(path).map(Shell::Zsh),
ShellType::Bash => get_bash_shell(path).map(Shell::Bash),
ShellType::PowerShell => get_powershell_shell(path).map(Shell::PowerShell),
}
}
pub fn detect_shell_type(shell_path: &PathBuf) -> Option<ShellType> {
match shell_path.as_os_str().to_str() {
Some("zsh") => Some(ShellType::Zsh),
Some("bash") => Some(ShellType::Bash),
Some("pwsh") => Some(ShellType::PowerShell),
Some("powershell") => Some(ShellType::PowerShell),
_ => {
let shell_name = shell_path.file_stem();
if let Some(shell_name) = shell_name
&& shell_name != shell_path
{
detect_shell_type(&PathBuf::from(shell_name))
} else {
None
}
}
}
Shell::Unknown
}
#[cfg(unix)]
pub async fn default_user_shell() -> Shell {
detect_default_user_shell()
}
#[cfg(target_os = "windows")]
pub async fn default_user_shell() -> Shell {
use tokio::process::Command;
// Prefer PowerShell 7+ (`pwsh`) if available, otherwise fall back to Windows PowerShell.
let has_pwsh = Command::new("pwsh")
.arg("-NoLogo")
.arg("-NoProfile")
.arg("-Command")
.arg("$PSVersionTable.PSVersion.Major")
.output()
.await
.map(|o| o.status.success())
.unwrap_or(false);
let bash_exe = if Command::new("bash.exe")
.arg("--version")
.stdin(std::process::Stdio::null())
.output()
.await
.ok()
.map(|o| o.status.success())
.unwrap_or(false)
{
which::which("bash.exe").ok()
if cfg!(windows) {
get_shell(ShellType::PowerShell, None).unwrap_or(Shell::Unknown)
} else {
None
};
if has_pwsh {
Shell::PowerShell(PowerShellConfig {
exe: "pwsh.exe".to_string(),
bash_exe_fallback: bash_exe,
})
} else {
Shell::PowerShell(PowerShellConfig {
exe: "powershell.exe".to_string(),
bash_exe_fallback: bash_exe,
})
get_user_shell_path()
.and_then(|shell| detect_shell_type(&shell))
.and_then(|shell_type| get_shell(shell_type, None))
.unwrap_or(Shell::Unknown)
}
}
#[cfg(all(not(target_os = "windows"), not(unix)))]
pub async fn default_user_shell() -> Shell {
Shell::Unknown
#[cfg(test)]
mod detect_shell_type_tests {
use super::*;
#[test]
fn test_detect_shell_type() {
assert_eq!(
detect_shell_type(&PathBuf::from("zsh")),
Some(ShellType::Zsh)
);
assert_eq!(
detect_shell_type(&PathBuf::from("bash")),
Some(ShellType::Bash)
);
assert_eq!(
detect_shell_type(&PathBuf::from("pwsh")),
Some(ShellType::PowerShell)
);
assert_eq!(
detect_shell_type(&PathBuf::from("powershell")),
Some(ShellType::PowerShell)
);
assert_eq!(detect_shell_type(&PathBuf::from("fish")), None);
assert_eq!(detect_shell_type(&PathBuf::from("other")), None);
assert_eq!(
detect_shell_type(&PathBuf::from("/bin/zsh")),
Some(ShellType::Zsh)
);
assert_eq!(
detect_shell_type(&PathBuf::from("/bin/bash")),
Some(ShellType::Bash)
);
assert_eq!(
detect_shell_type(&PathBuf::from("powershell.exe")),
Some(ShellType::PowerShell)
);
assert_eq!(
detect_shell_type(&PathBuf::from(if cfg!(windows) {
"C:\\windows\\System32\\WindowsPowerShell\\v1.0\\powershell.exe"
} else {
"/usr/local/bin/pwsh"
})),
Some(ShellType::PowerShell)
);
assert_eq!(
detect_shell_type(&PathBuf::from("pwsh.exe")),
Some(ShellType::PowerShell)
);
assert_eq!(
detect_shell_type(&PathBuf::from("/usr/local/bin/pwsh")),
Some(ShellType::PowerShell)
);
}
}
#[cfg(test)]
@@ -156,6 +273,34 @@ mod tests {
use std::path::PathBuf;
use std::process::Command;
#[test]
#[cfg(target_os = "macos")]
fn detects_zsh() {
let zsh_shell = get_shell(ShellType::Zsh, None).unwrap();
let ZshShell { shell_path } = match zsh_shell {
Shell::Zsh(zsh_shell) => zsh_shell,
_ => panic!("expected zsh shell"),
};
assert_eq!(shell_path, PathBuf::from("/bin/zsh"));
}
#[test]
fn detects_bash() {
let bash_shell = get_shell(ShellType::Bash, None).unwrap();
let BashShell { shell_path } = match bash_shell {
Shell::Bash(bash_shell) => bash_shell,
_ => panic!("expected bash shell"),
};
assert!(
shell_path == PathBuf::from("/bin/bash")
|| shell_path == PathBuf::from("/usr/bin/bash"),
"shell path: {shell_path:?}",
);
}
#[tokio::test]
async fn test_current_shell_detects_zsh() {
let shell = Command::new("sh")
@@ -164,292 +309,44 @@ mod tests {
.output()
.unwrap();
let home = std::env::var("HOME").unwrap();
let shell_path = String::from_utf8_lossy(&shell.stdout).trim().to_string();
if shell_path.ends_with("/zsh") {
assert_eq!(
default_user_shell().await,
Shell::Zsh(ZshShell {
shell_path: shell_path.to_string(),
zshrc_path: format!("{home}/.zshrc",),
shell_path: PathBuf::from(shell_path),
})
);
}
}
#[tokio::test]
async fn test_run_with_profile_bash_escaping_and_execution() {
let shell_path = "/bin/bash";
let cases = vec![
(
vec!["myecho"],
vec![shell_path, "-lc", "source BASHRC_PATH && (myecho)"],
Some("It works!\n"),
),
(
vec!["bash", "-lc", "echo 'single' \"double\""],
vec![
shell_path,
"-lc",
"source BASHRC_PATH && (echo 'single' \"double\")",
],
Some("single double\n"),
),
];
for (input, expected_cmd, expected_output) in cases {
use std::collections::HashMap;
use crate::exec::ExecParams;
use crate::exec::SandboxType;
use crate::exec::process_exec_tool_call;
use crate::protocol::SandboxPolicy;
let temp_home = tempfile::tempdir().unwrap();
let bashrc_path = temp_home.path().join(".bashrc");
std::fs::write(
&bashrc_path,
r#"
set -x
function myecho {
echo 'It works!'
}
"#,
)
.unwrap();
let command = expected_cmd
.iter()
.map(|s| s.replace("BASHRC_PATH", bashrc_path.to_str().unwrap()))
.collect::<Vec<_>>();
let output = process_exec_tool_call(
ExecParams {
command: command.clone(),
cwd: PathBuf::from(temp_home.path()),
timeout_ms: None,
env: HashMap::from([(
"HOME".to_string(),
temp_home.path().to_str().unwrap().to_string(),
)]),
with_escalated_permissions: None,
justification: None,
arg0: None,
},
SandboxType::None,
&SandboxPolicy::DangerFullAccess,
temp_home.path(),
&None,
None,
)
.await
.unwrap();
assert_eq!(output.exit_code, 0, "input: {input:?} output: {output:?}");
if let Some(expected) = expected_output {
assert_eq!(
output.stdout.text, expected,
"input: {input:?} output: {output:?}"
);
}
async fn detects_powershell_as_default() {
if !cfg!(windows) {
return;
}
let powershell_shell = default_user_shell().await;
let PowerShellConfig { shell_path } = match powershell_shell {
Shell::PowerShell(powershell_shell) => powershell_shell,
_ => panic!("expected powershell shell"),
};
assert!(shell_path.ends_with("pwsh.exe") || shell_path.ends_with("powershell.exe"));
}
}
#[cfg(test)]
#[cfg(target_os = "macos")]
mod macos_tests {
use std::path::PathBuf;
#[tokio::test]
async fn test_run_with_profile_escaping_and_execution() {
let shell_path = "/bin/zsh";
let cases = vec![
(
vec!["myecho"],
vec![shell_path, "-lc", "source ZSHRC_PATH && (myecho)"],
Some("It works!\n"),
),
(
vec!["myecho"],
vec![shell_path, "-lc", "source ZSHRC_PATH && (myecho)"],
Some("It works!\n"),
),
(
vec!["bash", "-c", "echo 'single' \"double\""],
vec![
shell_path,
"-lc",
"source ZSHRC_PATH && (bash -c \"echo 'single' \\\"double\\\"\")",
],
Some("single double\n"),
),
(
vec!["bash", "-lc", "echo 'single' \"double\""],
vec![
shell_path,
"-lc",
"source ZSHRC_PATH && (echo 'single' \"double\")",
],
Some("single double\n"),
),
];
for (input, expected_cmd, expected_output) in cases {
use std::collections::HashMap;
use crate::exec::ExecParams;
use crate::exec::SandboxType;
use crate::exec::process_exec_tool_call;
use crate::protocol::SandboxPolicy;
let temp_home = tempfile::tempdir().unwrap();
let zshrc_path = temp_home.path().join(".zshrc");
std::fs::write(
&zshrc_path,
r#"
set -x
function myecho {
echo 'It works!'
}
"#,
)
.unwrap();
let command = expected_cmd
.iter()
.map(|s| s.replace("ZSHRC_PATH", zshrc_path.to_str().unwrap()))
.collect::<Vec<_>>();
let output = process_exec_tool_call(
ExecParams {
command: command.clone(),
cwd: PathBuf::from(temp_home.path()),
timeout_ms: None,
env: HashMap::from([(
"HOME".to_string(),
temp_home.path().to_str().unwrap().to_string(),
)]),
with_escalated_permissions: None,
justification: None,
arg0: None,
},
SandboxType::None,
&SandboxPolicy::DangerFullAccess,
temp_home.path(),
&None,
None,
)
.await
.unwrap();
assert_eq!(output.exit_code, 0, "input: {input:?} output: {output:?}");
if let Some(expected) = expected_output {
assert_eq!(
output.stdout.text, expected,
"input: {input:?} output: {output:?}"
);
}
}
}
}
#[cfg(test)]
#[cfg(target_os = "windows")]
mod tests_windows {
use super::*;
#[test]
fn test_format_default_shell_invocation_powershell() {
use std::path::PathBuf;
let cases = vec![
(
PowerShellConfig {
exe: "pwsh.exe".to_string(),
bash_exe_fallback: None,
},
vec!["bash", "-lc", "echo hello"],
vec!["pwsh.exe", "-NoProfile", "-Command", "echo hello"],
),
(
PowerShellConfig {
exe: "powershell.exe".to_string(),
bash_exe_fallback: None,
},
vec!["bash", "-lc", "echo hello"],
vec!["powershell.exe", "-NoProfile", "-Command", "echo hello"],
),
(
PowerShellConfig {
exe: "pwsh.exe".to_string(),
bash_exe_fallback: Some(PathBuf::from("bash.exe")),
},
vec!["bash", "-lc", "echo hello"],
vec!["bash.exe", "-lc", "echo hello"],
),
(
PowerShellConfig {
exe: "pwsh.exe".to_string(),
bash_exe_fallback: Some(PathBuf::from("bash.exe")),
},
vec![
"bash",
"-lc",
"apply_patch <<'EOF'\n*** Begin Patch\n*** Update File: destination_file.txt\n-original content\n+modified content\n*** End Patch\nEOF",
],
vec![
"bash.exe",
"-lc",
"apply_patch <<'EOF'\n*** Begin Patch\n*** Update File: destination_file.txt\n-original content\n+modified content\n*** End Patch\nEOF",
],
),
(
PowerShellConfig {
exe: "pwsh.exe".to_string(),
bash_exe_fallback: Some(PathBuf::from("bash.exe")),
},
vec!["echo", "hello"],
vec!["pwsh.exe", "-NoProfile", "-Command", "echo hello"],
),
(
PowerShellConfig {
exe: "pwsh.exe".to_string(),
bash_exe_fallback: Some(PathBuf::from("bash.exe")),
},
vec!["pwsh.exe", "-NoProfile", "-Command", "echo hello"],
vec!["pwsh.exe", "-NoProfile", "-Command", "echo hello"],
),
(
PowerShellConfig {
exe: "powershell.exe".to_string(),
bash_exe_fallback: Some(PathBuf::from("bash.exe")),
},
vec![
"codex-mcp-server.exe",
"--codex-run-as-apply-patch",
"*** Begin Patch\n*** Update File: C:\\Users\\person\\destination_file.txt\n-original content\n+modified content\n*** End Patch",
],
vec![
"codex-mcp-server.exe",
"--codex-run-as-apply-patch",
"*** Begin Patch\n*** Update File: C:\\Users\\person\\destination_file.txt\n-original content\n+modified content\n*** End Patch",
],
),
];
for (config, input, expected_cmd) in cases {
let command = expected_cmd
.iter()
.map(|s| (*s).to_string())
.collect::<Vec<_>>();
// These tests assert the final command for each scenario now that the helper
// has been removed. The inputs remain to document the original coverage.
let expected = expected_cmd
.iter()
.map(|s| (*s).to_string())
.collect::<Vec<_>>();
assert_eq!(command, expected, "input: {input:?} config: {config:?}");
fn finds_poweshell() {
if !cfg!(windows) {
return;
}
let powershell_shell = get_shell(ShellType::PowerShell, None).unwrap();
let PowerShellConfig { shell_path } = match powershell_shell {
Shell::PowerShell(powershell_shell) => powershell_shell,
_ => panic!("expected powershell shell"),
};
assert!(shell_path.ends_with("pwsh.exe") || shell_path.ends_with("powershell.exe"));
}
}
codex-rs/core/src/state/service.rs
+4 -1
View File
@@ -8,9 +8,12 @@ use crate::unified_exec::UnifiedExecSessionManager;
use crate::user_notification::UserNotifier;
use codex_otel::otel_event_manager::OtelEventManager;
use tokio::sync::Mutex;
use tokio::sync::RwLock;
use tokio_util::sync::CancellationToken;
pub(crate) struct SessionServices {
pub(crate) mcp_connection_manager: McpConnectionManager,
pub(crate) mcp_connection_manager: Arc<RwLock<McpConnectionManager>>,
pub(crate) mcp_startup_cancellation_token: CancellationToken,
pub(crate) unified_exec_manager: UnifiedExecSessionManager,
pub(crate) notifier: UserNotifier,
pub(crate) rollout: Mutex<Option<RolloutRecorder>>,
codex-rs/core/src/state/session.rs
+4
View File
@@ -42,6 +42,10 @@ impl SessionState {
self.history.replace(items);
}
pub(crate) fn set_token_info(&mut self, info: Option<TokenUsageInfo>) {
self.history.set_token_info(info);
}
// Token/rate limit helpers
pub(crate) fn update_token_info_from_usage(
&mut self,
codex-rs/core/src/tasks/user_shell.rs
+30 -8
View File
@@ -20,6 +20,7 @@ use crate::parse_command::parse_command;
use crate::protocol::EventMsg;
use crate::protocol::ExecCommandBeginEvent;
use crate::protocol::ExecCommandEndEvent;
use crate::protocol::ExecCommandSource;
use crate::protocol::SandboxPolicy;
use crate::protocol::TaskStartedEvent;
use crate::sandboxing::ExecEnv;
@@ -64,30 +65,33 @@ impl SessionTask for UserShellCommandTask {
// allows commands that use shell features (pipes, &&, redirects, etc.).
// We do not source rc files or otherwise reformat the script.
let use_login_shell = true;
let shell_invocation = session
let command = session
.user_shell()
.derive_exec_args(&self.command, use_login_shell);
let call_id = Uuid::new_v4().to_string();
let raw_command = self.command.clone();
let cwd = turn_context.cwd.clone();
let parsed_cmd = parse_command(&shell_invocation);
let parsed_cmd = parse_command(&command);
session
.send_event(
turn_context.as_ref(),
EventMsg::ExecCommandBegin(ExecCommandBeginEvent {
call_id: call_id.clone(),
command: shell_invocation.clone(),
cwd: turn_context.cwd.clone(),
parsed_cmd,
is_user_shell_command: true,
turn_id: turn_context.sub_id.clone(),
command: command.clone(),
cwd: cwd.clone(),
parsed_cmd: parsed_cmd.clone(),
source: ExecCommandSource::UserShell,
interaction_input: None,
}),
)
.await;
let exec_env = ExecEnv {
command: shell_invocation,
cwd: turn_context.cwd.clone(),
command: command.clone(),
cwd: cwd.clone(),
env: create_env(&turn_context.shell_environment_policy),
timeout_ms: None,
sandbox: SandboxType::None,
@@ -127,6 +131,12 @@ impl SessionTask for UserShellCommandTask {
turn_context.as_ref(),
EventMsg::ExecCommandEnd(ExecCommandEndEvent {
call_id,
turn_id: turn_context.sub_id.clone(),
command: command.clone(),
cwd: cwd.clone(),
parsed_cmd: parsed_cmd.clone(),
source: ExecCommandSource::UserShell,
interaction_input: None,
stdout: String::new(),
stderr: aborted_message.clone(),
aggregated_output: aborted_message.clone(),
@@ -143,6 +153,12 @@ impl SessionTask for UserShellCommandTask {
turn_context.as_ref(),
EventMsg::ExecCommandEnd(ExecCommandEndEvent {
call_id: call_id.clone(),
turn_id: turn_context.sub_id.clone(),
command: command.clone(),
cwd: cwd.clone(),
parsed_cmd: parsed_cmd.clone(),
source: ExecCommandSource::UserShell,
interaction_input: None,
stdout: output.stdout.text.clone(),
stderr: output.stderr.text.clone(),
aggregated_output: output.aggregated_output.text.clone(),
@@ -174,6 +190,12 @@ impl SessionTask for UserShellCommandTask {
turn_context.as_ref(),
EventMsg::ExecCommandEnd(ExecCommandEndEvent {
call_id,
turn_id: turn_context.sub_id.clone(),
command,
cwd,
parsed_cmd,
source: ExecCommandSource::UserShell,
interaction_input: None,
stdout: exec_output.stdout.text.clone(),
stderr: exec_output.stderr.text.clone(),
aggregated_output: exec_output.aggregated_output.text.clone(),
codex-rs/core/src/tools/context.rs
-26
View File
@@ -4,17 +4,13 @@ use crate::tools::TELEMETRY_PREVIEW_MAX_BYTES;
use crate::tools::TELEMETRY_PREVIEW_MAX_LINES;
use crate::tools::TELEMETRY_PREVIEW_TRUNCATION_NOTICE;
use crate::turn_diff_tracker::TurnDiffTracker;
use codex_otel::otel_event_manager::OtelEventManager;
use codex_protocol::models::FunctionCallOutputContentItem;
use codex_protocol::models::FunctionCallOutputPayload;
use codex_protocol::models::ResponseInputItem;
use codex_protocol::models::ShellToolCallParams;
use codex_protocol::protocol::FileChange;
use codex_utils_string::take_bytes_at_char_boundary;
use mcp_types::CallToolResult;
use std::borrow::Cow;
use std::collections::HashMap;
use std::path::PathBuf;
use std::sync::Arc;
use tokio::sync::Mutex;
@@ -244,25 +240,3 @@ mod tests {
assert_eq!(lines.last(), Some(&TELEMETRY_PREVIEW_TRUNCATION_NOTICE));
}
}
#[derive(Clone, Debug)]
#[allow(dead_code)]
pub(crate) struct ExecCommandContext {
pub(crate) turn: Arc<TurnContext>,
pub(crate) call_id: String,
pub(crate) command_for_display: Vec<String>,
pub(crate) cwd: PathBuf,
pub(crate) apply_patch: Option<ApplyPatchCommandContext>,
pub(crate) tool_name: String,
pub(crate) otel_event_manager: OtelEventManager,
// TODO(abhisek-oai): Find a better way to track this.
// https://github.com/openai/codex/pull/2471/files#r2470352242
pub(crate) is_user_shell_command: bool,
}
#[derive(Clone, Debug)]
#[allow(dead_code)]
pub(crate) struct ApplyPatchCommandContext {
pub(crate) user_explicitly_approved_this_action: bool,
pub(crate) changes: HashMap<PathBuf, FileChange>,
}
codex-rs/core/src/tools/events.rs
+142 -96
View File
@@ -8,12 +8,14 @@ use crate::parse_command::parse_command;
use crate::protocol::EventMsg;
use crate::protocol::ExecCommandBeginEvent;
use crate::protocol::ExecCommandEndEvent;
use crate::protocol::ExecCommandSource;
use crate::protocol::FileChange;
use crate::protocol::PatchApplyBeginEvent;
use crate::protocol::PatchApplyEndEvent;
use crate::protocol::TurnDiffEvent;
use crate::tools::context::SharedTurnDiffTracker;
use crate::tools::sandboxing::ToolError;
use codex_protocol::parse_command::ParsedCommand;
use std::collections::HashMap;
use std::path::Path;
use std::path::PathBuf;
@@ -60,17 +62,21 @@ pub(crate) async fn emit_exec_command_begin(
ctx: ToolEventCtx<'_>,
command: &[String],
cwd: &Path,
is_user_shell_command: bool,
parsed_cmd: &[ParsedCommand],
source: ExecCommandSource,
interaction_input: Option<String>,
) {
ctx.session
.send_event(
ctx.turn,
EventMsg::ExecCommandBegin(ExecCommandBeginEvent {
call_id: ctx.call_id.to_string(),
turn_id: ctx.turn.sub_id.clone(),
command: command.to_vec(),
cwd: cwd.to_path_buf(),
parsed_cmd: parse_command(command),
is_user_shell_command,
parsed_cmd: parsed_cmd.to_vec(),
source,
interaction_input,
}),
)
.await;
@@ -80,27 +86,30 @@ pub(crate) enum ToolEmitter {
Shell {
command: Vec<String>,
cwd: PathBuf,
is_user_shell_command: bool,
source: ExecCommandSource,
parsed_cmd: Vec<ParsedCommand>,
},
ApplyPatch {
changes: HashMap<PathBuf, FileChange>,
auto_approved: bool,
},
UnifiedExec {
command: String,
command: Vec<String>,
cwd: PathBuf,
// True for `exec_command` and false for `write_stdin`.
#[allow(dead_code)]
is_startup_command: bool,
source: ExecCommandSource,
interaction_input: Option<String>,
parsed_cmd: Vec<ParsedCommand>,
},
}
impl ToolEmitter {
pub fn shell(command: Vec<String>, cwd: PathBuf, is_user_shell_command: bool) -> Self {
pub fn shell(command: Vec<String>, cwd: PathBuf, source: ExecCommandSource) -> Self {
let parsed_cmd = parse_command(&command);
Self::Shell {
command,
cwd,
is_user_shell_command,
source,
parsed_cmd,
}
}
@@ -111,11 +120,19 @@ impl ToolEmitter {
}
}
pub fn unified_exec(command: String, cwd: PathBuf, is_startup_command: bool) -> Self {
pub fn unified_exec(
command: &[String],
cwd: PathBuf,
source: ExecCommandSource,
interaction_input: Option<String>,
) -> Self {
let parsed_cmd = parse_command(command);
Self::UnifiedExec {
command,
command: command.to_vec(),
cwd,
is_startup_command,
source,
interaction_input,
parsed_cmd,
}
}
@@ -125,45 +142,15 @@ impl ToolEmitter {
Self::Shell {
command,
cwd,
is_user_shell_command,
source,
parsed_cmd,
},
ToolEventStage::Begin,
stage,
) => {
emit_exec_command_begin(ctx, command, cwd.as_path(), *is_user_shell_command).await;
}
(Self::Shell { .. }, ToolEventStage::Success(output)) => {
emit_exec_end(
emit_exec_stage(
ctx,
output.stdout.text.clone(),
output.stderr.text.clone(),
output.aggregated_output.text.clone(),
output.exit_code,
output.duration,
format_exec_output_str(&output),
)
.await;
}
(Self::Shell { .. }, ToolEventStage::Failure(ToolEventFailure::Output(output))) => {
emit_exec_end(
ctx,
output.stdout.text.clone(),
output.stderr.text.clone(),
output.aggregated_output.text.clone(),
output.exit_code,
output.duration,
format_exec_output_str(&output),
)
.await;
}
(Self::Shell { .. }, ToolEventStage::Failure(ToolEventFailure::Message(message))) => {
emit_exec_end(
ctx,
String::new(),
(*message).to_string(),
(*message).to_string(),
-1,
Duration::ZERO,
message.clone(),
ExecCommandInput::new(command, cwd.as_path(), parsed_cmd, *source, None),
stage,
)
.await;
}
@@ -217,48 +204,26 @@ impl ToolEmitter {
) => {
emit_patch_end(ctx, String::new(), (*message).to_string(), false).await;
}
(Self::UnifiedExec { command, cwd, .. }, ToolEventStage::Begin) => {
emit_exec_command_begin(ctx, &[command.to_string()], cwd.as_path(), false).await;
}
(Self::UnifiedExec { .. }, ToolEventStage::Success(output)) => {
emit_exec_end(
ctx,
output.stdout.text.clone(),
output.stderr.text.clone(),
output.aggregated_output.text.clone(),
output.exit_code,
output.duration,
format_exec_output_str(&output),
)
.await;
}
(
Self::UnifiedExec { .. },
ToolEventStage::Failure(ToolEventFailure::Output(output)),
Self::UnifiedExec {
command,
cwd,
source,
interaction_input,
parsed_cmd,
},
stage,
) => {
emit_exec_end(
emit_exec_stage(
ctx,
output.stdout.text.clone(),
output.stderr.text.clone(),
output.aggregated_output.text.clone(),
output.exit_code,
output.duration,
format_exec_output_str(&output),
)
.await;
}
(
Self::UnifiedExec { .. },
ToolEventStage::Failure(ToolEventFailure::Message(message)),
) => {
emit_exec_end(
ctx,
String::new(),
(*message).to_string(),
(*message).to_string(),
-1,
Duration::ZERO,
message.clone(),
ExecCommandInput::new(
command,
cwd.as_path(),
parsed_cmd,
*source,
interaction_input.as_deref(),
),
stage,
)
.await;
}
@@ -317,26 +282,107 @@ impl ToolEmitter {
}
}
async fn emit_exec_end(
ctx: ToolEventCtx<'_>,
struct ExecCommandInput<'a> {
command: &'a [String],
cwd: &'a Path,
parsed_cmd: &'a [ParsedCommand],
source: ExecCommandSource,
interaction_input: Option<&'a str>,
}
impl<'a> ExecCommandInput<'a> {
fn new(
command: &'a [String],
cwd: &'a Path,
parsed_cmd: &'a [ParsedCommand],
source: ExecCommandSource,
interaction_input: Option<&'a str>,
) -> Self {
Self {
command,
cwd,
parsed_cmd,
source,
interaction_input,
}
}
}
struct ExecCommandResult {
stdout: String,
stderr: String,
aggregated_output: String,
exit_code: i32,
duration: Duration,
formatted_output: String,
}
async fn emit_exec_stage(
ctx: ToolEventCtx<'_>,
exec_input: ExecCommandInput<'_>,
stage: ToolEventStage,
) {
match stage {
ToolEventStage::Begin => {
emit_exec_command_begin(
ctx,
exec_input.command,
exec_input.cwd,
exec_input.parsed_cmd,
exec_input.source,
exec_input.interaction_input.map(str::to_owned),
)
.await;
}
ToolEventStage::Success(output)
| ToolEventStage::Failure(ToolEventFailure::Output(output)) => {
let exec_result = ExecCommandResult {
stdout: output.stdout.text.clone(),
stderr: output.stderr.text.clone(),
aggregated_output: output.aggregated_output.text.clone(),
exit_code: output.exit_code,
duration: output.duration,
formatted_output: format_exec_output_str(&output),
};
emit_exec_end(ctx, exec_input, exec_result).await;
}
ToolEventStage::Failure(ToolEventFailure::Message(message)) => {
let text = message.to_string();
let exec_result = ExecCommandResult {
stdout: String::new(),
stderr: text.clone(),
aggregated_output: text.clone(),
exit_code: -1,
duration: Duration::ZERO,
formatted_output: text,
};
emit_exec_end(ctx, exec_input, exec_result).await;
}
}
}
async fn emit_exec_end(
ctx: ToolEventCtx<'_>,
exec_input: ExecCommandInput<'_>,
exec_result: ExecCommandResult,
) {
ctx.session
.send_event(
ctx.turn,
EventMsg::ExecCommandEnd(ExecCommandEndEvent {
call_id: ctx.call_id.to_string(),
stdout,
stderr,
aggregated_output,
exit_code,
duration,
formatted_output,
turn_id: ctx.turn.sub_id.clone(),
command: exec_input.command.to_vec(),
cwd: exec_input.cwd.to_path_buf(),
parsed_cmd: exec_input.parsed_cmd.to_vec(),
source: exec_input.source,
interaction_input: exec_input.interaction_input.map(str::to_owned),
stdout: exec_result.stdout,
stderr: exec_result.stderr,
aggregated_output: exec_result.aggregated_output,
exit_code: exec_result.exit_code,
duration: exec_result.duration,
formatted_output: exec_result.formatted_output,
}),
)
.await;
codex-rs/core/src/tools/handlers/mcp_resource.rs
+4
View File
@@ -287,6 +287,8 @@ async fn handle_list_resources(
let resources = session
.services
.mcp_connection_manager
.read()
.await
.list_all_resources()
.await;
Ok(ListResourcesPayload::from_all_servers(resources))
@@ -396,6 +398,8 @@ async fn handle_list_resource_templates(
let templates = session
.services
.mcp_connection_manager
.read()
.await
.list_all_resource_templates()
.await;
Ok(ListResourceTemplatesPayload::from_all_servers(templates))
codex-rs/core/src/tools/handlers/shell.rs
+17 -19
View File
@@ -11,6 +11,7 @@ use crate::exec::ExecParams;
use crate::exec_env::create_env;
use crate::function_tool::FunctionCallError;
use crate::is_safe_command::is_known_safe_command;
use crate::protocol::ExecCommandSource;
use crate::tools::context::ToolInvocation;
use crate::tools::context::ToolOutput;
use crate::tools::context::ToolPayload;
@@ -285,11 +286,13 @@ impl ShellHandler {
}
// Regular shell execution path.
let emitter = ToolEmitter::shell(
exec_params.command.clone(),
exec_params.cwd.clone(),
is_user_shell_command,
);
let source = if is_user_shell_command {
ExecCommandSource::UserShell
} else {
ExecCommandSource::Agent
};
let emitter =
ToolEmitter::shell(exec_params.command.clone(), exec_params.cwd.clone(), source);
let event_ctx = ToolEventCtx::new(session.as_ref(), turn.as_ref(), &call_id, None);
emitter.begin(event_ctx).await;
@@ -324,8 +327,11 @@ impl ShellHandler {
#[cfg(test)]
mod tests {
use std::path::PathBuf;
use crate::is_safe_command::is_known_safe_command;
use crate::shell::BashShell;
use crate::shell::PowerShellConfig;
use crate::shell::Shell;
use crate::shell::ZshShell;
@@ -335,27 +341,19 @@ mod tests {
#[test]
fn commands_generated_by_shell_command_handler_can_be_matched_by_is_known_safe_command() {
let bash_shell = Shell::Bash(BashShell {
shell_path: "/bin/bash".to_string(),
bashrc_path: "/home/user/.bashrc".to_string(),
shell_path: PathBuf::from("/bin/bash"),
});
assert_safe(&bash_shell, "ls -la");
let zsh_shell = Shell::Zsh(ZshShell {
shell_path: "/bin/zsh".to_string(),
zshrc_path: "/home/user/.zshrc".to_string(),
shell_path: PathBuf::from("/bin/zsh"),
});
assert_safe(&zsh_shell, "ls -la");
#[cfg(target_os = "windows")]
{
use crate::shell::PowerShellConfig;
let powershell = Shell::PowerShell(PowerShellConfig {
exe: "pwsh.exe".to_string(),
bash_exe_fallback: None,
});
assert_safe(&powershell, "ls -Name");
}
let powershell = Shell::PowerShell(PowerShellConfig {
shell_path: PathBuf::from("pwsh.exe"),
});
assert_safe(&powershell, "ls -Name");
}
fn assert_safe(shell: &Shell, command: &str) {
codex-rs/core/src/tools/handlers/unified_exec.rs
+32 -12
View File
@@ -4,7 +4,9 @@ use crate::function_tool::FunctionCallError;
use crate::is_safe_command::is_known_safe_command;
use crate::protocol::EventMsg;
use crate::protocol::ExecCommandOutputDeltaEvent;
use crate::protocol::ExecCommandSource;
use crate::protocol::ExecOutputStream;
use crate::shell::get_shell_by_model_provided_path;
use crate::tools::context::ToolInvocation;
use crate::tools::context::ToolOutput;
use crate::tools::context::ToolPayload;
@@ -32,8 +34,8 @@ struct ExecCommandArgs {
shell: String,
#[serde(default = "default_login")]
login: bool,
#[serde(default)]
yield_time_ms: Option<u64>,
#[serde(default = "default_exec_yield_time_ms")]
yield_time_ms: u64,
#[serde(default)]
max_output_tokens: Option<usize>,
#[serde(default)]
@@ -47,12 +49,20 @@ struct WriteStdinArgs {
session_id: i32,
#[serde(default)]
chars: String,
#[serde(default)]
yield_time_ms: Option<u64>,
#[serde(default = "default_write_stdin_yield_time_ms")]
yield_time_ms: u64,
#[serde(default)]
max_output_tokens: Option<usize>,
}
fn default_exec_yield_time_ms() -> u64 {
10000
}
fn default_write_stdin_yield_time_ms() -> u64 {
250
}
fn default_shell() -> String {
"/bin/bash".to_string()
}
@@ -84,7 +94,8 @@ impl ToolHandler for UnifiedExecHandler {
let Ok(params) = serde_json::from_str::<ExecCommandArgs>(arguments) else {
return true;
};
!is_known_safe_command(&["bash".to_string(), "-lc".to_string(), params.cmd])
let command = get_command(&params);
!is_known_safe_command(&command)
}
async fn handle(&self, invocation: ToolInvocation) -> Result<ToolOutput, FunctionCallError> {
@@ -117,15 +128,15 @@ impl ToolHandler for UnifiedExecHandler {
"failed to parse exec_command arguments: {err:?}"
))
})?;
let command = get_command(&args);
let ExecCommandArgs {
cmd,
workdir,
shell,
login,
yield_time_ms,
max_output_tokens,
with_escalated_permissions,
justification,
..
} = args;
if with_escalated_permissions.unwrap_or(false)
@@ -152,15 +163,18 @@ impl ToolHandler for UnifiedExecHandler {
&context.call_id,
None,
);
let emitter = ToolEmitter::unified_exec(cmd.clone(), cwd.clone(), true);
let emitter = ToolEmitter::unified_exec(
&command,
cwd.clone(),
ExecCommandSource::UnifiedExecStartup,
None,
);
emitter.emit(event_ctx, ToolEventStage::Begin).await;
manager
.exec_command(
ExecCommandRequest {
command: &cmd,
shell: &shell,
login,
command,
yield_time_ms,
max_output_tokens,
workdir,
@@ -182,6 +196,7 @@ impl ToolHandler for UnifiedExecHandler {
})?;
manager
.write_stdin(WriteStdinRequest {
call_id: &call_id,
session_id: args.session_id,
input: &args.chars,
yield_time_ms: args.yield_time_ms,
@@ -221,6 +236,11 @@ impl ToolHandler for UnifiedExecHandler {
}
}
fn get_command(args: &ExecCommandArgs) -> Vec<String> {
let shell = get_shell_by_model_provided_path(&PathBuf::from(args.shell.clone()));
shell.derive_exec_args(&args.cmd, args.login)
}
fn format_response(response: &UnifiedExecResponse) -> String {
let mut sections = Vec::new();
codex-rs/core/src/tools/mod.rs
+3 -1
View File
@@ -9,6 +9,8 @@ pub mod runtimes;
pub mod sandboxing;
pub mod spec;
use crate::context_manager::MODEL_FORMAT_MAX_BYTES;
use crate::context_manager::MODEL_FORMAT_MAX_LINES;
use crate::context_manager::format_output_for_model_body;
use crate::exec::ExecToolCallOutput;
pub use router::ToolRouter;
@@ -75,5 +77,5 @@ pub fn format_exec_output_str(exec_output: &ExecToolCallOutput) -> String {
};
// Truncate for model consumption before serialization.
format_output_for_model_body(&body)
format_output_for_model_body(&body, MODEL_FORMAT_MAX_BYTES, MODEL_FORMAT_MAX_LINES)
}
codex-rs/core/src/tools/parallel.rs
+1 -1
View File
@@ -112,7 +112,7 @@ impl ToolCallRuntime {
fn abort_message(call: &ToolCall, secs: f32) -> String {
match call.tool_name.as_str() {
"shell" | "container.exec" | "local_shell" | "unified_exec" => {
"shell" | "container.exec" | "local_shell" | "shell_command" | "unified_exec" => {
format!("Wall time: {secs:.1} seconds\naborted by user")
}
_ => format!("aborted by user after {secs:.1}s"),
codex-rs/core/src/tools/router.rs
+2 -2
View File
@@ -54,7 +54,7 @@ impl ToolRouter {
.any(|config| config.spec.name() == tool_name)
}
pub fn build_tool_call(
pub async fn build_tool_call(
session: &Session,
item: ResponseItem,
) -> Result<Option<ToolCall>, FunctionCallError> {
@@ -65,7 +65,7 @@ impl ToolRouter {
call_id,
..
} => {
if let Some((server, tool)) = session.parse_mcp_tool_name(&name) {
if let Some((server, tool)) = session.parse_mcp_tool_name(&name).await {
Ok(Some(ToolCall {
tool_name: name,
call_id,
codex-rs/core/src/tools/spec.rs
+102 -15
View File
@@ -20,6 +20,11 @@ pub enum ConfigShellToolType {
Default,
Local,
UnifiedExec,
/// Do not include a shell tool by default. Useful when using Codex
/// with tools provided exclusively provided by MCP servers. Often used
/// with `--config base_instructions=CUSTOM_INSTRUCTIONS`
/// to customize agent behavior.
Disabled,
/// Takes a command as a single string to be run in the user's default shell.
ShellCommand,
}
@@ -48,7 +53,9 @@ impl ToolsConfig {
let include_web_search_request = features.enabled(Feature::WebSearchRequest);
let include_view_image_tool = features.enabled(Feature::ViewImageTool);
let shell_type = if features.enabled(Feature::UnifiedExec) {
let shell_type = if !features.enabled(Feature::ShellTool) {
ConfigShellToolType::Disabled
} else if features.enabled(Feature::UnifiedExec) {
ConfigShellToolType::UnifiedExec
} else if features.enabled(Feature::ShellCommandTool) {
ConfigShellToolType::ShellCommand
@@ -294,9 +301,26 @@ fn create_shell_tool() -> ToolSpec {
},
);
let description = if cfg!(windows) {
r#"Runs a Powershell command (Windows) and returns its output. Arguments to `shell` will be passed to CreateProcessW(). Most commands should be prefixed with ["powershell.exe", "-Command"].
Examples of valid command strings:
- ls -a (show hidden): ["powershell.exe", "-Command", "Get-ChildItem -Force"]
- recursive find by name: ["powershell.exe", "-Command", "Get-ChildItem -Recurse -Filter *.py"]
- recursive grep: ["powershell.exe", "-Command", "Get-ChildItem -Path C:\\myrepo -Recurse | Select-String -Pattern 'TODO' -CaseSensitive"]
- ps aux | grep python: ["powershell.exe", "-Command", "Get-Process | Where-Object { $_.ProcessName -like '*python*' }"]
- setting an env var: ["powershell.exe", "-Command", "$env:FOO='bar'; echo $env:FOO"]
- running an inline Python script: ["powershell.exe", "-Command", "@'\\nprint('Hello, world!')\\n'@ | python -"]"#
} else {
r#"Runs a shell command and returns its output.
- The arguments to `shell` will be passed to execvp(). Most terminal commands should be prefixed with ["bash", "-lc"].
- Always set the `workdir` param when using the shell function. Do not use `cd` unless absolutely necessary."#
}.to_string();
ToolSpec::Function(ResponsesApiTool {
name: "shell".to_string(),
description: "Runs a shell command and returns its output.".to_string(),
description,
strict: false,
parameters: JsonSchema::Object {
properties,
@@ -341,9 +365,25 @@ fn create_shell_command_tool() -> ToolSpec {
},
);
let description = if cfg!(windows) {
r#"Runs a Powershell command (Windows) and returns its output.
Examples of valid command strings:
- ls -a (show hidden): "Get-ChildItem -Force"
- recursive find by name: "Get-ChildItem -Recurse -Filter *.py"
- recursive grep: "Get-ChildItem -Path C:\\myrepo -Recurse | Select-String -Pattern 'TODO' -CaseSensitive"
- ps aux | grep python: "Get-Process | Where-Object { $_.ProcessName -like '*python*' }"
- setting an env var: "$env:FOO='bar'; echo $env:FOO"
- running an inline Python script: "@'\\nprint('Hello, world!')\\n'@ | python -"#
} else {
r#"Runs a shell command and returns its output.
- Always set the `workdir` param when using the shell_command function. Do not use `cd` unless absolutely necessary."#
}.to_string();
ToolSpec::Function(ResponsesApiTool {
name: "shell_command".to_string(),
description: "Runs a shell command string and returns its output.".to_string(),
description,
strict: false,
parameters: JsonSchema::Object {
properties,
@@ -973,16 +1013,21 @@ pub(crate) fn build_specs(
builder.register_handler("exec_command", unified_exec_handler.clone());
builder.register_handler("write_stdin", unified_exec_handler);
}
ConfigShellToolType::Disabled => {
// Do nothing.
}
ConfigShellToolType::ShellCommand => {
builder.push_spec(create_shell_command_tool());
}
}
// Always register shell aliases so older prompts remain compatible.
builder.register_handler("shell", shell_handler.clone());
builder.register_handler("container.exec", shell_handler.clone());
builder.register_handler("local_shell", shell_handler);
builder.register_handler("shell_command", shell_command_handler);
if config.shell_type != ConfigShellToolType::Disabled {
// Always register shell aliases so older prompts remain compatible.
builder.register_handler("shell", shell_handler.clone());
builder.register_handler("container.exec", shell_handler.clone());
builder.register_handler("local_shell", shell_handler);
builder.register_handler("shell_command", shell_command_handler);
}
builder.push_spec_with_parallel_support(create_list_mcp_resources_tool(), true);
builder.push_spec_with_parallel_support(create_list_mcp_resource_templates_tool(), true);
@@ -1118,6 +1163,7 @@ mod tests {
ConfigShellToolType::Default => Some("shell"),
ConfigShellToolType::Local => Some("local_shell"),
ConfigShellToolType::UnifiedExec => None,
ConfigShellToolType::Disabled => None,
ConfigShellToolType::ShellCommand => Some("shell_command"),
}
}
@@ -1246,7 +1292,11 @@ mod tests {
"gpt-5-codex",
&Features::with_defaults(),
&[
"shell",
if cfg!(windows) {
"shell_command"
} else {
"shell"
},
"list_mcp_resources",
"list_mcp_resource_templates",
"read_mcp_resource",
@@ -1263,7 +1313,11 @@ mod tests {
"gpt-5.1-codex",
&Features::with_defaults(),
&[
"shell",
if cfg!(windows) {
"shell_command"
} else {
"shell"
},
"list_mcp_resources",
"list_mcp_resource_templates",
"read_mcp_resource",
@@ -1338,7 +1392,11 @@ mod tests {
"gpt-5.1-codex-mini",
&Features::with_defaults(),
&[
"shell",
if cfg!(windows) {
"shell_command"
} else {
"shell"
},
"list_mcp_resources",
"list_mcp_resource_templates",
"read_mcp_resource",
@@ -1873,8 +1931,23 @@ mod tests {
};
assert_eq!(name, "shell");
let expected = "Runs a shell command and returns its output.";
assert_eq!(description, expected);
let expected = if cfg!(windows) {
r#"Runs a Powershell command (Windows) and returns its output. Arguments to `shell` will be passed to CreateProcessW(). Most commands should be prefixed with ["powershell.exe", "-Command"].
Examples of valid command strings:
- ls -a (show hidden): ["powershell.exe", "-Command", "Get-ChildItem -Force"]
- recursive find by name: ["powershell.exe", "-Command", "Get-ChildItem -Recurse -Filter *.py"]
- recursive grep: ["powershell.exe", "-Command", "Get-ChildItem -Path C:\\myrepo -Recurse | Select-String -Pattern 'TODO' -CaseSensitive"]
- ps aux | grep python: ["powershell.exe", "-Command", "Get-Process | Where-Object { $_.ProcessName -like '*python*' }"]
- setting an env var: ["powershell.exe", "-Command", "$env:FOO='bar'; echo $env:FOO"]
- running an inline Python script: ["powershell.exe", "-Command", "@'\\nprint('Hello, world!')\\n'@ | python -"]"#
} else {
r#"Runs a shell command and returns its output.
- The arguments to `shell` will be passed to execvp(). Most terminal commands should be prefixed with ["bash", "-lc"].
- Always set the `workdir` param when using the shell function. Do not use `cd` unless absolutely necessary."#
}.to_string();
assert_eq!(description, &expected);
}
#[test]
@@ -1888,8 +1961,22 @@ mod tests {
};
assert_eq!(name, "shell_command");
let expected = "Runs a shell command string and returns its output.";
assert_eq!(description, expected);
let expected = if cfg!(windows) {
r#"Runs a Powershell command (Windows) and returns its output.
Examples of valid command strings:
- ls -a (show hidden): "Get-ChildItem -Force"
- recursive find by name: "Get-ChildItem -Recurse -Filter *.py"
- recursive grep: "Get-ChildItem -Path C:\\myrepo -Recurse | Select-String -Pattern 'TODO' -CaseSensitive"
- ps aux | grep python: "Get-Process | Where-Object { $_.ProcessName -like '*python*' }"
- setting an env var: "$env:FOO='bar'; echo $env:FOO"
- running an inline Python script: "@'\\nprint('Hello, world!')\\n'@ | python -"#.to_string()
} else {
r#"Runs a shell command and returns its output.
- Always set the `workdir` param when using the shell_command function. Do not use `cd` unless absolutely necessary."#.to_string()
};
assert_eq!(description, &expected);
}
#[test]
codex-rs/core/src/truncate.rs
+416 -1
View File
@@ -1,8 +1,194 @@
//! Utilities for truncating large chunks of output while preserving a prefix
//! and suffix on UTF-8 boundaries.
//! and suffix on UTF-8 boundaries, and helpers for line/token‑based truncation
//! used across the core crate.
use codex_protocol::models::FunctionCallOutputContentItem;
use codex_utils_string::take_bytes_at_char_boundary;
use codex_utils_string::take_last_bytes_at_char_boundary;
use codex_utils_tokenizer::Tokenizer;
/// Model-formatting limits: clients get full streams; only content sent to the model is truncated.
pub const MODEL_FORMAT_MAX_BYTES: usize = 10 * 1024; // 10 KiB
pub const MODEL_FORMAT_MAX_LINES: usize = 256; // lines
/// Globally truncate function output items to fit within `MODEL_FORMAT_MAX_BYTES`
/// by preserving as many text/image items as possible and appending a summary
/// for any omitted text items.
pub(crate) fn globally_truncate_function_output_items(
items: &[FunctionCallOutputContentItem],
) -> Vec<FunctionCallOutputContentItem> {
let mut out: Vec<FunctionCallOutputContentItem> = Vec::with_capacity(items.len());
let mut remaining = MODEL_FORMAT_MAX_BYTES;
let mut omitted_text_items = 0usize;
for it in items {
match it {
FunctionCallOutputContentItem::InputText { text } => {
if remaining == 0 {
omitted_text_items += 1;
continue;
}
let len = text.len();
if len <= remaining {
out.push(FunctionCallOutputContentItem::InputText { text: text.clone() });
remaining -= len;
} else {
let slice = take_bytes_at_char_boundary(text, remaining);
if !slice.is_empty() {
out.push(FunctionCallOutputContentItem::InputText {
text: slice.to_string(),
});
}
remaining = 0;
}
}
// todo(aibrahim): handle input images; resize
FunctionCallOutputContentItem::InputImage { image_url } => {
out.push(FunctionCallOutputContentItem::InputImage {
image_url: image_url.clone(),
});
}
}
}
if omitted_text_items > 0 {
out.push(FunctionCallOutputContentItem::InputText {
text: format!("[omitted {omitted_text_items} text items ...]"),
});
}
out
}
/// Format a block of exec/tool output for model consumption, truncating by
/// lines and bytes while preserving head and tail segments.
pub(crate) fn format_output_for_model_body(
content: &str,
limit_bytes: usize,
limit_lines: usize,
) -> String {
// Head+tail truncation for the model: show the beginning and end with an elision.
// Clients still receive full streams; only this formatted summary is capped.
let total_lines = content.lines().count();
if content.len() <= limit_bytes && total_lines <= limit_lines {
return content.to_string();
}
let output = truncate_formatted_exec_output(content, total_lines, limit_bytes, limit_lines);
format!("Total output lines: {total_lines}\n\n{output}")
}
fn truncate_formatted_exec_output(
content: &str,
total_lines: usize,
limit_bytes: usize,
limit_lines: usize,
) -> String {
error_on_double_truncation(content);
let head_lines: usize = limit_lines / 2;
let tail_lines: usize = limit_lines - head_lines; // 128
let head_bytes: usize = limit_bytes / 2;
let segments: Vec<&str> = content.split_inclusive('\n').collect();
let head_take = head_lines.min(segments.len());
let tail_take = tail_lines.min(segments.len().saturating_sub(head_take));
let omitted = segments.len().saturating_sub(head_take + tail_take);
let head_slice_end: usize = segments
.iter()
.take(head_take)
.map(|segment| segment.len())
.sum();
let tail_slice_start: usize = if tail_take == 0 {
content.len()
} else {
content.len()
- segments
.iter()
.rev()
.take(tail_take)
.map(|segment| segment.len())
.sum::<usize>()
};
let head_slice = &content[..head_slice_end];
let tail_slice = &content[tail_slice_start..];
let truncated_by_bytes = content.len() > limit_bytes;
// this is a bit wrong. We are counting metadata lines and not just shell output lines.
let marker = if omitted > 0 {
Some(format!(
"\n[... omitted {omitted} of {total_lines} lines ...]\n\n"
))
} else if truncated_by_bytes {
Some(format!(
"\n[... output truncated to fit {limit_bytes} bytes ...]\n\n"
))
} else {
None
};
let marker_len = marker.as_ref().map_or(0, String::len);
let base_head_budget = head_bytes.min(limit_bytes);
let head_budget = base_head_budget.min(limit_bytes.saturating_sub(marker_len));
let head_part = take_bytes_at_char_boundary(head_slice, head_budget);
let mut result = String::with_capacity(limit_bytes.min(content.len()));
result.push_str(head_part);
if let Some(marker_text) = marker.as_ref() {
result.push_str(marker_text);
}
let remaining = limit_bytes.saturating_sub(result.len());
if remaining == 0 {
return result;
}
let tail_part = take_last_bytes_at_char_boundary(tail_slice, remaining);
result.push_str(tail_part);
result
}
fn error_on_double_truncation(content: &str) {
if content.contains("Total output lines:") && content.contains("omitted") {
tracing::error!(
"FunctionCallOutput content was already truncated before ContextManager::record_items; this would cause double truncation {content}"
);
}
}
/// Truncate an output string to a maximum number of “tokens”, where tokens are
/// approximated as individual `char`s. Preserves a prefix and suffix with an
/// elision marker describing how many tokens were omitted.
pub(crate) fn truncate_output_to_tokens(
output: &str,
max_tokens: usize,
) -> (String, Option<usize>) {
if max_tokens == 0 {
let total_tokens = output.chars().count();
let message = format!("…{total_tokens} tokens truncated…");
return (message, Some(total_tokens));
}
let tokens: Vec<char> = output.chars().collect();
let total_tokens = tokens.len();
if total_tokens <= max_tokens {
return (output.to_string(), None);
}
let half = max_tokens / 2;
if half == 0 {
let truncated = total_tokens.saturating_sub(max_tokens);
let message = format!("…{truncated} tokens truncated…");
return (message, Some(total_tokens));
}
let truncated = total_tokens.saturating_sub(half * 2);
let mut truncated_output = String::new();
truncated_output.extend(&tokens[..half]);
truncated_output.push_str(&format!("…{truncated} tokens truncated…"));
truncated_output.extend(&tokens[total_tokens - half..]);
(truncated_output, Some(total_tokens))
}
/// Truncate the middle of a UTF-8 string to at most `max_bytes` bytes,
/// preserving the beginning and the end. Returns the possibly truncated
/// string and `Some(original_token_count)` (counted with the local tokenizer;
@@ -136,8 +322,33 @@ pub(crate) fn truncate_middle(s: &str, max_bytes: usize) -> (String, Option<u64>
#[cfg(test)]
mod tests {
use super::MODEL_FORMAT_MAX_BYTES;
use super::MODEL_FORMAT_MAX_LINES;
use super::format_output_for_model_body;
use super::globally_truncate_function_output_items;
use super::truncate_middle;
use super::truncate_output_to_tokens;
use codex_protocol::models::FunctionCallOutputContentItem;
use codex_utils_tokenizer::Tokenizer;
use pretty_assertions::assert_eq;
use regex_lite::Regex;
fn truncated_message_pattern(line: &str, total_lines: usize) -> String {
let head_lines = MODEL_FORMAT_MAX_LINES / 2;
let tail_lines = MODEL_FORMAT_MAX_LINES - head_lines;
let head_take = head_lines.min(total_lines);
let tail_take = tail_lines.min(total_lines.saturating_sub(head_take));
let omitted = total_lines.saturating_sub(head_take + tail_take);
let escaped_line = regex_lite::escape(line);
if omitted == 0 {
return format!(
r"(?s)^Total output lines: {total_lines}\n\n(?P<body>{escaped_line}.*\n\[\.{{3}} output truncated to fit {MODEL_FORMAT_MAX_BYTES} bytes \.{{3}}]\n\n.*)$",
);
}
format!(
r"(?s)^Total output lines: {total_lines}\n\n(?P<body>{escaped_line}.*\n\[\.{{3}} omitted {omitted} of {total_lines} lines \.{{3}}]\n\n.*)$",
)
}
#[test]
fn truncate_middle_no_newlines_fallback() {
@@ -197,4 +408,208 @@ mod tests {
assert!(out.ends_with("017\n018\n019\n020\n"));
assert_eq!(total, Some(tok.count(&s) as u64));
}
#[test]
fn truncate_output_to_tokens_returns_original_when_under_limit() {
let s = "short output";
let (truncated, original) = truncate_output_to_tokens(s, 100);
assert_eq!(truncated, s);
assert_eq!(original, None);
}
#[test]
fn truncate_output_to_tokens_reports_truncation_at_zero_limit() {
let s = "abcdef";
let (truncated, original) = truncate_output_to_tokens(s, 0);
assert!(truncated.contains("tokens truncated"));
assert_eq!(original, Some(s.chars().count()));
}
#[test]
fn truncate_output_to_tokens_preserves_prefix_and_suffix() {
let s = "abcdefghijklmnopqrstuvwxyz";
let max_tokens = 10;
let (truncated, original) = truncate_output_to_tokens(s, max_tokens);
assert!(truncated.starts_with("abcde"));
assert!(truncated.ends_with("vwxyz"));
assert_eq!(original, Some(s.chars().count()));
}
#[test]
fn format_exec_output_truncates_large_error() {
let line = "very long execution error line that should trigger truncation\n";
let large_error = line.repeat(2_500); // way beyond both byte and line limits
let truncated = format_output_for_model_body(
&large_error,
MODEL_FORMAT_MAX_BYTES,
MODEL_FORMAT_MAX_LINES,
);
let total_lines = large_error.lines().count();
let pattern = truncated_message_pattern(line, total_lines);
let regex = Regex::new(&pattern).unwrap_or_else(|err| {
panic!("failed to compile regex {pattern}: {err}");
});
let captures = regex
.captures(&truncated)
.unwrap_or_else(|| panic!("message failed to match pattern {pattern}: {truncated}"));
let body = captures
.name("body")
.expect("missing body capture")
.as_str();
assert!(
body.len() <= MODEL_FORMAT_MAX_BYTES,
"body exceeds byte limit: {} bytes",
body.len()
);
assert_ne!(truncated, large_error);
}
#[test]
fn format_exec_output_marks_byte_truncation_without_omitted_lines() {
let long_line = "a".repeat(MODEL_FORMAT_MAX_BYTES + 50);
let truncated = format_output_for_model_body(
&long_line,
MODEL_FORMAT_MAX_BYTES,
MODEL_FORMAT_MAX_LINES,
);
assert_ne!(truncated, long_line);
let marker_line =
format!("[... output truncated to fit {MODEL_FORMAT_MAX_BYTES} bytes ...]");
assert!(
truncated.contains(&marker_line),
"missing byte truncation marker: {truncated}"
);
assert!(
!truncated.contains("omitted"),
"line omission marker should not appear when no lines were dropped: {truncated}"
);
}
#[test]
fn format_exec_output_returns_original_when_within_limits() {
let content = "example output\n".repeat(10);
assert_eq!(
format_output_for_model_body(&content, MODEL_FORMAT_MAX_BYTES, MODEL_FORMAT_MAX_LINES),
content
);
}
#[test]
fn format_exec_output_reports_omitted_lines_and_keeps_head_and_tail() {
let total_lines = MODEL_FORMAT_MAX_LINES + 100;
let content: String = (0..total_lines)
.map(|idx| format!("line-{idx}\n"))
.collect();
let truncated =
format_output_for_model_body(&content, MODEL_FORMAT_MAX_BYTES, MODEL_FORMAT_MAX_LINES);
let omitted = total_lines - MODEL_FORMAT_MAX_LINES;
let expected_marker = format!("[... omitted {omitted} of {total_lines} lines ...]");
assert!(
truncated.contains(&expected_marker),
"missing omitted marker: {truncated}"
);
assert!(
truncated.contains("line-0\n"),
"expected head line to remain: {truncated}"
);
let last_line = format!("line-{}\n", total_lines - 1);
assert!(
truncated.contains(&last_line),
"expected tail line to remain: {truncated}"
);
}
#[test]
fn format_exec_output_prefers_line_marker_when_both_limits_exceeded() {
let total_lines = MODEL_FORMAT_MAX_LINES + 42;
let long_line = "x".repeat(256);
let content: String = (0..total_lines)
.map(|idx| format!("line-{idx}-{long_line}\n"))
.collect();
let truncated =
format_output_for_model_body(&content, MODEL_FORMAT_MAX_BYTES, MODEL_FORMAT_MAX_LINES);
assert!(
truncated.contains("[... omitted 42 of 298 lines ...]"),
"expected omitted marker when line count exceeds limit: {truncated}"
);
assert!(
!truncated.contains("output truncated to fit"),
"line omission marker should take precedence over byte marker: {truncated}"
);
}
#[test]
fn truncates_across_multiple_under_limit_texts_and_reports_omitted() {
// Arrange: several text items, none exceeding per-item limit, but total exceeds budget.
let budget = MODEL_FORMAT_MAX_BYTES;
let t1_len = (budget / 2).saturating_sub(10);
let t2_len = (budget / 2).saturating_sub(10);
let remaining_after_t1_t2 = budget.saturating_sub(t1_len + t2_len);
let t3_len = 50; // gets truncated to remaining_after_t1_t2
let t4_len = 5; // omitted
let t5_len = 7; // omitted
let t1 = "a".repeat(t1_len);
let t2 = "b".repeat(t2_len);
let t3 = "c".repeat(t3_len);
let t4 = "d".repeat(t4_len);
let t5 = "e".repeat(t5_len);
let items = vec![
FunctionCallOutputContentItem::InputText { text: t1 },
FunctionCallOutputContentItem::InputText { text: t2 },
FunctionCallOutputContentItem::InputImage {
image_url: "img:mid".to_string(),
},
FunctionCallOutputContentItem::InputText { text: t3 },
FunctionCallOutputContentItem::InputText { text: t4 },
FunctionCallOutputContentItem::InputText { text: t5 },
];
let output = globally_truncate_function_output_items(&items);
// Expect: t1 (full), t2 (full), image, t3 (truncated), summary mentioning 2 omitted.
assert_eq!(output.len(), 5);
let first_text = match &output[0] {
FunctionCallOutputContentItem::InputText { text } => text,
other => panic!("unexpected first item: {other:?}"),
};
assert_eq!(first_text.len(), t1_len);
let second_text = match &output[1] {
FunctionCallOutputContentItem::InputText { text } => text,
other => panic!("unexpected second item: {other:?}"),
};
assert_eq!(second_text.len(), t2_len);
assert_eq!(
output[2],
FunctionCallOutputContentItem::InputImage {
image_url: "img:mid".to_string()
}
);
let fourth_text = match &output[3] {
FunctionCallOutputContentItem::InputText { text } => text,
other => panic!("unexpected fourth item: {other:?}"),
};
assert_eq!(fourth_text.len(), remaining_after_t1_t2);
let summary_text = match &output[4] {
FunctionCallOutputContentItem::InputText { text } => text,
other => panic!("unexpected summary item: {other:?}"),
};
assert!(summary_text.contains("omitted 2 text items"));
}
}
codex-rs/core/src/unified_exec/mod.rs
+30 -71
View File
@@ -41,7 +41,6 @@ mod session_manager;
pub(crate) use errors::UnifiedExecError;
pub(crate) use session::UnifiedExecSession;
pub(crate) const DEFAULT_YIELD_TIME_MS: u64 = 10_000;
pub(crate) const MIN_YIELD_TIME_MS: u64 = 250;
pub(crate) const MAX_YIELD_TIME_MS: u64 = 30_000;
pub(crate) const DEFAULT_MAX_OUTPUT_TOKENS: usize = 10_000;
@@ -64,11 +63,9 @@ impl UnifiedExecContext {
}
#[derive(Debug)]
pub(crate) struct ExecCommandRequest<'a> {
pub command: &'a str,
pub shell: &'a str,
pub login: bool,
pub yield_time_ms: Option<u64>,
pub(crate) struct ExecCommandRequest {
pub command: Vec<String>,
pub yield_time_ms: u64,
pub max_output_tokens: Option<usize>,
pub workdir: Option<PathBuf>,
pub with_escalated_permissions: Option<bool>,
@@ -77,9 +74,10 @@ pub(crate) struct ExecCommandRequest<'a> {
#[derive(Debug)]
pub(crate) struct WriteStdinRequest<'a> {
pub call_id: &'a str,
pub session_id: i32,
pub input: &'a str,
pub yield_time_ms: Option<u64>,
pub yield_time_ms: u64,
pub max_output_tokens: Option<usize>,
}
@@ -92,6 +90,7 @@ pub(crate) struct UnifiedExecResponse {
pub session_id: Option<i32>,
pub exit_code: Option<i32>,
pub original_token_count: Option<usize>,
pub session_command: Option<Vec<String>>,
}
#[derive(Default)]
@@ -105,16 +104,13 @@ struct SessionEntry {
session_ref: Arc<Session>,
turn_ref: Arc<TurnContext>,
call_id: String,
command: String,
command: Vec<String>,
cwd: PathBuf,
started_at: tokio::time::Instant,
}
pub(crate) fn clamp_yield_time(yield_time_ms: Option<u64>) -> u64 {
match yield_time_ms {
Some(value) => value.clamp(MIN_YIELD_TIME_MS, MAX_YIELD_TIME_MS),
None => DEFAULT_YIELD_TIME_MS,
}
pub(crate) fn clamp_yield_time(yield_time_ms: u64) -> u64 {
yield_time_ms.clamp(MIN_YIELD_TIME_MS, MAX_YIELD_TIME_MS)
}
pub(crate) fn resolve_max_tokens(max_tokens: Option<usize>) -> usize {
@@ -128,37 +124,6 @@ pub(crate) fn generate_chunk_id() -> String {
.collect()
}
pub(crate) fn truncate_output_to_tokens(
output: &str,
max_tokens: usize,
) -> (String, Option<usize>) {
if max_tokens == 0 {
let total_tokens = output.chars().count();
let message = format!("…{total_tokens} tokens truncated…");
return (message, Some(total_tokens));
}
let tokens: Vec<char> = output.chars().collect();
let total_tokens = tokens.len();
if total_tokens <= max_tokens {
return (output.to_string(), None);
}
let half = max_tokens / 2;
if half == 0 {
let truncated = total_tokens.saturating_sub(max_tokens);
let message = format!("…{truncated} tokens truncated…");
return (message, Some(total_tokens));
}
let truncated = total_tokens.saturating_sub(half * 2);
let mut truncated_output = String::new();
truncated_output.extend(&tokens[..half]);
truncated_output.push_str(&format!("…{truncated} tokens truncated…"));
truncated_output.extend(&tokens[total_tokens - half..]);
(truncated_output, Some(total_tokens))
}
#[cfg(test)]
#[cfg(unix)]
mod tests {
@@ -187,7 +152,7 @@ mod tests {
session: &Arc<Session>,
turn: &Arc<TurnContext>,
cmd: &str,
yield_time_ms: Option<u64>,
yield_time_ms: u64,
) -> Result<UnifiedExecResponse, UnifiedExecError> {
let context =
UnifiedExecContext::new(Arc::clone(session), Arc::clone(turn), "call".to_string());
@@ -197,9 +162,7 @@ mod tests {
.unified_exec_manager
.exec_command(
ExecCommandRequest {
command: cmd,
shell: "/bin/bash",
login: true,
command: vec!["bash".to_string(), "-lc".to_string(), cmd.to_string()],
yield_time_ms,
max_output_tokens: None,
workdir: None,
@@ -215,12 +178,13 @@ mod tests {
session: &Arc<Session>,
session_id: i32,
input: &str,
yield_time_ms: Option<u64>,
yield_time_ms: u64,
) -> Result<UnifiedExecResponse, UnifiedExecError> {
session
.services
.unified_exec_manager
.write_stdin(WriteStdinRequest {
call_id: "write-stdin",
session_id,
input,
yield_time_ms,
@@ -253,14 +217,14 @@ mod tests {
let (session, turn) = test_session_and_turn();
let open_shell = exec_command(&session, &turn, "bash -i", Some(2_500)).await?;
let open_shell = exec_command(&session, &turn, "bash -i", 2_500).await?;
let session_id = open_shell.session_id.expect("expected session_id");
write_stdin(
&session,
session_id,
"export CODEX_INTERACTIVE_SHELL_VAR=codex\n",
Some(2_500),
2_500,
)
.await?;
@@ -268,7 +232,7 @@ mod tests {
&session,
session_id,
"echo $CODEX_INTERACTIVE_SHELL_VAR\n",
Some(2_500),
2_500,
)
.await?;
assert!(
@@ -285,24 +249,19 @@ mod tests {
let (session, turn) = test_session_and_turn();
let shell_a = exec_command(&session, &turn, "bash -i", Some(2_500)).await?;
let shell_a = exec_command(&session, &turn, "bash -i", 2_500).await?;
let session_a = shell_a.session_id.expect("expected session id");
write_stdin(
&session,
session_a,
"export CODEX_INTERACTIVE_SHELL_VAR=codex\n",
Some(2_500),
2_500,
)
.await?;
let out_2 = exec_command(
&session,
&turn,
"echo $CODEX_INTERACTIVE_SHELL_VAR",
Some(2_500),
)
.await?;
let out_2 =
exec_command(&session, &turn, "echo $CODEX_INTERACTIVE_SHELL_VAR", 2_500).await?;
assert!(
out_2.session_id.is_none(),
"short command should not retain a session"
@@ -316,7 +275,7 @@ mod tests {
&session,
session_a,
"echo $CODEX_INTERACTIVE_SHELL_VAR\n",
Some(2_500),
2_500,
)
.await?;
assert!(
@@ -333,14 +292,14 @@ mod tests {
let (session, turn) = test_session_and_turn();
let open_shell = exec_command(&session, &turn, "bash -i", Some(2_500)).await?;
let open_shell = exec_command(&session, &turn, "bash -i", 2_500).await?;
let session_id = open_shell.session_id.expect("expected session id");
write_stdin(
&session,
session_id,
"export CODEX_INTERACTIVE_SHELL_VAR=codex\n",
Some(2_500),
2_500,
)
.await?;
@@ -348,7 +307,7 @@ mod tests {
&session,
session_id,
"sleep 5 && echo $CODEX_INTERACTIVE_SHELL_VAR\n",
Some(10),
10,
)
.await?;
assert!(
@@ -358,7 +317,7 @@ mod tests {
tokio::time::sleep(Duration::from_secs(7)).await;
let out_3 = write_stdin(&session, session_id, "", Some(100)).await?;
let out_3 = write_stdin(&session, session_id, "", 100).await?;
assert!(
out_3.output.contains("codex"),
@@ -373,7 +332,7 @@ mod tests {
async fn requests_with_large_timeout_are_capped() -> anyhow::Result<()> {
let (session, turn) = test_session_and_turn();
let result = exec_command(&session, &turn, "echo codex", Some(120_000)).await?;
let result = exec_command(&session, &turn, "echo codex", 120_000).await?;
assert!(result.session_id.is_none());
assert!(result.output.contains("codex"));
@@ -385,7 +344,7 @@ mod tests {
#[ignore] // Ignored while we have a better way to test this.
async fn completed_commands_do_not_persist_sessions() -> anyhow::Result<()> {
let (session, turn) = test_session_and_turn();
let result = exec_command(&session, &turn, "echo codex", Some(2_500)).await?;
let result = exec_command(&session, &turn, "echo codex", 2_500).await?;
assert!(
result.session_id.is_none(),
@@ -412,14 +371,14 @@ mod tests {
let (session, turn) = test_session_and_turn();
let open_shell = exec_command(&session, &turn, "bash -i", Some(2_500)).await?;
let open_shell = exec_command(&session, &turn, "bash -i", 2_500).await?;
let session_id = open_shell.session_id.expect("expected session id");
write_stdin(&session, session_id, "exit\n", Some(2_500)).await?;
write_stdin(&session, session_id, "exit\n", 2_500).await?;
tokio::time::sleep(Duration::from_millis(200)).await;
let err = write_stdin(&session, session_id, "", Some(100))
let err = write_stdin(&session, session_id, "", 100)
.await
.expect_err("expected unknown session error");
codex-rs/core/src/unified_exec/session_manager.rs
+146 -37
View File
@@ -6,12 +6,18 @@ use tokio::sync::mpsc;
use tokio::time::Duration;
use tokio::time::Instant;
use crate::codex::Session;
use crate::codex::TurnContext;
use crate::exec::ExecToolCallOutput;
use crate::exec::StreamOutput;
use crate::exec_env::create_env;
use crate::protocol::BackgroundEventEvent;
use crate::protocol::EventMsg;
use crate::protocol::ExecCommandSource;
use crate::sandboxing::ExecEnv;
use crate::tools::events::ToolEmitter;
use crate::tools::events::ToolEventCtx;
use crate::tools::events::ToolEventFailure;
use crate::tools::events::ToolEventStage;
use crate::tools::orchestrator::ToolOrchestrator;
use crate::tools::runtimes::unified_exec::UnifiedExecRequest as UnifiedExecToolRequest;
@@ -19,7 +25,6 @@ use crate::tools::runtimes::unified_exec::UnifiedExecRuntime;
use crate::tools::sandboxing::ToolCtx;
use super::ExecCommandRequest;
use super::MIN_YIELD_TIME_MS;
use super::SessionEntry;
use super::UnifiedExecContext;
use super::UnifiedExecError;
@@ -31,28 +36,22 @@ use super::generate_chunk_id;
use super::resolve_max_tokens;
use super::session::OutputBuffer;
use super::session::UnifiedExecSession;
use super::truncate_output_to_tokens;
use crate::truncate::truncate_output_to_tokens;
impl UnifiedExecSessionManager {
pub(crate) async fn exec_command(
&self,
request: ExecCommandRequest<'_>,
request: ExecCommandRequest,
context: &UnifiedExecContext,
) -> Result<UnifiedExecResponse, UnifiedExecError> {
let cwd = request
.workdir
.clone()
.unwrap_or_else(|| context.turn.cwd.clone());
let shell_flag = if request.login { "-lc" } else { "-c" };
let command = vec![
request.shell.to_string(),
shell_flag.to_string(),
request.command.to_string(),
];
let session = self
.open_session_with_sandbox(
command,
&request.command,
cwd.clone(),
request.with_escalated_permissions,
request.justification,
@@ -61,8 +60,7 @@ impl UnifiedExecSessionManager {
.await?;
let max_tokens = resolve_max_tokens(request.max_output_tokens);
let yield_time_ms =
clamp_yield_time(Some(request.yield_time_ms.unwrap_or(MIN_YIELD_TIME_MS)));
let yield_time_ms = clamp_yield_time(request.yield_time_ms);
let start = Instant::now();
let (output_buffer, output_notify) = session.output_handles();
@@ -74,15 +72,18 @@ impl UnifiedExecSessionManager {
let text = String::from_utf8_lossy(&collected).to_string();
let (output, original_token_count) = truncate_output_to_tokens(&text, max_tokens);
let chunk_id = generate_chunk_id();
let exit_code = session.exit_code();
let session_id = if session.has_exited() {
None
} else {
Some(
self.store_session(session, context, request.command, cwd.clone(), start)
.await,
)
};
let has_exited = session.has_exited();
let stored_id = self
.store_session(session, context, &request.command, cwd.clone(), start)
.await;
let exit_code = self
.sessions
.lock()
.await
.get(&stored_id)
.map(|entry| entry.session.exit_code());
// Only include a session_id in the response if the process is still alive.
let session_id = if has_exited { None } else { Some(stored_id) };
let response = UnifiedExecResponse {
event_call_id: context.call_id.clone(),
@@ -90,16 +91,21 @@ impl UnifiedExecSessionManager {
wall_time,
output,
session_id,
exit_code,
exit_code: exit_code.flatten(),
original_token_count,
session_command: Some(request.command.clone()),
};
if response.session_id.is_some() {
Self::emit_waiting_status(&context.session, &context.turn, &request.command).await;
}
// If the command completed during this call, emit an ExecCommandEnd via the emitter.
if response.session_id.is_none() {
let exit = response.exit_code.unwrap_or(-1);
Self::emit_exec_end_from_context(
context,
request.command.to_string(),
&request.command,
cwd,
response.output.clone(),
exit,
@@ -117,11 +123,46 @@ impl UnifiedExecSessionManager {
) -> Result<UnifiedExecResponse, UnifiedExecError> {
let session_id = request.session_id;
let (writer_tx, output_buffer, output_notify) =
self.prepare_session_handles(session_id).await?;
let (
writer_tx,
output_buffer,
output_notify,
session_ref,
turn_ref,
session_command,
session_cwd,
) = self.prepare_session_handles(session_id).await?;
let interaction_emitter = ToolEmitter::unified_exec(
&session_command,
session_cwd.clone(),
ExecCommandSource::UnifiedExecInteraction,
(!request.input.is_empty()).then(|| request.input.to_string()),
);
let make_event_ctx = || {
ToolEventCtx::new(
session_ref.as_ref(),
turn_ref.as_ref(),
request.call_id,
None,
)
};
interaction_emitter
.emit(make_event_ctx(), ToolEventStage::Begin)
.await;
if !request.input.is_empty() {
Self::send_input(&writer_tx, request.input.as_bytes()).await?;
if let Err(err) = Self::send_input(&writer_tx, request.input.as_bytes()).await {
interaction_emitter
.emit(
make_event_ctx(),
ToolEventStage::Failure(ToolEventFailure::Message(format!(
"write_stdin failed: {err:?}"
))),
)
.await;
return Err(err);
}
tokio::time::sleep(Duration::from_millis(100)).await;
}
@@ -159,8 +200,28 @@ impl UnifiedExecSessionManager {
session_id,
exit_code,
original_token_count,
session_command: Some(session_command.clone()),
};
let interaction_output = ExecToolCallOutput {
exit_code: response.exit_code.unwrap_or(0),
stdout: StreamOutput::new(response.output.clone()),
stderr: StreamOutput::new(String::new()),
aggregated_output: StreamOutput::new(response.output.clone()),
duration: response.wall_time,
timed_out: false,
};
interaction_emitter
.emit(
make_event_ctx(),
ToolEventStage::Success(interaction_output),
)
.await;
if response.session_id.is_some() {
Self::emit_waiting_status(&session_ref, &turn_ref, &session_command).await;
}
if let (Some(exit), Some(entry)) = (response.exit_code, completion_entry) {
let total_duration = Instant::now().saturating_duration_since(entry.started_at);
Self::emit_exec_end_from_entry(entry, response.output.clone(), exit, total_duration)
@@ -197,17 +258,44 @@ impl UnifiedExecSessionManager {
async fn prepare_session_handles(
&self,
session_id: i32,
) -> Result<(mpsc::Sender<Vec<u8>>, OutputBuffer, Arc<Notify>), UnifiedExecError> {
) -> Result<
(
mpsc::Sender<Vec<u8>>,
OutputBuffer,
Arc<Notify>,
Arc<Session>,
Arc<TurnContext>,
Vec<String>,
PathBuf,
),
UnifiedExecError,
> {
let sessions = self.sessions.lock().await;
let (output_buffer, output_notify, writer_tx) =
let (output_buffer, output_notify, writer_tx, session, turn, command, cwd) =
if let Some(entry) = sessions.get(&session_id) {
let (buffer, notify) = entry.session.output_handles();
(buffer, notify, entry.session.writer_sender())
(
buffer,
notify,
entry.session.writer_sender(),
Arc::clone(&entry.session_ref),
Arc::clone(&entry.turn_ref),
entry.command.clone(),
entry.cwd.clone(),
)
} else {
return Err(UnifiedExecError::UnknownSessionId { session_id });
};
Ok((writer_tx, output_buffer, output_notify))
Ok((
writer_tx,
output_buffer,
output_notify,
session,
turn,
command,
cwd,
))
}
async fn send_input(
@@ -224,7 +312,7 @@ impl UnifiedExecSessionManager {
&self,
session: UnifiedExecSession,
context: &UnifiedExecContext,
command: &str,
command: &[String],
cwd: PathBuf,
started_at: Instant,
) -> i32 {
@@ -236,7 +324,7 @@ impl UnifiedExecSessionManager {
session_ref: Arc::clone(&context.session),
turn_ref: Arc::clone(&context.turn),
call_id: context.call_id.clone(),
command: command.to_string(),
command: command.to_vec(),
cwd,
started_at,
};
@@ -264,7 +352,12 @@ impl UnifiedExecSessionManager {
&entry.call_id,
None,
);
let emitter = ToolEmitter::unified_exec(entry.command, entry.cwd, true);
let emitter = ToolEmitter::unified_exec(
&entry.command,
entry.cwd,
ExecCommandSource::UnifiedExecStartup,
None,
);
emitter
.emit(event_ctx, ToolEventStage::Success(output))
.await;
@@ -272,7 +365,7 @@ impl UnifiedExecSessionManager {
async fn emit_exec_end_from_context(
context: &UnifiedExecContext,
command: String,
command: &[String],
cwd: PathBuf,
aggregated_output: String,
exit_code: i32,
@@ -292,12 +385,28 @@ impl UnifiedExecSessionManager {
&context.call_id,
None,
);
let emitter = ToolEmitter::unified_exec(command, cwd, true);
let emitter =
ToolEmitter::unified_exec(command, cwd, ExecCommandSource::UnifiedExecStartup, None);
emitter
.emit(event_ctx, ToolEventStage::Success(output))
.await;
}
async fn emit_waiting_status(
session: &Arc<Session>,
turn: &Arc<TurnContext>,
command: &[String],
) {
let command_display = command.join(" ");
let message = format!("Waiting for `{command_display}`");
session
.send_event(
turn.as_ref(),
EventMsg::BackgroundEvent(BackgroundEventEvent { message }),
)
.await;
}
pub(crate) async fn open_session_with_exec_env(
&self,
env: &ExecEnv,
@@ -321,7 +430,7 @@ impl UnifiedExecSessionManager {
pub(super) async fn open_session_with_sandbox(
&self,
command: Vec<String>,
command: &[String],
cwd: PathBuf,
with_escalated_permissions: Option<bool>,
justification: Option<String>,
@@ -330,7 +439,7 @@ impl UnifiedExecSessionManager {
let mut orchestrator = ToolOrchestrator::new();
let mut runtime = UnifiedExecRuntime::new(self);
let req = UnifiedExecToolRequest::new(
command,
command.to_vec(),
cwd,
create_env(&context.turn.shell_environment_policy),
with_escalated_permissions,
codex-rs/core/templates/compact/summary_prefix.md
+1
View File
@@ -0,0 +1 @@
Another language model started to solve this problem and produced a summary of its thinking process. You also have access to the state of the tools that were used by that language model. Use this to build on the work that has already been done and avoid duplicating work. Here is the summary produced by the other language model, use the information in this summary to assist with your own analysis:
codex-rs/core/tests/chat_completions_payload.rs
+9 -53
View File
@@ -10,11 +10,11 @@ use codex_core::ModelProviderInfo;
use codex_core::Prompt;
use codex_core::ResponseItem;
use codex_core::WireApi;
use codex_core::spawn::CODEX_SANDBOX_NETWORK_DISABLED_ENV_VAR;
use codex_otel::otel_event_manager::OtelEventManager;
use codex_protocol::ConversationId;
use codex_protocol::models::ReasoningItemContent;
use core_test_support::load_default_config_for_test;
use core_test_support::skip_if_no_network;
use futures::StreamExt;
use serde_json::Value;
use tempfile::TempDir;
@@ -24,10 +24,6 @@ use wiremock::ResponseTemplate;
use wiremock::matchers::method;
use wiremock::matchers::path;
fn network_disabled() -> bool {
std::env::var(CODEX_SANDBOX_NETWORK_DISABLED_ENV_VAR).is_ok()
}
async fn run_request(input: Vec<ResponseItem>) -> Value {
let server = MockServer::start().await;
@@ -191,12 +187,7 @@ fn first_assistant(messages: &[Value]) -> &Value {
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn omits_reasoning_when_none_present() {
if network_disabled() {
println!(
"Skipping test because it cannot execute when network is disabled in a Codex sandbox."
);
return;
}
skip_if_no_network!();
let body = run_request(vec![user_message("u1"), assistant_message("a1")]).await;
let messages = messages_from(&body);
@@ -208,12 +199,7 @@ async fn omits_reasoning_when_none_present() {
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn attaches_reasoning_to_previous_assistant() {
if network_disabled() {
println!(
"Skipping test because it cannot execute when network is disabled in a Codex sandbox."
);
return;
}
skip_if_no_network!();
let body = run_request(vec![
user_message("u1"),
@@ -230,12 +216,7 @@ async fn attaches_reasoning_to_previous_assistant() {
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn attaches_reasoning_to_function_call_anchor() {
if network_disabled() {
println!(
"Skipping test because it cannot execute when network is disabled in a Codex sandbox."
);
return;
}
skip_if_no_network!();
let body = run_request(vec![
user_message("u1"),
@@ -257,12 +238,7 @@ async fn attaches_reasoning_to_function_call_anchor() {
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn attaches_reasoning_to_local_shell_call() {
if network_disabled() {
println!(
"Skipping test because it cannot execute when network is disabled in a Codex sandbox."
);
return;
}
skip_if_no_network!();
let body = run_request(vec![
user_message("u1"),
@@ -282,12 +258,7 @@ async fn attaches_reasoning_to_local_shell_call() {
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn drops_reasoning_when_last_role_is_user() {
if network_disabled() {
println!(
"Skipping test because it cannot execute when network is disabled in a Codex sandbox."
);
return;
}
skip_if_no_network!();
let body = run_request(vec![
assistant_message("aPrev"),
@@ -301,12 +272,7 @@ async fn drops_reasoning_when_last_role_is_user() {
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn ignores_reasoning_before_last_user() {
if network_disabled() {
println!(
"Skipping test because it cannot execute when network is disabled in a Codex sandbox."
);
return;
}
skip_if_no_network!();
let body = run_request(vec![
user_message("u1"),
@@ -321,12 +287,7 @@ async fn ignores_reasoning_before_last_user() {
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn skips_empty_reasoning_segments() {
if network_disabled() {
println!(
"Skipping test because it cannot execute when network is disabled in a Codex sandbox."
);
return;
}
skip_if_no_network!();
let body = run_request(vec![
user_message("u1"),
@@ -342,12 +303,7 @@ async fn skips_empty_reasoning_segments() {
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn suppresses_duplicate_assistant_messages() {
if network_disabled() {
println!(
"Skipping test because it cannot execute when network is disabled in a Codex sandbox."
);
return;
}
skip_if_no_network!();
let body = run_request(vec![assistant_message("dup"), assistant_message("dup")]).await;
let messages = messages_from(&body);
codex-rs/core/tests/chat_completions_sse.rs
+44 -58
View File
@@ -10,11 +10,11 @@ use codex_core::Prompt;
use codex_core::ResponseEvent;
use codex_core::ResponseItem;
use codex_core::WireApi;
use codex_core::spawn::CODEX_SANDBOX_NETWORK_DISABLED_ENV_VAR;
use codex_otel::otel_event_manager::OtelEventManager;
use codex_protocol::ConversationId;
use codex_protocol::models::ReasoningItemContent;
use core_test_support::load_default_config_for_test;
use core_test_support::skip_if_no_network;
use futures::StreamExt;
use tempfile::TempDir;
use wiremock::Mock;
@@ -23,10 +23,6 @@ use wiremock::ResponseTemplate;
use wiremock::matchers::method;
use wiremock::matchers::path;
fn network_disabled() -> bool {
std::env::var(CODEX_SANDBOX_NETWORK_DISABLED_ENV_VAR).is_ok()
}
async fn run_stream(sse_body: &str) -> Vec<ResponseEvent> {
run_stream_with_bytes(sse_body.as_bytes()).await
}
@@ -157,12 +153,7 @@ fn assert_reasoning(item: &ResponseItem, expected: &str) {
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn streams_text_without_reasoning() {
if network_disabled() {
println!(
"Skipping test because it cannot execute when network is disabled in a Codex sandbox."
);
return;
}
skip_if_no_network!();
let sse = concat!(
"data: {\"choices\":[{\"delta\":{\"content\":\"hi\"}}]}\n\n",
@@ -193,12 +184,7 @@ async fn streams_text_without_reasoning() {
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn streams_reasoning_from_string_delta() {
if network_disabled() {
println!(
"Skipping test because it cannot execute when network is disabled in a Codex sandbox."
);
return;
}
skip_if_no_network!();
let sse = concat!(
"data: {\"choices\":[{\"delta\":{\"reasoning\":\"think1\"}}]}\n\n",
@@ -215,7 +201,13 @@ async fn streams_reasoning_from_string_delta() {
}
match &events[1] {
ResponseEvent::ReasoningContentDelta(text) => assert_eq!(text, "think1"),
ResponseEvent::ReasoningContentDelta {
delta,
content_index,
} => {
assert_eq!(delta, "think1");
assert_eq!(content_index, &0);
}
other => panic!("expected reasoning delta, got {other:?}"),
}
@@ -244,12 +236,7 @@ async fn streams_reasoning_from_string_delta() {
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn streams_reasoning_from_object_delta() {
if network_disabled() {
println!(
"Skipping test because it cannot execute when network is disabled in a Codex sandbox."
);
return;
}
skip_if_no_network!();
let sse = concat!(
"data: {\"choices\":[{\"delta\":{\"reasoning\":{\"text\":\"partA\"}}}]}\n\n",
@@ -267,12 +254,24 @@ async fn streams_reasoning_from_object_delta() {
}
match &events[1] {
ResponseEvent::ReasoningContentDelta(text) => assert_eq!(text, "partA"),
ResponseEvent::ReasoningContentDelta {
delta,
content_index,
} => {
assert_eq!(delta, "partA");
assert_eq!(content_index, &0);
}
other => panic!("expected reasoning delta, got {other:?}"),
}
match &events[2] {
ResponseEvent::ReasoningContentDelta(text) => assert_eq!(text, "partB"),
ResponseEvent::ReasoningContentDelta {
delta,
content_index,
} => {
assert_eq!(delta, "partB");
assert_eq!(content_index, &1);
}
other => panic!("expected reasoning delta, got {other:?}"),
}
@@ -301,12 +300,7 @@ async fn streams_reasoning_from_object_delta() {
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn streams_reasoning_from_final_message() {
if network_disabled() {
println!(
"Skipping test because it cannot execute when network is disabled in a Codex sandbox."
);
return;
}
skip_if_no_network!();
let sse = "data: {\"choices\":[{\"message\":{\"reasoning\":\"final-cot\"},\"finish_reason\":\"stop\"}]}\n\n";
@@ -319,7 +313,13 @@ async fn streams_reasoning_from_final_message() {
}
match &events[1] {
ResponseEvent::ReasoningContentDelta(text) => assert_eq!(text, "final-cot"),
ResponseEvent::ReasoningContentDelta {
delta,
content_index,
} => {
assert_eq!(delta, "final-cot");
assert_eq!(content_index, &0);
}
other => panic!("expected reasoning delta, got {other:?}"),
}
@@ -333,12 +333,7 @@ async fn streams_reasoning_from_final_message() {
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn streams_reasoning_before_tool_call() {
if network_disabled() {
println!(
"Skipping test because it cannot execute when network is disabled in a Codex sandbox."
);
return;
}
skip_if_no_network!();
let sse = concat!(
"data: {\"choices\":[{\"delta\":{\"reasoning\":\"pre-tool\"}}]}\n\n",
@@ -354,7 +349,13 @@ async fn streams_reasoning_before_tool_call() {
}
match &events[1] {
ResponseEvent::ReasoningContentDelta(text) => assert_eq!(text, "pre-tool"),
ResponseEvent::ReasoningContentDelta {
delta,
content_index,
} => {
assert_eq!(delta, "pre-tool");
assert_eq!(content_index, &0);
}
other => panic!("expected reasoning delta, got {other:?}"),
}
@@ -383,12 +384,7 @@ async fn streams_reasoning_before_tool_call() {
#[tokio::test]
#[traced_test]
async fn chat_sse_emits_failed_on_parse_error() {
if network_disabled() {
println!(
"Skipping test because it cannot execute when network is disabled in a Codex sandbox."
);
return;
}
skip_if_no_network!();
let sse_body = concat!("data: not-json\n\n", "data: [DONE]\n\n");
@@ -420,12 +416,7 @@ async fn chat_sse_emits_failed_on_parse_error() {
#[tokio::test]
#[traced_test]
async fn chat_sse_done_chunk_emits_event() {
if network_disabled() {
println!(
"Skipping test because it cannot execute when network is disabled in a Codex sandbox."
);
return;
}
skip_if_no_network!();
let sse_body = "data: [DONE]\n\n";
@@ -443,12 +434,7 @@ async fn chat_sse_done_chunk_emits_event() {
#[tokio::test]
#[traced_test]
async fn chat_sse_emits_error_on_invalid_utf8() {
if network_disabled() {
println!(
"Skipping test because it cannot execute when network is disabled in a Codex sandbox."
);
return;
}
skip_if_no_network!();
let _ = run_stream_with_bytes(b"data: \x80\x80\n\n").await;
codex-rs/core/tests/common/responses.rs
+106
View File
@@ -1,6 +1,7 @@
use std::sync::Arc;
use std::sync::Mutex;
use anyhow::Result;
use serde_json::Value;
use wiremock::BodyPrintLimit;
use wiremock::Match;
@@ -12,6 +13,8 @@ use wiremock::ResponseTemplate;
use wiremock::matchers::method;
use wiremock::matchers::path_regex;
use crate::test_codex::ApplyPatchModelOutput;
#[derive(Debug, Clone)]
pub struct ResponseMock {
requests: Arc<Mutex<Vec<ResponsesRequest>>>,
@@ -36,6 +39,10 @@ impl ResponseMock {
self.requests.lock().unwrap().clone()
}
pub fn last_request(&self) -> Option<ResponsesRequest> {
self.requests.lock().unwrap().last().cloned()
}
/// Returns true if any captured request contains a `function_call` with the
/// provided `call_id`.
pub fn saw_function_call(&self, call_id: &str) -> bool {
@@ -128,6 +135,42 @@ impl ResponsesRequest {
.map(str::to_string)
}
pub fn function_call_output_content_and_success(
&self,
call_id: &str,
) -> Option<(Option<String>, Option<bool>)> {
self.call_output_content_and_success(call_id, "function_call_output")
}
pub fn custom_tool_call_output_content_and_success(
&self,
call_id: &str,
) -> Option<(Option<String>, Option<bool>)> {
self.call_output_content_and_success(call_id, "custom_tool_call_output")
}
fn call_output_content_and_success(
&self,
call_id: &str,
call_type: &str,
) -> Option<(Option<String>, Option<bool>)> {
let output = self
.call_output(call_id, call_type)
.get("output")
.cloned()
.unwrap_or(Value::Null);
match output {
Value::String(text) => Some((Some(text), None)),
Value::Object(obj) => Some((
obj.get("content")
.and_then(Value::as_str)
.map(str::to_string),
obj.get("success").and_then(Value::as_bool),
)),
_ => Some((None, None)),
}
}
pub fn header(&self, name: &str) -> Option<String> {
self.0
.headers
@@ -294,6 +337,7 @@ pub fn ev_reasoning_summary_text_delta(delta: &str) -> Value {
serde_json::json!({
"type": "response.reasoning_summary_text.delta",
"delta": delta,
"summary_index": 0,
})
}
@@ -301,6 +345,7 @@ pub fn ev_reasoning_text_delta(delta: &str) -> Value {
serde_json::json!({
"type": "response.reasoning_text.delta",
"delta": delta,
"content_index": 0,
})
}
@@ -367,6 +412,21 @@ pub fn ev_local_shell_call(call_id: &str, status: &str, command: Vec<&str>) -> V
})
}
pub fn ev_apply_patch_call(
call_id: &str,
patch: &str,
output_type: ApplyPatchModelOutput,
) -> Value {
match output_type {
ApplyPatchModelOutput::Freeform => ev_apply_patch_custom_tool_call(call_id, patch),
ApplyPatchModelOutput::Function => ev_apply_patch_function_call(call_id, patch),
ApplyPatchModelOutput::Shell => ev_apply_patch_shell_call(call_id, patch),
ApplyPatchModelOutput::ShellViaHeredoc => {
ev_apply_patch_shell_call_via_heredoc(call_id, patch)
}
}
}
/// Convenience: SSE event for an `apply_patch` custom tool call with raw patch
/// text. This mirrors the payload produced by the Responses API when the model
/// invokes `apply_patch` directly (before we convert it to a function call).
@@ -400,6 +460,21 @@ pub fn ev_apply_patch_function_call(call_id: &str, patch: &str) -> Value {
})
}
pub fn ev_apply_patch_shell_call(call_id: &str, patch: &str) -> Value {
let args = serde_json::json!({ "command": ["apply_patch", patch] });
let arguments = serde_json::to_string(&args).expect("serialize apply_patch arguments");
ev_function_call(call_id, "shell", &arguments)
}
pub fn ev_apply_patch_shell_call_via_heredoc(call_id: &str, patch: &str) -> Value {
let script = format!("apply_patch <<'EOF'\n{patch}\nEOF\n");
let args = serde_json::json!({ "command": ["bash", "-lc", script] });
let arguments = serde_json::to_string(&args).expect("serialize apply_patch arguments");
ev_function_call(call_id, "shell", &arguments)
}
pub fn sse_failed(id: &str, code: &str, message: &str) -> String {
sse(vec![serde_json::json!({
"type": "response.failed",
@@ -453,6 +528,37 @@ pub async fn start_mock_server() -> MockServer {
.await
}
#[derive(Clone)]
pub struct FunctionCallResponseMocks {
pub function_call: ResponseMock,
pub completion: ResponseMock,
}
pub async fn mount_function_call_agent_response(
server: &MockServer,
call_id: &str,
arguments: &str,
tool_name: &str,
) -> FunctionCallResponseMocks {
let first_response = sse(vec![
ev_response_created("resp-1"),
ev_function_call(call_id, tool_name, arguments),
ev_completed("resp-1"),
]);
let function_call = mount_sse_once(server, first_response).await;
let second_response = sse(vec![
ev_assistant_message("msg-1", "done"),
ev_completed("resp-2"),
]);
let completion = mount_sse_once(server, second_response).await;
FunctionCallResponseMocks {
function_call,
completion,
}
}
/// Mounts a sequence of SSE response bodies and serves them in order for each
/// POST to `/v1/responses`. Panics if more requests are received than bodies
/// provided. Also asserts the exact number of expected calls.
codex-rs/core/tests/common/test_codex.rs
+57 -5
View File
@@ -29,6 +29,24 @@ use crate::wait_for_event;
type ConfigMutator = dyn FnOnce(&mut Config) + Send;
/// A collection of different ways the model can output an apply_patch call
#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash)]
pub enum ApplyPatchModelOutput {
Freeform,
Function,
Shell,
ShellViaHeredoc,
}
/// A collection of different ways the model can output an apply_patch call
#[derive(Clone, Copy, Debug, PartialEq, Eq, Hash)]
pub enum ShellModelOutput {
Shell,
ShellCommand,
LocalShell,
// UnifiedExec has its own set of tests
}
pub struct TestCodexBuilder {
config_mutators: Vec<Box<ConfigMutator>>,
}
@@ -71,6 +89,7 @@ impl TestCodexBuilder {
resume_from: Option<PathBuf>,
) -> anyhow::Result<TestCodex> {
let (config, cwd) = self.prepare_config(server, &home).await?;
let conversation_manager = ConversationManager::with_auth(CodexAuth::from_api_key("dummy"));
let new_conversation = match resume_from {
@@ -79,15 +98,20 @@ impl TestCodexBuilder {
CodexAuth::from_api_key("dummy"),
);
conversation_manager
.resume_conversation_from_rollout(config, path, auth_manager)
.resume_conversation_from_rollout(config.clone(), path, auth_manager)
.await?
}
None => {
conversation_manager
.new_conversation(config.clone())
.await?
}
None => conversation_manager.new_conversation(config).await?,
};
Ok(TestCodex {
home,
cwd,
config,
codex: new_conversation.conversation,
session_configured: new_conversation.session_configured,
})
@@ -131,6 +155,7 @@ pub struct TestCodex {
pub cwd: Arc<TempDir>,
pub codex: Arc<CodexConversation>,
pub session_configured: SessionConfiguredEvent,
pub config: Config,
}
impl TestCodex {
@@ -143,14 +168,28 @@ impl TestCodex {
}
pub async fn submit_turn(&self, prompt: &str) -> Result<()> {
self.submit_turn_with_policy(prompt, SandboxPolicy::DangerFullAccess)
.await
self.submit_turn_with_policies(
prompt,
AskForApproval::Never,
SandboxPolicy::DangerFullAccess,
)
.await
}
pub async fn submit_turn_with_policy(
&self,
prompt: &str,
sandbox_policy: SandboxPolicy,
) -> Result<()> {
self.submit_turn_with_policies(prompt, AskForApproval::Never, sandbox_policy)
.await
}
pub async fn submit_turn_with_policies(
&self,
prompt: &str,
approval_policy: AskForApproval,
sandbox_policy: SandboxPolicy,
) -> Result<()> {
let session_model = self.session_configured.model.clone();
self.codex
@@ -160,7 +199,7 @@ impl TestCodex {
}],
final_output_json_schema: None,
cwd: self.cwd.path().to_path_buf(),
approval_policy: AskForApproval::Never,
approval_policy,
sandbox_policy,
model: session_model,
effort: None,
@@ -258,6 +297,19 @@ impl TestCodexHarness {
.expect("output string")
.to_string()
}
pub async fn apply_patch_output(
&self,
call_id: &str,
output_type: ApplyPatchModelOutput,
) -> String {
match output_type {
ApplyPatchModelOutput::Freeform => self.custom_tool_call_output(call_id).await,
ApplyPatchModelOutput::Function
| ApplyPatchModelOutput::Shell
| ApplyPatchModelOutput::ShellViaHeredoc => self.function_call_stdout(call_id).await,
}
}
}
fn custom_tool_call_output<'a>(bodies: &'a [Value], call_id: &str) -> &'a Value {
codex-rs/core/tests/suite/apply_patch_cli.rs
+208 -84
View File
@@ -1,6 +1,8 @@
#![allow(clippy::expect_used)]
use anyhow::Result;
use core_test_support::responses::ev_apply_patch_call;
use core_test_support::test_codex::ApplyPatchModelOutput;
use pretty_assertions::assert_eq;
use std::fs;
@@ -25,8 +27,9 @@ use core_test_support::skip_if_no_network;
use core_test_support::test_codex::TestCodexHarness;
use core_test_support::wait_for_event;
use serde_json::json;
use test_case::test_case;
async fn apply_patch_harness() -> Result<TestCodexHarness> {
pub async fn apply_patch_harness() -> Result<TestCodexHarness> {
apply_patch_harness_with(|_| {}).await
}
@@ -40,24 +43,30 @@ async fn apply_patch_harness_with(
.await
}
async fn mount_apply_patch(
pub async fn mount_apply_patch(
harness: &TestCodexHarness,
call_id: &str,
patch: &str,
assistant_msg: &str,
output_type: ApplyPatchModelOutput,
) {
mount_sse_sequence(
harness.server(),
apply_patch_responses(call_id, patch, assistant_msg),
apply_patch_responses(call_id, patch, assistant_msg, output_type),
)
.await;
}
fn apply_patch_responses(call_id: &str, patch: &str, assistant_msg: &str) -> Vec<String> {
fn apply_patch_responses(
call_id: &str,
patch: &str,
assistant_msg: &str,
output_type: ApplyPatchModelOutput,
) -> Vec<String> {
vec![
sse(vec![
ev_response_created("resp-1"),
ev_apply_patch_function_call(call_id, patch),
ev_apply_patch_call(call_id, patch, output_type),
ev_completed("resp-1"),
]),
sse(vec![
@@ -68,12 +77,18 @@ fn apply_patch_responses(call_id: &str, patch: &str, assistant_msg: &str) -> Vec
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn apply_patch_cli_multiple_operations_integration() -> Result<()> {
#[test_case(ApplyPatchModelOutput::Freeform)]
#[test_case(ApplyPatchModelOutput::Function)]
#[test_case(ApplyPatchModelOutput::Shell)]
#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
async fn apply_patch_cli_multiple_operations_integration(
output_type: ApplyPatchModelOutput,
) -> Result<()> {
skip_if_no_network!(Ok(()));
let harness = apply_patch_harness_with(|config| {
config.model = "gpt-5".to_string();
config.model_family = find_family_for_model("gpt-5").expect("gpt-5 is valid");
config.model = "gpt-5.1".to_string();
config.model_family = find_family_for_model("gpt-5.1").expect("gpt-5.1 is valid");
})
.await?;
@@ -86,11 +101,11 @@ async fn apply_patch_cli_multiple_operations_integration() -> Result<()> {
let patch = "*** Begin Patch\n*** Add File: nested/new.txt\n+created\n*** Delete File: delete.txt\n*** Update File: modify.txt\n@@\n-line2\n+changed\n*** End Patch";
let call_id = "apply-multi-ops";
mount_apply_patch(&harness, call_id, patch, "done").await;
mount_apply_patch(&harness, call_id, patch, "done", output_type).await;
harness.submit("please apply multi-ops patch").await?;
let out = harness.function_call_stdout(call_id).await;
let out = harness.apply_patch_output(call_id, output_type).await;
let expected = r"(?s)^Exit code: 0
Wall time: [0-9]+(?:\.[0-9]+)? seconds
@@ -113,7 +128,11 @@ D delete.txt
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn apply_patch_cli_multiple_chunks() -> Result<()> {
#[test_case(ApplyPatchModelOutput::Freeform)]
#[test_case(ApplyPatchModelOutput::Function)]
#[test_case(ApplyPatchModelOutput::Shell)]
#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
async fn apply_patch_cli_multiple_chunks(model_output: ApplyPatchModelOutput) -> Result<()> {
skip_if_no_network!(Ok(()));
let harness = apply_patch_harness().await?;
@@ -123,7 +142,7 @@ async fn apply_patch_cli_multiple_chunks() -> Result<()> {
let patch = "*** Begin Patch\n*** Update File: multi.txt\n@@\n-line2\n+changed2\n@@\n-line4\n+changed4\n*** End Patch";
let call_id = "apply-multi-chunks";
mount_apply_patch(&harness, call_id, patch, "ok").await;
mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
harness.submit("apply multi-chunk patch").await?;
@@ -135,7 +154,13 @@ async fn apply_patch_cli_multiple_chunks() -> Result<()> {
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn apply_patch_cli_moves_file_to_new_directory() -> Result<()> {
#[test_case(ApplyPatchModelOutput::Freeform)]
#[test_case(ApplyPatchModelOutput::Function)]
#[test_case(ApplyPatchModelOutput::Shell)]
#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
async fn apply_patch_cli_moves_file_to_new_directory(
model_output: ApplyPatchModelOutput,
) -> Result<()> {
skip_if_no_network!(Ok(()));
let harness = apply_patch_harness().await?;
@@ -147,7 +172,7 @@ async fn apply_patch_cli_moves_file_to_new_directory() -> Result<()> {
let patch = "*** Begin Patch\n*** Update File: old/name.txt\n*** Move to: renamed/dir/name.txt\n@@\n-old content\n+new content\n*** End Patch";
let call_id = "apply-move";
mount_apply_patch(&harness, call_id, patch, "ok").await;
mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
harness.submit("apply move patch").await?;
@@ -157,7 +182,13 @@ async fn apply_patch_cli_moves_file_to_new_directory() -> Result<()> {
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn apply_patch_cli_updates_file_appends_trailing_newline() -> Result<()> {
#[test_case(ApplyPatchModelOutput::Freeform)]
#[test_case(ApplyPatchModelOutput::Function)]
#[test_case(ApplyPatchModelOutput::Shell)]
#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
async fn apply_patch_cli_updates_file_appends_trailing_newline(
model_output: ApplyPatchModelOutput,
) -> Result<()> {
skip_if_no_network!(Ok(()));
let harness = apply_patch_harness().await?;
@@ -167,7 +198,7 @@ async fn apply_patch_cli_updates_file_appends_trailing_newline() -> Result<()> {
let patch = "*** Begin Patch\n*** Update File: no_newline.txt\n@@\n-no newline at end\n+first line\n+second line\n*** End Patch";
let call_id = "apply-append-nl";
mount_apply_patch(&harness, call_id, patch, "ok").await;
mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
harness.submit("apply newline patch").await?;
@@ -178,7 +209,13 @@ async fn apply_patch_cli_updates_file_appends_trailing_newline() -> Result<()> {
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn apply_patch_cli_insert_only_hunk_modifies_file() -> Result<()> {
#[test_case(ApplyPatchModelOutput::Freeform)]
#[test_case(ApplyPatchModelOutput::Function)]
#[test_case(ApplyPatchModelOutput::Shell)]
#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
async fn apply_patch_cli_insert_only_hunk_modifies_file(
model_output: ApplyPatchModelOutput,
) -> Result<()> {
skip_if_no_network!(Ok(()));
let harness = apply_patch_harness().await?;
@@ -188,7 +225,7 @@ async fn apply_patch_cli_insert_only_hunk_modifies_file() -> Result<()> {
let patch = "*** Begin Patch\n*** Update File: insert_only.txt\n@@\n alpha\n+beta\n omega\n*** End Patch";
let call_id = "apply-insert-only";
mount_apply_patch(&harness, call_id, patch, "ok").await;
mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
harness.submit("insert lines via apply_patch").await?;
@@ -197,7 +234,13 @@ async fn apply_patch_cli_insert_only_hunk_modifies_file() -> Result<()> {
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn apply_patch_cli_move_overwrites_existing_destination() -> Result<()> {
#[test_case(ApplyPatchModelOutput::Freeform)]
#[test_case(ApplyPatchModelOutput::Function)]
#[test_case(ApplyPatchModelOutput::Shell)]
#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
async fn apply_patch_cli_move_overwrites_existing_destination(
model_output: ApplyPatchModelOutput,
) -> Result<()> {
skip_if_no_network!(Ok(()));
let harness = apply_patch_harness().await?;
@@ -211,7 +254,7 @@ async fn apply_patch_cli_move_overwrites_existing_destination() -> Result<()> {
let patch = "*** Begin Patch\n*** Update File: old/name.txt\n*** Move to: renamed/dir/name.txt\n@@\n-from\n+new\n*** End Patch";
let call_id = "apply-move-overwrite";
mount_apply_patch(&harness, call_id, patch, "ok").await;
mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
harness.submit("apply move overwrite patch").await?;
@@ -221,7 +264,13 @@ async fn apply_patch_cli_move_overwrites_existing_destination() -> Result<()> {
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn apply_patch_cli_move_without_content_change_has_no_turn_diff() -> Result<()> {
#[test_case(ApplyPatchModelOutput::Freeform)]
#[test_case(ApplyPatchModelOutput::Function)]
#[test_case(ApplyPatchModelOutput::Shell)]
#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
async fn apply_patch_cli_move_without_content_change_has_no_turn_diff(
model_output: ApplyPatchModelOutput,
) -> Result<()> {
skip_if_no_network!(Ok(()));
let harness = apply_patch_harness().await?;
@@ -236,7 +285,7 @@ async fn apply_patch_cli_move_without_content_change_has_no_turn_diff() -> Resul
let patch = "*** Begin Patch\n*** Update File: old/name.txt\n*** Move to: renamed/name.txt\n@@\n same\n*** End Patch";
let call_id = "apply-move-no-change";
mount_apply_patch(&harness, call_id, patch, "ok").await;
mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
let model = test.session_configured.model.clone();
codex
@@ -272,7 +321,13 @@ async fn apply_patch_cli_move_without_content_change_has_no_turn_diff() -> Resul
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn apply_patch_cli_add_overwrites_existing_file() -> Result<()> {
#[test_case(ApplyPatchModelOutput::Freeform)]
#[test_case(ApplyPatchModelOutput::Function)]
#[test_case(ApplyPatchModelOutput::Shell)]
#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
async fn apply_patch_cli_add_overwrites_existing_file(
model_output: ApplyPatchModelOutput,
) -> Result<()> {
skip_if_no_network!(Ok(()));
let harness = apply_patch_harness().await?;
@@ -282,7 +337,7 @@ async fn apply_patch_cli_add_overwrites_existing_file() -> Result<()> {
let patch = "*** Begin Patch\n*** Add File: duplicate.txt\n+new content\n*** End Patch";
let call_id = "apply-add-overwrite";
mount_apply_patch(&harness, call_id, patch, "ok").await;
mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
harness.submit("apply add overwrite patch").await?;
@@ -291,18 +346,24 @@ async fn apply_patch_cli_add_overwrites_existing_file() -> Result<()> {
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn apply_patch_cli_rejects_invalid_hunk_header() -> Result<()> {
#[test_case(ApplyPatchModelOutput::Freeform)]
#[test_case(ApplyPatchModelOutput::Function)]
#[test_case(ApplyPatchModelOutput::Shell)]
#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
async fn apply_patch_cli_rejects_invalid_hunk_header(
model_output: ApplyPatchModelOutput,
) -> Result<()> {
skip_if_no_network!(Ok(()));
let harness = apply_patch_harness().await?;
let patch = "*** Begin Patch\n*** Frobnicate File: foo\n*** End Patch";
let call_id = "apply-invalid-header";
mount_apply_patch(&harness, call_id, patch, "ok").await;
mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
harness.submit("apply invalid header patch").await?;
let out = harness.function_call_stdout(call_id).await;
let out = harness.apply_patch_output(call_id, model_output).await;
assert!(
out.contains("apply_patch verification failed"),
@@ -316,7 +377,13 @@ async fn apply_patch_cli_rejects_invalid_hunk_header() -> Result<()> {
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn apply_patch_cli_reports_missing_context() -> Result<()> {
#[test_case(ApplyPatchModelOutput::Freeform)]
#[test_case(ApplyPatchModelOutput::Function)]
#[test_case(ApplyPatchModelOutput::Shell)]
#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
async fn apply_patch_cli_reports_missing_context(
model_output: ApplyPatchModelOutput,
) -> Result<()> {
skip_if_no_network!(Ok(()));
let harness = apply_patch_harness().await?;
@@ -327,11 +394,11 @@ async fn apply_patch_cli_reports_missing_context() -> Result<()> {
let patch =
"*** Begin Patch\n*** Update File: modify.txt\n@@\n-missing\n+changed\n*** End Patch";
let call_id = "apply-missing-context";
mount_apply_patch(&harness, call_id, patch, "ok").await;
mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
harness.submit("apply missing context patch").await?;
let out = harness.function_call_stdout(call_id).await;
let out = harness.apply_patch_output(call_id, model_output).await;
assert!(
out.contains("apply_patch verification failed"),
@@ -343,18 +410,24 @@ async fn apply_patch_cli_reports_missing_context() -> Result<()> {
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn apply_patch_cli_reports_missing_target_file() -> Result<()> {
#[test_case(ApplyPatchModelOutput::Freeform)]
#[test_case(ApplyPatchModelOutput::Function)]
#[test_case(ApplyPatchModelOutput::Shell)]
#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
async fn apply_patch_cli_reports_missing_target_file(
model_output: ApplyPatchModelOutput,
) -> Result<()> {
skip_if_no_network!(Ok(()));
let harness = apply_patch_harness().await?;
let patch = "*** Begin Patch\n*** Update File: missing.txt\n@@\n-nope\n+better\n*** End Patch";
let call_id = "apply-missing-file";
mount_apply_patch(&harness, call_id, patch, "fail").await;
mount_apply_patch(&harness, call_id, patch, "fail", model_output).await;
harness.submit("attempt to update a missing file").await?;
let out = harness.function_call_stdout(call_id).await;
let out = harness.apply_patch_output(call_id, model_output).await;
assert!(
out.contains("apply_patch verification failed"),
"expected verification failure message"
@@ -372,18 +445,24 @@ async fn apply_patch_cli_reports_missing_target_file() -> Result<()> {
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn apply_patch_cli_delete_missing_file_reports_error() -> Result<()> {
#[test_case(ApplyPatchModelOutput::Freeform)]
#[test_case(ApplyPatchModelOutput::Function)]
#[test_case(ApplyPatchModelOutput::Shell)]
#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
async fn apply_patch_cli_delete_missing_file_reports_error(
model_output: ApplyPatchModelOutput,
) -> Result<()> {
skip_if_no_network!(Ok(()));
let harness = apply_patch_harness().await?;
let patch = "*** Begin Patch\n*** Delete File: missing.txt\n*** End Patch";
let call_id = "apply-delete-missing";
mount_apply_patch(&harness, call_id, patch, "fail").await;
mount_apply_patch(&harness, call_id, patch, "fail", model_output).await;
harness.submit("attempt to delete missing file").await?;
let out = harness.function_call_stdout(call_id).await;
let out = harness.apply_patch_output(call_id, model_output).await;
assert!(
out.contains("apply_patch verification failed"),
@@ -402,18 +481,22 @@ async fn apply_patch_cli_delete_missing_file_reports_error() -> Result<()> {
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn apply_patch_cli_rejects_empty_patch() -> Result<()> {
#[test_case(ApplyPatchModelOutput::Freeform)]
#[test_case(ApplyPatchModelOutput::Function)]
#[test_case(ApplyPatchModelOutput::Shell)]
#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
async fn apply_patch_cli_rejects_empty_patch(model_output: ApplyPatchModelOutput) -> Result<()> {
skip_if_no_network!(Ok(()));
let harness = apply_patch_harness().await?;
let patch = "*** Begin Patch\n*** End Patch";
let call_id = "apply-empty";
mount_apply_patch(&harness, call_id, patch, "ok").await;
mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
harness.submit("apply empty patch").await?;
let out = harness.function_call_stdout(call_id).await;
let out = harness.apply_patch_output(call_id, model_output).await;
assert!(
out.contains("patch rejected: empty patch"),
"expected rejection for empty patch: {out}"
@@ -422,7 +505,13 @@ async fn apply_patch_cli_rejects_empty_patch() -> Result<()> {
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn apply_patch_cli_delete_directory_reports_verification_error() -> Result<()> {
#[test_case(ApplyPatchModelOutput::Freeform)]
#[test_case(ApplyPatchModelOutput::Function)]
#[test_case(ApplyPatchModelOutput::Shell)]
#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
async fn apply_patch_cli_delete_directory_reports_verification_error(
model_output: ApplyPatchModelOutput,
) -> Result<()> {
skip_if_no_network!(Ok(()));
let harness = apply_patch_harness().await?;
@@ -431,18 +520,24 @@ async fn apply_patch_cli_delete_directory_reports_verification_error() -> Result
let patch = "*** Begin Patch\n*** Delete File: dir\n*** End Patch";
let call_id = "apply-delete-dir";
mount_apply_patch(&harness, call_id, patch, "ok").await;
mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
harness.submit("delete a directory via apply_patch").await?;
let out = harness.function_call_stdout(call_id).await;
let out = harness.apply_patch_output(call_id, model_output).await;
assert!(out.contains("apply_patch verification failed"));
assert!(out.contains("Failed to read"));
Ok(())
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn apply_patch_cli_rejects_path_traversal_outside_workspace() -> Result<()> {
#[test_case(ApplyPatchModelOutput::Freeform)]
#[test_case(ApplyPatchModelOutput::Function)]
#[test_case(ApplyPatchModelOutput::Shell)]
#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
async fn apply_patch_cli_rejects_path_traversal_outside_workspace(
model_output: ApplyPatchModelOutput,
) -> Result<()> {
skip_if_no_network!(Ok(()));
let harness = apply_patch_harness().await?;
@@ -458,7 +553,7 @@ async fn apply_patch_cli_rejects_path_traversal_outside_workspace() -> Result<()
let patch = "*** Begin Patch\n*** Add File: ../escape.txt\n+outside\n*** End Patch";
let call_id = "apply-path-traversal";
mount_apply_patch(&harness, call_id, patch, "fail").await;
mount_apply_patch(&harness, call_id, patch, "fail", model_output).await;
let sandbox_policy = SandboxPolicy::WorkspaceWrite {
writable_roots: vec![],
@@ -473,7 +568,7 @@ async fn apply_patch_cli_rejects_path_traversal_outside_workspace() -> Result<()
)
.await?;
let out = harness.function_call_stdout(call_id).await;
let out = harness.apply_patch_output(call_id, model_output).await;
assert!(
out.contains(
"patch rejected: writing outside of the project; rejected by user approval settings"
@@ -488,7 +583,13 @@ async fn apply_patch_cli_rejects_path_traversal_outside_workspace() -> Result<()
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn apply_patch_cli_rejects_move_path_traversal_outside_workspace() -> Result<()> {
#[test_case(ApplyPatchModelOutput::Freeform)]
#[test_case(ApplyPatchModelOutput::Function)]
#[test_case(ApplyPatchModelOutput::Shell)]
#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
async fn apply_patch_cli_rejects_move_path_traversal_outside_workspace(
model_output: ApplyPatchModelOutput,
) -> Result<()> {
skip_if_no_network!(Ok(()));
let harness = apply_patch_harness().await?;
@@ -507,7 +608,7 @@ async fn apply_patch_cli_rejects_move_path_traversal_outside_workspace() -> Resu
let patch = "*** Begin Patch\n*** Update File: stay.txt\n*** Move to: ../escape-move.txt\n@@\n-from\n+to\n*** End Patch";
let call_id = "apply-move-traversal";
mount_apply_patch(&harness, call_id, patch, "fail").await;
mount_apply_patch(&harness, call_id, patch, "fail", model_output).await;
let sandbox_policy = SandboxPolicy::WorkspaceWrite {
writable_roots: vec![],
@@ -519,7 +620,7 @@ async fn apply_patch_cli_rejects_move_path_traversal_outside_workspace() -> Resu
.submit_with_policy("attempt move traversal via apply_patch", sandbox_policy)
.await?;
let out = harness.function_call_stdout(call_id).await;
let out = harness.apply_patch_output(call_id, model_output).await;
assert!(
out.contains(
"patch rejected: writing outside of the project; rejected by user approval settings"
@@ -535,7 +636,13 @@ async fn apply_patch_cli_rejects_move_path_traversal_outside_workspace() -> Resu
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn apply_patch_cli_verification_failure_has_no_side_effects() -> Result<()> {
#[test_case(ApplyPatchModelOutput::Freeform)]
#[test_case(ApplyPatchModelOutput::Function)]
#[test_case(ApplyPatchModelOutput::Shell)]
#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
async fn apply_patch_cli_verification_failure_has_no_side_effects(
model_output: ApplyPatchModelOutput,
) -> Result<()> {
skip_if_no_network!(Ok(()));
let harness = apply_patch_harness_with(|config| {
@@ -547,7 +654,7 @@ async fn apply_patch_cli_verification_failure_has_no_side_effects() -> Result<()
let call_id = "apply-partial-no-side-effects";
let patch = "*** Begin Patch\n*** Add File: created.txt\n+hello\n*** Update File: missing.txt\n@@\n-old\n+new\n*** End Patch";
mount_apply_patch(&harness, call_id, patch, "failed").await;
mount_apply_patch(&harness, call_id, patch, "failed", model_output).await;
harness.submit("attempt partial apply patch").await?;
@@ -564,8 +671,8 @@ async fn apply_patch_shell_heredoc_with_cd_updates_relative_workdir() -> Result<
skip_if_no_network!(Ok(()));
let harness = apply_patch_harness_with(|config| {
config.model = "gpt-5".to_string();
config.model_family = find_family_for_model("gpt-5").expect("gpt-5 is valid");
config.model = "gpt-5.1".to_string();
config.model_family = find_family_for_model("gpt-5.1").expect("gpt-5.1 is valid");
})
.await?;
@@ -610,8 +717,8 @@ async fn apply_patch_shell_failure_propagates_error_and_skips_diff() -> Result<(
skip_if_no_network!(Ok(()));
let harness = apply_patch_harness_with(|config| {
config.model = "gpt-5".to_string();
config.model_family = find_family_for_model("gpt-5").expect("gpt-5 is valid");
config.model = "gpt-5.1".to_string();
config.model_family = find_family_for_model("gpt-5.1").expect("gpt-5.1 is valid");
})
.await?;
let test = harness.test();
@@ -700,7 +807,14 @@ async fn apply_patch_function_accepts_lenient_heredoc_wrapped_patch() -> Result<
format!("*** Begin Patch\n*** Add File: {file_name}\n+lenient\n*** End Patch\n");
let wrapped = format!("<<'EOF'\n{patch_inner}EOF\n");
let call_id = "apply-lenient";
mount_apply_patch(&harness, call_id, &wrapped, "ok").await;
mount_apply_patch(
&harness,
call_id,
wrapped.as_str(),
"ok",
ApplyPatchModelOutput::Function,
)
.await;
harness.submit("apply lenient heredoc patch").await?;
@@ -710,7 +824,11 @@ async fn apply_patch_function_accepts_lenient_heredoc_wrapped_patch() -> Result<
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn apply_patch_cli_end_of_file_anchor() -> Result<()> {
#[test_case(ApplyPatchModelOutput::Freeform)]
#[test_case(ApplyPatchModelOutput::Function)]
#[test_case(ApplyPatchModelOutput::Shell)]
#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
async fn apply_patch_cli_end_of_file_anchor(model_output: ApplyPatchModelOutput) -> Result<()> {
skip_if_no_network!(Ok(()));
let harness = apply_patch_harness().await?;
@@ -720,7 +838,7 @@ async fn apply_patch_cli_end_of_file_anchor() -> Result<()> {
let patch = "*** Begin Patch\n*** Update File: tail.txt\n@@\n-last\n+end\n*** End of File\n*** End Patch";
let call_id = "apply-eof";
mount_apply_patch(&harness, call_id, patch, "ok").await;
mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
harness.submit("apply EOF-anchored patch").await?;
assert_eq!(fs::read_to_string(&target)?, "alpha\nend\n");
@@ -728,7 +846,13 @@ async fn apply_patch_cli_end_of_file_anchor() -> Result<()> {
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn apply_patch_cli_missing_second_chunk_context_rejected() -> Result<()> {
#[test_case(ApplyPatchModelOutput::Freeform)]
#[test_case(ApplyPatchModelOutput::Function)]
#[test_case(ApplyPatchModelOutput::Shell)]
#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
async fn apply_patch_cli_missing_second_chunk_context_rejected(
model_output: ApplyPatchModelOutput,
) -> Result<()> {
skip_if_no_network!(Ok(()));
let harness = apply_patch_harness().await?;
@@ -740,11 +864,11 @@ async fn apply_patch_cli_missing_second_chunk_context_rejected() -> Result<()> {
let patch =
"*** Begin Patch\n*** Update File: two_chunks.txt\n@@\n-b\n+B\n\n-d\n+D\n*** End Patch";
let call_id = "apply-missing-ctx-2nd";
mount_apply_patch(&harness, call_id, patch, "fail").await;
mount_apply_patch(&harness, call_id, patch, "fail", model_output).await;
harness.submit("apply missing context second chunk").await?;
let out = harness.function_call_stdout(call_id).await;
let out = harness.apply_patch_output(call_id, model_output).await;
assert!(out.contains("apply_patch verification failed"));
assert!(
out.contains("Failed to find expected lines in"),
@@ -756,7 +880,13 @@ async fn apply_patch_cli_missing_second_chunk_context_rejected() -> Result<()> {
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn apply_patch_emits_turn_diff_event_with_unified_diff() -> Result<()> {
#[test_case(ApplyPatchModelOutput::Freeform)]
#[test_case(ApplyPatchModelOutput::Function)]
#[test_case(ApplyPatchModelOutput::Shell)]
#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
async fn apply_patch_emits_turn_diff_event_with_unified_diff(
model_output: ApplyPatchModelOutput,
) -> Result<()> {
skip_if_no_network!(Ok(()));
let harness = apply_patch_harness().await?;
@@ -767,16 +897,7 @@ async fn apply_patch_emits_turn_diff_event_with_unified_diff() -> Result<()> {
let call_id = "apply-diff-event";
let file = "udiff.txt";
let patch = format!("*** Begin Patch\n*** Add File: {file}\n+hello\n*** End Patch\n");
let first = sse(vec![
ev_response_created("resp-1"),
ev_apply_patch_function_call(call_id, &patch),
ev_completed("resp-1"),
]);
let second = sse(vec![
ev_assistant_message("msg-1", "ok"),
ev_completed("resp-2"),
]);
mount_sse_sequence(harness.server(), vec![first, second]).await;
mount_apply_patch(&harness, call_id, patch.as_str(), "ok", model_output).await;
let model = test.session_configured.model.clone();
codex
@@ -814,7 +935,13 @@ async fn apply_patch_emits_turn_diff_event_with_unified_diff() -> Result<()> {
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn apply_patch_turn_diff_for_rename_with_content_change() -> Result<()> {
#[test_case(ApplyPatchModelOutput::Freeform)]
#[test_case(ApplyPatchModelOutput::Function)]
#[test_case(ApplyPatchModelOutput::Shell)]
#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
async fn apply_patch_turn_diff_for_rename_with_content_change(
model_output: ApplyPatchModelOutput,
) -> Result<()> {
skip_if_no_network!(Ok(()));
let harness = apply_patch_harness().await?;
@@ -829,16 +956,7 @@ async fn apply_patch_turn_diff_for_rename_with_content_change() -> Result<()> {
// Patch: update + move
let call_id = "apply-rename-change";
let patch = "*** Begin Patch\n*** Update File: old.txt\n*** Move to: new.txt\n@@\n-old\n+new\n*** End Patch";
let first = sse(vec![
ev_response_created("resp-1"),
ev_apply_patch_function_call(call_id, patch),
ev_completed("resp-1"),
]);
let second = sse(vec![
ev_assistant_message("msg-1", "ok"),
ev_completed("resp-2"),
]);
mount_sse_sequence(harness.server(), vec![first, second]).await;
mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
let model = test.session_configured.model.clone();
codex
@@ -1031,7 +1149,13 @@ async fn apply_patch_aggregates_diff_preserves_success_after_failure() -> Result
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn apply_patch_change_context_disambiguates_target() -> Result<()> {
#[test_case(ApplyPatchModelOutput::Freeform)]
#[test_case(ApplyPatchModelOutput::Function)]
#[test_case(ApplyPatchModelOutput::Shell)]
#[test_case(ApplyPatchModelOutput::ShellViaHeredoc)]
async fn apply_patch_change_context_disambiguates_target(
model_output: ApplyPatchModelOutput,
) -> Result<()> {
skip_if_no_network!(Ok(()));
let harness = apply_patch_harness().await?;
@@ -1042,7 +1166,7 @@ async fn apply_patch_change_context_disambiguates_target() -> Result<()> {
let patch =
"*** Begin Patch\n*** Update File: multi_ctx.txt\n@@ fn b\n-x=10\n+x=11\n*** End Patch";
let call_id = "apply-ctx";
mount_apply_patch(&harness, call_id, patch, "ok").await;
mount_apply_patch(&harness, call_id, patch, "ok", model_output).await;
harness.submit("apply with change_context").await?;
codex-rs/core/tests/suite/apply_patch_freeform.rs
-1001
View File
File diff suppressed because it is too large Load Diff
codex-rs/core/tests/suite/approvals.rs
+278 -28
View File
@@ -240,6 +240,10 @@ enum Expectation {
target: TargetPath,
content: &'static str,
},
FileCreatedNoExitCode {
target: TargetPath,
content: &'static str,
},
PatchApplied {
target: TargetPath,
content: &'static str,
@@ -251,12 +255,18 @@ enum Expectation {
NetworkSuccess {
body_contains: &'static str,
},
NetworkSuccessNoExitCode {
body_contains: &'static str,
},
NetworkFailure {
expect_tag: &'static str,
},
CommandSuccess {
stdout_contains: &'static str,
},
CommandSuccessNoExitCode {
stdout_contains: &'static str,
},
CommandFailure {
output_contains: &'static str,
},
@@ -270,8 +280,7 @@ impl Expectation {
assert_eq!(
result.exit_code,
Some(0),
"expected successful exit for {:?}",
path
"expected successful exit for {path:?}"
);
assert!(
result.stdout.contains(content),
@@ -285,6 +294,21 @@ impl Expectation {
);
let _ = fs::remove_file(path);
}
Expectation::FileCreatedNoExitCode { target, content } => {
let (path, _) = target.resolve_for_patch(test);
assert_eq!(result.exit_code, None, "expected no exit code for {path:?}");
assert!(
result.stdout.contains(content),
"stdout missing {content:?}: {}",
result.stdout
);
let file_contents = fs::read_to_string(&path)?;
assert!(
file_contents.contains(content),
"file contents missing {content:?}: {file_contents}"
);
let _ = fs::remove_file(path);
}
Expectation::PatchApplied { target, content } => {
let (path, _) = target.resolve_for_patch(test);
match result.exit_code {
@@ -360,6 +384,23 @@ impl Expectation {
result.stdout
);
}
Expectation::NetworkSuccessNoExitCode { body_contains } => {
assert_eq!(
result.exit_code, None,
"expected no exit code for successful network call: {}",
result.stdout
);
assert!(
result.stdout.contains("OK:"),
"stdout missing OK prefix: {}",
result.stdout
);
assert!(
result.stdout.contains(body_contains),
"stdout missing body text {body_contains:?}: {}",
result.stdout
);
}
Expectation::NetworkFailure { expect_tag } => {
assert_ne!(
result.exit_code,
@@ -391,6 +432,18 @@ impl Expectation {
result.stdout
);
}
Expectation::CommandSuccessNoExitCode { stdout_contains } => {
assert_eq!(
result.exit_code, None,
"expected no exit code for trusted command: {}",
result.stdout
);
assert!(
result.stdout.contains(stdout_contains),
"trusted command stdout missing {stdout_contains:?}: {}",
result.stdout
);
}
Expectation::CommandFailure { output_contains } => {
assert_ne!(
result.exit_code,
@@ -588,13 +641,30 @@ fn scenarios() -> Vec<ScenarioSpec> {
},
with_escalated_permissions: false,
features: vec![],
model_override: None,
model_override: Some("gpt-5"),
outcome: Outcome::Auto,
expectation: Expectation::FileCreated {
target: TargetPath::OutsideWorkspace("dfa_on_request.txt"),
content: "danger-on-request",
},
},
ScenarioSpec {
name: "danger_full_access_on_request_allows_outside_write_gpt_5_1_no_exit",
approval_policy: OnRequest,
sandbox_policy: SandboxPolicy::DangerFullAccess,
action: ActionKind::WriteFile {
target: TargetPath::OutsideWorkspace("dfa_on_request_5_1.txt"),
content: "danger-on-request",
},
with_escalated_permissions: false,
features: vec![],
model_override: Some("gpt-5.1"),
outcome: Outcome::Auto,
expectation: Expectation::FileCreatedNoExitCode {
target: TargetPath::OutsideWorkspace("dfa_on_request_5_1.txt"),
content: "danger-on-request",
},
},
ScenarioSpec {
name: "danger_full_access_on_request_allows_network",
approval_policy: OnRequest,
@@ -605,12 +675,28 @@ fn scenarios() -> Vec<ScenarioSpec> {
},
with_escalated_permissions: false,
features: vec![],
model_override: None,
model_override: Some("gpt-5"),
outcome: Outcome::Auto,
expectation: Expectation::NetworkSuccess {
body_contains: "danger-network-ok",
},
},
ScenarioSpec {
name: "danger_full_access_on_request_allows_network_gpt_5_1_no_exit",
approval_policy: OnRequest,
sandbox_policy: SandboxPolicy::DangerFullAccess,
action: ActionKind::FetchUrl {
endpoint: "/dfa/network",
response_body: "danger-network-ok",
},
with_escalated_permissions: false,
features: vec![],
model_override: Some("gpt-5.1"),
outcome: Outcome::Auto,
expectation: Expectation::NetworkSuccessNoExitCode {
body_contains: "danger-network-ok",
},
},
ScenarioSpec {
name: "trusted_command_unless_trusted_runs_without_prompt",
approval_policy: UnlessTrusted,
@@ -620,12 +706,27 @@ fn scenarios() -> Vec<ScenarioSpec> {
},
with_escalated_permissions: false,
features: vec![],
model_override: None,
model_override: Some("gpt-5"),
outcome: Outcome::Auto,
expectation: Expectation::CommandSuccess {
stdout_contains: "trusted-unless",
},
},
ScenarioSpec {
name: "trusted_command_unless_trusted_runs_without_prompt_gpt_5_1_no_exit",
approval_policy: UnlessTrusted,
sandbox_policy: SandboxPolicy::DangerFullAccess,
action: ActionKind::RunCommand {
command: &["echo", "trusted-unless"],
},
with_escalated_permissions: false,
features: vec![],
model_override: Some("gpt-5.1"),
outcome: Outcome::Auto,
expectation: Expectation::CommandSuccessNoExitCode {
stdout_contains: "trusted-unless",
},
},
ScenarioSpec {
name: "danger_full_access_on_failure_allows_outside_write",
approval_policy: OnFailure,
@@ -636,13 +737,30 @@ fn scenarios() -> Vec<ScenarioSpec> {
},
with_escalated_permissions: false,
features: vec![],
model_override: None,
model_override: Some("gpt-5"),
outcome: Outcome::Auto,
expectation: Expectation::FileCreated {
target: TargetPath::OutsideWorkspace("dfa_on_failure.txt"),
content: "danger-on-failure",
},
},
ScenarioSpec {
name: "danger_full_access_on_failure_allows_outside_write_gpt_5_1_no_exit",
approval_policy: OnFailure,
sandbox_policy: SandboxPolicy::DangerFullAccess,
action: ActionKind::WriteFile {
target: TargetPath::OutsideWorkspace("dfa_on_failure_5_1.txt"),
content: "danger-on-failure",
},
with_escalated_permissions: false,
features: vec![],
model_override: Some("gpt-5.1"),
outcome: Outcome::Auto,
expectation: Expectation::FileCreatedNoExitCode {
target: TargetPath::OutsideWorkspace("dfa_on_failure_5_1.txt"),
content: "danger-on-failure",
},
},
ScenarioSpec {
name: "danger_full_access_unless_trusted_requests_approval",
approval_policy: UnlessTrusted,
@@ -653,7 +771,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
},
with_escalated_permissions: false,
features: vec![],
model_override: None,
model_override: Some("gpt-5"),
outcome: Outcome::ExecApproval {
decision: ReviewDecision::Approved,
expected_reason: None,
@@ -663,6 +781,26 @@ fn scenarios() -> Vec<ScenarioSpec> {
content: "danger-unless-trusted",
},
},
ScenarioSpec {
name: "danger_full_access_unless_trusted_requests_approval_gpt_5_1_no_exit",
approval_policy: UnlessTrusted,
sandbox_policy: SandboxPolicy::DangerFullAccess,
action: ActionKind::WriteFile {
target: TargetPath::OutsideWorkspace("dfa_unless_trusted_5_1.txt"),
content: "danger-unless-trusted",
},
with_escalated_permissions: false,
features: vec![],
model_override: Some("gpt-5.1"),
outcome: Outcome::ExecApproval {
decision: ReviewDecision::Approved,
expected_reason: None,
},
expectation: Expectation::FileCreatedNoExitCode {
target: TargetPath::OutsideWorkspace("dfa_unless_trusted_5_1.txt"),
content: "danger-unless-trusted",
},
},
ScenarioSpec {
name: "danger_full_access_never_allows_outside_write",
approval_policy: Never,
@@ -673,13 +811,30 @@ fn scenarios() -> Vec<ScenarioSpec> {
},
with_escalated_permissions: false,
features: vec![],
model_override: None,
model_override: Some("gpt-5"),
outcome: Outcome::Auto,
expectation: Expectation::FileCreated {
target: TargetPath::OutsideWorkspace("dfa_never.txt"),
content: "danger-never",
},
},
ScenarioSpec {
name: "danger_full_access_never_allows_outside_write_gpt_5_1_no_exit",
approval_policy: Never,
sandbox_policy: SandboxPolicy::DangerFullAccess,
action: ActionKind::WriteFile {
target: TargetPath::OutsideWorkspace("dfa_never_5_1.txt"),
content: "danger-never",
},
with_escalated_permissions: false,
features: vec![],
model_override: Some("gpt-5.1"),
outcome: Outcome::Auto,
expectation: Expectation::FileCreatedNoExitCode {
target: TargetPath::OutsideWorkspace("dfa_never_5_1.txt"),
content: "danger-never",
},
},
ScenarioSpec {
name: "read_only_on_request_requires_approval",
approval_policy: OnRequest,
@@ -690,7 +845,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
},
with_escalated_permissions: true,
features: vec![],
model_override: None,
model_override: Some("gpt-5"),
outcome: Outcome::ExecApproval {
decision: ReviewDecision::Approved,
expected_reason: None,
@@ -700,6 +855,26 @@ fn scenarios() -> Vec<ScenarioSpec> {
content: "read-only-approval",
},
},
ScenarioSpec {
name: "read_only_on_request_requires_approval_gpt_5_1_no_exit",
approval_policy: OnRequest,
sandbox_policy: SandboxPolicy::ReadOnly,
action: ActionKind::WriteFile {
target: TargetPath::Workspace("ro_on_request_5_1.txt"),
content: "read-only-approval",
},
with_escalated_permissions: true,
features: vec![],
model_override: Some("gpt-5.1"),
outcome: Outcome::ExecApproval {
decision: ReviewDecision::Approved,
expected_reason: None,
},
expectation: Expectation::FileCreatedNoExitCode {
target: TargetPath::Workspace("ro_on_request_5_1.txt"),
content: "read-only-approval",
},
},
ScenarioSpec {
name: "trusted_command_on_request_read_only_runs_without_prompt",
approval_policy: OnRequest,
@@ -709,12 +884,27 @@ fn scenarios() -> Vec<ScenarioSpec> {
},
with_escalated_permissions: false,
features: vec![],
model_override: None,
model_override: Some("gpt-5"),
outcome: Outcome::Auto,
expectation: Expectation::CommandSuccess {
stdout_contains: "trusted-read-only",
},
},
ScenarioSpec {
name: "trusted_command_on_request_read_only_runs_without_prompt_gpt_5_1_no_exit",
approval_policy: OnRequest,
sandbox_policy: SandboxPolicy::ReadOnly,
action: ActionKind::RunCommand {
command: &["echo", "trusted-read-only"],
},
with_escalated_permissions: false,
features: vec![],
model_override: Some("gpt-5.1"),
outcome: Outcome::Auto,
expectation: Expectation::CommandSuccessNoExitCode {
stdout_contains: "trusted-read-only",
},
},
ScenarioSpec {
name: "read_only_on_request_blocks_network",
approval_policy: OnRequest,
@@ -760,7 +950,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
},
with_escalated_permissions: false,
features: vec![],
model_override: None,
model_override: Some("gpt-5"),
outcome: Outcome::ExecApproval {
decision: ReviewDecision::Approved,
expected_reason: Some("command failed; retry without sandbox?"),
@@ -770,6 +960,27 @@ fn scenarios() -> Vec<ScenarioSpec> {
content: "read-only-on-failure",
},
},
#[cfg(not(target_os = "linux"))]
ScenarioSpec {
name: "read_only_on_failure_escalates_after_sandbox_error_gpt_5_1_no_exit",
approval_policy: OnFailure,
sandbox_policy: SandboxPolicy::ReadOnly,
action: ActionKind::WriteFile {
target: TargetPath::Workspace("ro_on_failure_5_1.txt"),
content: "read-only-on-failure",
},
with_escalated_permissions: false,
features: vec![],
model_override: Some("gpt-5.1"),
outcome: Outcome::ExecApproval {
decision: ReviewDecision::Approved,
expected_reason: Some("command failed; retry without sandbox?"),
},
expectation: Expectation::FileCreatedNoExitCode {
target: TargetPath::Workspace("ro_on_failure_5_1.txt"),
content: "read-only-on-failure",
},
},
ScenarioSpec {
name: "read_only_on_request_network_escalates_when_approved",
approval_policy: OnRequest,
@@ -780,7 +991,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
},
with_escalated_permissions: true,
features: vec![],
model_override: None,
model_override: Some("gpt-5"),
outcome: Outcome::ExecApproval {
decision: ReviewDecision::Approved,
expected_reason: None,
@@ -789,6 +1000,25 @@ fn scenarios() -> Vec<ScenarioSpec> {
body_contains: "read-only-network-ok",
},
},
ScenarioSpec {
name: "read_only_on_request_network_escalates_when_approved_gpt_5_1_no_exit",
approval_policy: OnRequest,
sandbox_policy: SandboxPolicy::ReadOnly,
action: ActionKind::FetchUrl {
endpoint: "/ro/network-approved",
response_body: "read-only-network-ok",
},
with_escalated_permissions: true,
features: vec![],
model_override: Some("gpt-5.1"),
outcome: Outcome::ExecApproval {
decision: ReviewDecision::Approved,
expected_reason: None,
},
expectation: Expectation::NetworkSuccessNoExitCode {
body_contains: "read-only-network-ok",
},
},
ScenarioSpec {
name: "apply_patch_shell_requires_patch_approval",
approval_policy: UnlessTrusted,
@@ -819,7 +1049,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
},
with_escalated_permissions: false,
features: vec![],
model_override: Some("gpt-5-codex"),
model_override: Some("gpt-5.1-codex"),
outcome: Outcome::Auto,
expectation: Expectation::PatchApplied {
target: TargetPath::Workspace("apply_patch_function.txt"),
@@ -836,7 +1066,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
},
with_escalated_permissions: false,
features: vec![Feature::ApplyPatchFreeform],
model_override: Some("gpt-5-codex"),
model_override: Some("gpt-5.1-codex"),
outcome: Outcome::Auto,
expectation: Expectation::PatchApplied {
target: TargetPath::OutsideWorkspace("apply_patch_function_danger.txt"),
@@ -853,7 +1083,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
},
with_escalated_permissions: false,
features: vec![],
model_override: Some("gpt-5-codex"),
model_override: Some("gpt-5.1-codex"),
outcome: Outcome::PatchApproval {
decision: ReviewDecision::Approved,
expected_reason: None,
@@ -873,7 +1103,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
},
with_escalated_permissions: false,
features: vec![],
model_override: Some("gpt-5-codex"),
model_override: Some("gpt-5.1-codex"),
outcome: Outcome::PatchApproval {
decision: ReviewDecision::Denied,
expected_reason: None,
@@ -913,7 +1143,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
},
with_escalated_permissions: false,
features: vec![],
model_override: Some("gpt-5-codex"),
model_override: Some("gpt-5.1-codex"),
outcome: Outcome::PatchApproval {
decision: ReviewDecision::Approved,
expected_reason: None,
@@ -933,7 +1163,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
},
with_escalated_permissions: false,
features: vec![],
model_override: Some("gpt-5-codex"),
model_override: Some("gpt-5.1-codex"),
outcome: Outcome::Auto,
expectation: Expectation::FileNotCreated {
target: TargetPath::OutsideWorkspace("apply_patch_function_never.txt"),
@@ -952,7 +1182,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
},
with_escalated_permissions: false,
features: vec![],
model_override: None,
model_override: Some("gpt-5"),
outcome: Outcome::ExecApproval {
decision: ReviewDecision::Approved,
expected_reason: None,
@@ -962,6 +1192,26 @@ fn scenarios() -> Vec<ScenarioSpec> {
content: "read-only-unless-trusted",
},
},
ScenarioSpec {
name: "read_only_unless_trusted_requires_approval_gpt_5_1_no_exit",
approval_policy: UnlessTrusted,
sandbox_policy: SandboxPolicy::ReadOnly,
action: ActionKind::WriteFile {
target: TargetPath::Workspace("ro_unless_trusted_5_1.txt"),
content: "read-only-unless-trusted",
},
with_escalated_permissions: false,
features: vec![],
model_override: Some("gpt-5.1"),
outcome: Outcome::ExecApproval {
decision: ReviewDecision::Approved,
expected_reason: None,
},
expectation: Expectation::FileCreatedNoExitCode {
target: TargetPath::Workspace("ro_unless_trusted_5_1.txt"),
content: "read-only-unless-trusted",
},
},
ScenarioSpec {
name: "read_only_never_reports_sandbox_failure",
approval_policy: Never,
@@ -992,7 +1242,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
},
with_escalated_permissions: false,
features: vec![],
model_override: None,
model_override: Some("gpt-5"),
outcome: Outcome::Auto,
expectation: Expectation::CommandSuccess {
stdout_contains: "trusted-never",
@@ -1008,7 +1258,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
},
with_escalated_permissions: false,
features: vec![],
model_override: None,
model_override: Some("gpt-5"),
outcome: Outcome::Auto,
expectation: Expectation::FileCreated {
target: TargetPath::Workspace("ww_on_request.txt"),
@@ -1039,7 +1289,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
},
with_escalated_permissions: true,
features: vec![],
model_override: None,
model_override: Some("gpt-5"),
outcome: Outcome::ExecApproval {
decision: ReviewDecision::Approved,
expected_reason: None,
@@ -1059,7 +1309,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
},
with_escalated_permissions: false,
features: vec![],
model_override: None,
model_override: Some("gpt-5"),
outcome: Outcome::Auto,
expectation: Expectation::NetworkSuccess {
body_contains: "workspace-network-ok",
@@ -1076,7 +1326,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
},
with_escalated_permissions: false,
features: vec![],
model_override: None,
model_override: Some("gpt-5"),
outcome: Outcome::ExecApproval {
decision: ReviewDecision::Approved,
expected_reason: Some("command failed; retry without sandbox?"),
@@ -1096,7 +1346,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
},
with_escalated_permissions: false,
features: vec![],
model_override: None,
model_override: Some("gpt-5"),
outcome: Outcome::ExecApproval {
decision: ReviewDecision::Approved,
expected_reason: None,
@@ -1137,7 +1387,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
},
with_escalated_permissions: false,
features: vec![Feature::UnifiedExec],
model_override: None,
model_override: Some("gpt-5"),
outcome: Outcome::Auto,
expectation: Expectation::CommandSuccess {
stdout_contains: "hello unified exec",
@@ -1155,7 +1405,7 @@ fn scenarios() -> Vec<ScenarioSpec> {
},
with_escalated_permissions: true,
features: vec![Feature::UnifiedExec],
model_override: None,
model_override: Some("gpt-5"),
outcome: Outcome::ExecApproval {
decision: ReviewDecision::Approved,
expected_reason: Some(DEFAULT_UNIFIED_EXEC_JUSTIFICATION),
@@ -1208,7 +1458,7 @@ async fn run_scenario(scenario: &ScenarioSpec) -> Result<()> {
let mut builder = test_codex().with_config(move |config| {
config.approval_policy = approval_policy;
config.sandbox_policy = sandbox_policy.clone();
let model = model_override.unwrap_or("gpt-5");
let model = model_override.unwrap_or("gpt-5.1");
config.model = model.to_string();
config.model_family =
find_family_for_model(model).expect("model should map to a known family");
codex-rs/core/tests/suite/cli_stream.rs
+1 -6
View File
@@ -110,12 +110,7 @@ async fn exec_cli_applies_experimental_instructions_file() {
"data: {\"type\":\"response.created\",\"response\":{}}\n\n",
"data: {\"type\":\"response.completed\",\"response\":{\"id\":\"r1\"}}\n\n"
);
let resp_mock = core_test_support::responses::mount_sse_once_match(
&server,
path("/v1/responses"),
sse.to_string(),
)
.await;
let resp_mock = core_test_support::responses::mount_sse_once(&server, sse.to_string()).await;
// Create a temporary instructions file with a unique marker we can assert
// appears in the outbound request payload.
codex-rs/core/tests/suite/client.rs
+16 -35
View File
@@ -238,9 +238,7 @@ async fn resume_includes_initial_messages_and_sends_prior_items() {
// Mock server that will receive the resumed request
let server = MockServer::start().await;
let resp_mock =
responses::mount_sse_once_match(&server, path("/v1/responses"), sse_completed("resp1"))
.await;
let resp_mock = responses::mount_sse_once(&server, sse_completed("resp1")).await;
// Configure Codex to resume from our file
let model_provider = ModelProviderInfo {
@@ -381,9 +379,7 @@ async fn includes_base_instructions_override_in_request() {
skip_if_no_network!();
// Mock server
let server = MockServer::start().await;
let resp_mock =
responses::mount_sse_once_match(&server, path("/v1/responses"), sse_completed("resp1"))
.await;
let resp_mock = responses::mount_sse_once(&server, sse_completed("resp1")).await;
let model_provider = ModelProviderInfo {
base_url: Some(format!("{}/v1", server.uri())),
@@ -571,9 +567,7 @@ async fn includes_user_instructions_message_in_request() {
skip_if_no_network!();
let server = MockServer::start().await;
let resp_mock =
responses::mount_sse_once_match(&server, path("/v1/responses"), sse_completed("resp1"))
.await;
let resp_mock = responses::mount_sse_once(&server, sse_completed("resp1")).await;
let model_provider = ModelProviderInfo {
base_url: Some(format!("{}/v1", server.uri())),
@@ -631,9 +625,7 @@ async fn includes_configured_effort_in_request() -> anyhow::Result<()> {
skip_if_no_network!(Ok(()));
let server = MockServer::start().await;
let resp_mock =
responses::mount_sse_once_match(&server, path("/v1/responses"), sse_completed("resp1"))
.await;
let resp_mock = responses::mount_sse_once(&server, sse_completed("resp1")).await;
let TestCodex { codex, .. } = test_codex()
.with_model("gpt-5.1-codex")
.with_config(|config| {
@@ -672,9 +664,7 @@ async fn includes_no_effort_in_request() -> anyhow::Result<()> {
skip_if_no_network!(Ok(()));
let server = MockServer::start().await;
let resp_mock =
responses::mount_sse_once_match(&server, path("/v1/responses"), sse_completed("resp1"))
.await;
let resp_mock = responses::mount_sse_once(&server, sse_completed("resp1")).await;
let TestCodex { codex, .. } = test_codex()
.with_model("gpt-5.1-codex")
.build(&server)
@@ -711,9 +701,7 @@ async fn includes_default_reasoning_effort_in_request_when_defined_by_model_fami
skip_if_no_network!(Ok(()));
let server = MockServer::start().await;
let resp_mock =
responses::mount_sse_once_match(&server, path("/v1/responses"), sse_completed("resp1"))
.await;
let resp_mock = responses::mount_sse_once(&server, sse_completed("resp1")).await;
let TestCodex { codex, .. } = test_codex().with_model("gpt-5.1").build(&server).await?;
codex
@@ -746,9 +734,7 @@ async fn includes_default_verbosity_in_request() -> anyhow::Result<()> {
skip_if_no_network!(Ok(()));
let server = MockServer::start().await;
let resp_mock =
responses::mount_sse_once_match(&server, path("/v1/responses"), sse_completed("resp1"))
.await;
let resp_mock = responses::mount_sse_once(&server, sse_completed("resp1")).await;
let TestCodex { codex, .. } = test_codex().with_model("gpt-5.1").build(&server).await?;
codex
@@ -781,11 +767,9 @@ async fn configured_verbosity_not_sent_for_models_without_support() -> anyhow::R
skip_if_no_network!(Ok(()));
let server = MockServer::start().await;
let resp_mock =
responses::mount_sse_once_match(&server, path("/v1/responses"), sse_completed("resp1"))
.await;
let resp_mock = responses::mount_sse_once(&server, sse_completed("resp1")).await;
let TestCodex { codex, .. } = test_codex()
.with_model("gpt-5-codex")
.with_model("gpt-5.1-codex")
.with_config(|config| {
config.model_verbosity = Some(Verbosity::High);
})
@@ -821,11 +805,9 @@ async fn configured_verbosity_is_sent() -> anyhow::Result<()> {
skip_if_no_network!(Ok(()));
let server = MockServer::start().await;
let resp_mock =
responses::mount_sse_once_match(&server, path("/v1/responses"), sse_completed("resp1"))
.await;
let resp_mock = responses::mount_sse_once(&server, sse_completed("resp1")).await;
let TestCodex { codex, .. } = test_codex()
.with_model("gpt-5")
.with_model("gpt-5.1")
.with_config(|config| {
config.model_verbosity = Some(Verbosity::High);
})
@@ -862,9 +844,7 @@ async fn includes_developer_instructions_message_in_request() {
skip_if_no_network!();
let server = MockServer::start().await;
let resp_mock =
responses::mount_sse_once_match(&server, path("/v1/responses"), sse_completed("resp1"))
.await;
let resp_mock = responses::mount_sse_once(&server, sse_completed("resp1")).await;
let model_provider = ModelProviderInfo {
base_url: Some(format!("{}/v1", server.uri())),
@@ -1175,7 +1155,7 @@ async fn token_count_includes_rate_limits_snapshot() {
"reasoning_output_tokens": 0,
"total_tokens": 123
},
// Default model is gpt-5-codex in tests → 95% usable context window
// Default model is gpt-5.1-codex in tests → 95% usable context window
"model_context_window": 258400
},
"rate_limits": {
@@ -1324,8 +1304,9 @@ async fn context_window_error_sets_total_tokens_to_model_window() -> anyhow::Res
let TestCodex { codex, .. } = test_codex()
.with_config(|config| {
config.model = "gpt-5".to_string();
config.model_family = find_family_for_model("gpt-5").expect("known gpt-5 model family");
config.model = "gpt-5.1".to_string();
config.model_family =
find_family_for_model("gpt-5.1").expect("known gpt-5.1 model family");
config.model_context_window = Some(272_000);
})
.build(&server)
codex-rs/core/tests/suite/compact.rs
+641 -156
View File
File diff suppressed because it is too large Load Diff
codex-rs/core/tests/suite/compact_resume_fork.rs
+27 -15
View File
@@ -10,15 +10,14 @@
use super::compact::COMPACT_WARNING_MESSAGE;
use super::compact::FIRST_REPLY;
use super::compact::SUMMARY_TEXT;
use super::compact::TEST_COMPACT_PROMPT;
use codex_core::CodexAuth;
use codex_core::CodexConversation;
use codex_core::ConversationManager;
use codex_core::ModelProviderInfo;
use codex_core::NewConversation;
use codex_core::built_in_model_providers;
use codex_core::compact::SUMMARIZATION_PROMPT;
use codex_core::config::Config;
use codex_core::config::OPENAI_DEFAULT_MODEL;
use codex_core::protocol::EventMsg;
use codex_core::protocol::Op;
use codex_core::protocol::WarningEvent;
@@ -38,13 +37,22 @@ use tempfile::TempDir;
use wiremock::MockServer;
const AFTER_SECOND_RESUME: &str = "AFTER_SECOND_RESUME";
const COMPACT_PROMPT_MARKER: &str =
"You are performing a CONTEXT CHECKPOINT COMPACTION for a tool.";
fn network_disabled() -> bool {
std::env::var(CODEX_SANDBOX_NETWORK_DISABLED_ENV_VAR).is_ok()
}
fn body_contains_text(body: &str, text: &str) -> bool {
body.contains(&json_fragment(text))
}
fn json_fragment(text: &str) -> String {
serde_json::to_string(text)
.expect("serialize text to JSON")
.trim_matches('"')
.to_string()
}
fn filter_out_ghost_snapshot_entries(items: &[Value]) -> Vec<Value> {
items
.iter()
@@ -82,7 +90,8 @@ fn extract_summary_message(request: &Value, summary_text: &str) -> Value {
.and_then(|arr| arr.first())
.and_then(|entry| entry.get("text"))
.and_then(Value::as_str)
== Some(summary_text)
.map(|text| text.contains(summary_text))
.unwrap_or(false)
})
})
.cloned()
@@ -101,9 +110,10 @@ async fn compact_resume_and_fork_preserve_model_history_view() {
// 1. Arrange mocked SSE responses for the initial compact/resume/fork flow.
let server = MockServer::start().await;
mount_initial_flow(&server).await;
let expected_model = "gpt-5.1-codex";
// 2. Start a new conversation and drive it through the compact/resume/fork steps.
let (_home, config, manager, base) = start_test_conversation(&server).await;
let (_home, config, manager, base) =
start_test_conversation(&server, Some(expected_model)).await;
user_turn(&base, "hello world").await;
compact_conversation(&base).await;
@@ -179,7 +189,6 @@ async fn compact_resume_and_fork_preserve_model_history_view() {
.as_str()
.unwrap_or_default()
.to_string();
let expected_model = OPENAI_DEFAULT_MODEL;
let summary_after_compact = extract_summary_message(&requests[2], SUMMARY_TEXT);
let summary_after_resume = extract_summary_message(&requests[3], SUMMARY_TEXT);
let summary_after_fork = extract_summary_message(&requests[4], SUMMARY_TEXT);
@@ -283,7 +292,7 @@ async fn compact_resume_and_fork_preserve_model_history_view() {
"content": [
{
"type": "input_text",
"text": TEST_COMPACT_PROMPT
"text": SUMMARIZATION_PROMPT
}
]
}
@@ -548,7 +557,7 @@ async fn compact_resume_after_second_compaction_preserves_history() {
mount_second_compact_flow(&server).await;
// 2. Drive the conversation through compact -> resume -> fork -> compact -> resume.
let (_home, config, manager, base) = start_test_conversation(&server).await;
let (_home, config, manager, base) = start_test_conversation(&server, None).await;
user_turn(&base, "hello world").await;
compact_conversation(&base).await;
@@ -741,7 +750,7 @@ async fn mount_initial_flow(server: &MockServer) {
let match_first = |req: &wiremock::Request| {
let body = std::str::from_utf8(&req.body).unwrap_or("");
body.contains("\"text\":\"hello world\"")
&& !body.contains(COMPACT_PROMPT_MARKER)
&& !body_contains_text(body, SUMMARIZATION_PROMPT)
&& !body.contains(&format!("\"text\":\"{SUMMARY_TEXT}\""))
&& !body.contains("\"text\":\"AFTER_COMPACT\"")
&& !body.contains("\"text\":\"AFTER_RESUME\"")
@@ -751,7 +760,7 @@ async fn mount_initial_flow(server: &MockServer) {
let match_compact = |req: &wiremock::Request| {
let body = std::str::from_utf8(&req.body).unwrap_or("");
body.contains(COMPACT_PROMPT_MARKER)
body_contains_text(body, SUMMARIZATION_PROMPT)
};
mount_sse_once_match(server, match_compact, sse2).await;
@@ -785,7 +794,7 @@ async fn mount_second_compact_flow(server: &MockServer) {
let match_second_compact = |req: &wiremock::Request| {
let body = std::str::from_utf8(&req.body).unwrap_or("");
body.contains(COMPACT_PROMPT_MARKER) && body.contains("AFTER_FORK")
body_contains_text(body, SUMMARIZATION_PROMPT) && body.contains("AFTER_FORK")
};
mount_sse_once_match(server, match_second_compact, sse6).await;
@@ -798,6 +807,7 @@ async fn mount_second_compact_flow(server: &MockServer) {
async fn start_test_conversation(
server: &MockServer,
model: Option<&str>,
) -> (TempDir, Config, ConversationManager, Arc<CodexConversation>) {
let model_provider = ModelProviderInfo {
base_url: Some(format!("{}/v1", server.uri())),
@@ -806,8 +816,10 @@ async fn start_test_conversation(
let home = TempDir::new().expect("create temp dir");
let mut config = load_default_config_for_test(&home);
config.model_provider = model_provider;
config.compact_prompt = Some(TEST_COMPACT_PROMPT.to_string());
config.compact_prompt = Some(SUMMARIZATION_PROMPT.to_string());
if let Some(model) = model {
config.model = model.to_string();
}
let manager = ConversationManager::with_auth(CodexAuth::from_api_key("dummy"));
let NewConversation { conversation, .. } = manager
.new_conversation(config.clone())
codex-rs/core/tests/suite/grep_files.rs
+22 -109
View File
@@ -2,30 +2,16 @@
use anyhow::Result;
use codex_core::model_family::find_family_for_model;
use codex_core::protocol::AskForApproval;
use codex_core::protocol::EventMsg;
use codex_core::protocol::Op;
use codex_core::protocol::SandboxPolicy;
use codex_protocol::config_types::ReasoningSummary;
use codex_protocol::user_input::UserInput;
use core_test_support::responses;
use core_test_support::responses::ev_assistant_message;
use core_test_support::responses::ev_completed;
use core_test_support::responses::ev_function_call;
use core_test_support::responses::ev_response_created;
use core_test_support::responses::sse;
use core_test_support::responses::mount_function_call_agent_response;
use core_test_support::responses::start_mock_server;
use core_test_support::skip_if_no_network;
use core_test_support::test_codex::TestCodex;
use core_test_support::test_codex::test_codex;
use core_test_support::wait_for_event;
use serde_json::Value;
use std::collections::HashSet;
use std::path::Path;
use std::process::Command as StdCommand;
use wiremock::matchers::any;
const MODEL_WITH_TOOL: &str = "test-gpt-5-codex";
const MODEL_WITH_TOOL: &str = "test-gpt-5.1-codex";
fn ripgrep_available() -> bool {
StdCommand::new("rg")
@@ -69,18 +55,22 @@ async fn grep_files_tool_collects_matches() -> Result<()> {
})
.to_string();
mount_tool_sequence(&server, call_id, &arguments, "grep_files").await;
submit_turn(&test, "please find uses of needle").await?;
let mocks =
mount_function_call_agent_response(&server, call_id, &arguments, "grep_files").await;
test.submit_turn("please find uses of needle").await?;
let bodies = recorded_bodies(&server).await?;
let tool_output = find_tool_output(&bodies, call_id).expect("tool output present");
let payload = tool_output.get("output").expect("output field present");
let (content_opt, success_opt) = extract_content_and_success(payload);
let req = mocks.completion.single_request();
let (content_opt, success_opt) = req
.function_call_output_content_and_success(call_id)
.expect("tool output present");
let content = content_opt.expect("content present");
let success = success_opt.unwrap_or(true);
assert!(success, "expected success for matches, got {payload:?}");
assert!(
success,
"expected success for matches, got content={content}"
);
let entries = collect_file_names(content);
let entries = collect_file_names(&content);
assert_eq!(entries.len(), 2, "content: {content}");
assert!(
entries.contains("alpha.rs"),
@@ -118,16 +108,17 @@ async fn grep_files_tool_reports_empty_results() -> Result<()> {
})
.to_string();
mount_tool_sequence(&server, call_id, &arguments, "grep_files").await;
submit_turn(&test, "search again").await?;
let mocks =
mount_function_call_agent_response(&server, call_id, &arguments, "grep_files").await;
test.submit_turn("search again").await?;
let bodies = recorded_bodies(&server).await?;
let tool_output = find_tool_output(&bodies, call_id).expect("tool output present");
let payload = tool_output.get("output").expect("output field present");
let (content_opt, success_opt) = extract_content_and_success(payload);
let req = mocks.completion.single_request();
let (content_opt, success_opt) = req
.function_call_output_content_and_success(call_id)
.expect("tool output present");
let content = content_opt.expect("content present");
if let Some(success) = success_opt {
assert!(!success, "expected success=false payload: {payload:?}");
assert!(!success, "expected success=false content={content}");
}
assert_eq!(content, "No matches found.");
@@ -144,73 +135,6 @@ async fn build_test_codex(server: &wiremock::MockServer) -> Result<TestCodex> {
builder.build(server).await
}
async fn submit_turn(test: &TestCodex, prompt: &str) -> Result<()> {
let session_model = test.session_configured.model.clone();
test.codex
.submit(Op::UserTurn {
items: vec![UserInput::Text {
text: prompt.into(),
}],
final_output_json_schema: None,
cwd: test.cwd.path().to_path_buf(),
approval_policy: AskForApproval::Never,
sandbox_policy: SandboxPolicy::DangerFullAccess,
model: session_model,
effort: None,
summary: ReasoningSummary::Auto,
})
.await?;
wait_for_event(&test.codex, |event| {
matches!(event, EventMsg::TaskComplete(_))
})
.await;
Ok(())
}
async fn mount_tool_sequence(
server: &wiremock::MockServer,
call_id: &str,
arguments: &str,
tool_name: &str,
) {
let first_response = sse(vec![
ev_response_created("resp-1"),
ev_function_call(call_id, tool_name, arguments),
ev_completed("resp-1"),
]);
responses::mount_sse_once_match(server, any(), first_response).await;
let second_response = sse(vec![
ev_assistant_message("msg-1", "done"),
ev_completed("resp-2"),
]);
responses::mount_sse_once_match(server, any(), second_response).await;
}
#[allow(clippy::expect_used)]
async fn recorded_bodies(server: &wiremock::MockServer) -> Result<Vec<Value>> {
let requests = server.received_requests().await.expect("requests recorded");
Ok(requests
.iter()
.map(|req| req.body_json::<Value>().expect("request json"))
.collect())
}
fn find_tool_output<'a>(requests: &'a [Value], call_id: &str) -> Option<&'a Value> {
requests.iter().find_map(|body| {
body.get("input")
.and_then(Value::as_array)
.and_then(|items| {
items.iter().find(|item| {
item.get("type").and_then(Value::as_str) == Some("function_call_output")
&& item.get("call_id").and_then(Value::as_str) == Some(call_id)
})
})
})
}
fn collect_file_names(content: &str) -> HashSet<String> {
content
.lines()
@@ -224,14 +148,3 @@ fn collect_file_names(content: &str) -> HashSet<String> {
})
.collect()
}
fn extract_content_and_success(value: &Value) -> (Option<&str>, Option<bool>) {
match value {
Value::String(text) => (Some(text.as_str()), None),
Value::Object(obj) => (
obj.get("content").and_then(Value::as_str),
obj.get("success").and_then(Value::as_bool),
),
_ => (None, None),
}
}
codex-rs/core/tests/suite/items.rs
+8 -9
View File
@@ -18,7 +18,7 @@ use core_test_support::responses::ev_reasoning_text_delta;
use core_test_support::responses::ev_response_created;
use core_test_support::responses::ev_web_search_call_added;
use core_test_support::responses::ev_web_search_call_done;
use core_test_support::responses::mount_sse_once_match;
use core_test_support::responses::mount_sse_once;
use core_test_support::responses::sse;
use core_test_support::responses::start_mock_server;
use core_test_support::skip_if_no_network;
@@ -26,7 +26,6 @@ use core_test_support::test_codex::TestCodex;
use core_test_support::test_codex::test_codex;
use core_test_support::wait_for_event_match;
use pretty_assertions::assert_eq;
use wiremock::matchers::any;
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn user_message_item_is_emitted() -> anyhow::Result<()> {
@@ -37,7 +36,7 @@ async fn user_message_item_is_emitted() -> anyhow::Result<()> {
let TestCodex { codex, .. } = test_codex().build(&server).await?;
let first_response = sse(vec![ev_response_created("resp-1"), ev_completed("resp-1")]);
mount_sse_once_match(&server, any(), first_response).await;
mount_sse_once(&server, first_response).await;
codex
.submit(Op::UserInput {
@@ -93,7 +92,7 @@ async fn assistant_message_item_is_emitted() -> anyhow::Result<()> {
ev_assistant_message("msg-1", "all done"),
ev_completed("resp-1"),
]);
mount_sse_once_match(&server, any(), first_response).await;
mount_sse_once(&server, first_response).await;
codex
.submit(Op::UserInput {
@@ -149,7 +148,7 @@ async fn reasoning_item_is_emitted() -> anyhow::Result<()> {
reasoning_item,
ev_completed("resp-1"),
]);
mount_sse_once_match(&server, any(), first_response).await;
mount_sse_once(&server, first_response).await;
codex
.submit(Op::UserInput {
@@ -207,7 +206,7 @@ async fn web_search_item_is_emitted() -> anyhow::Result<()> {
web_search_done,
ev_completed("resp-1"),
]);
mount_sse_once_match(&server, any(), first_response).await;
mount_sse_once(&server, first_response).await;
codex
.submit(Op::UserInput {
@@ -259,7 +258,7 @@ async fn agent_message_content_delta_has_item_metadata() -> anyhow::Result<()> {
ev_assistant_message("msg-1", "streamed response"),
ev_completed("resp-1"),
]);
mount_sse_once_match(&server, any(), stream).await;
mount_sse_once(&server, stream).await;
codex
.submit(Op::UserInput {
@@ -324,7 +323,7 @@ async fn reasoning_content_delta_has_item_metadata() -> anyhow::Result<()> {
ev_reasoning_item("reasoning-1", &["step one"], &[]),
ev_completed("resp-1"),
]);
mount_sse_once_match(&server, any(), stream).await;
mount_sse_once(&server, stream).await;
codex
.submit(Op::UserInput {
@@ -381,7 +380,7 @@ async fn reasoning_raw_content_delta_respects_flag() -> anyhow::Result<()> {
ev_reasoning_item("reasoning-raw", &["complete"], &["raw detail"]),
ev_completed("resp-1"),
]);
mount_sse_once_match(&server, any(), stream).await;
mount_sse_once(&server, stream).await;
codex
.submit(Op::UserInput {
codex-rs/core/tests/suite/json_result.rs
+2 -2
View File
@@ -31,12 +31,12 @@ const SCHEMA: &str = r#"
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn codex_returns_json_result_for_gpt5() -> anyhow::Result<()> {
codex_returns_json_result("gpt-5".to_string()).await
codex_returns_json_result("gpt-5.1".to_string()).await
}
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
async fn codex_returns_json_result_for_gpt5_codex() -> anyhow::Result<()> {
codex_returns_json_result("gpt-5-codex".to_string()).await
codex_returns_json_result("gpt-5.1-codex".to_string()).await
}
async fn codex_returns_json_result(model: String) -> anyhow::Result<()> {
codex-rs/core/tests/suite/list_dir.rs
+46 -339
View File
@@ -1,25 +1,11 @@
#![cfg(not(target_os = "windows"))]
use codex_core::protocol::AskForApproval;
use codex_core::protocol::EventMsg;
use codex_core::protocol::Op;
use codex_core::protocol::SandboxPolicy;
use codex_protocol::config_types::ReasoningSummary;
use codex_protocol::user_input::UserInput;
use core_test_support::responses;
use core_test_support::responses::ev_assistant_message;
use core_test_support::responses::ev_completed;
use core_test_support::responses::ev_function_call;
use core_test_support::responses::ev_response_created;
use core_test_support::responses::sse;
use core_test_support::responses::mount_function_call_agent_response;
use core_test_support::responses::start_mock_server;
use core_test_support::skip_if_no_network;
use core_test_support::test_codex::TestCodex;
use core_test_support::test_codex::test_codex;
use core_test_support::wait_for_event;
use pretty_assertions::assert_eq;
use serde_json::Value;
use wiremock::matchers::any;
use serde_json::json;
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
#[ignore = "disabled until we enable list_dir tool"]
@@ -27,99 +13,30 @@ async fn list_dir_tool_returns_entries() -> anyhow::Result<()> {
skip_if_no_network!(Ok(()));
let server = start_mock_server().await;
let test = test_codex().build(&server).await?;
let TestCodex {
codex,
cwd,
session_configured,
..
} = test_codex().build(&server).await?;
let dir_path = cwd.path().join("sample_dir");
let dir_path = test.cwd.path().join("sample_dir");
std::fs::create_dir(&dir_path)?;
std::fs::write(dir_path.join("alpha.txt"), "first file")?;
std::fs::create_dir(dir_path.join("nested"))?;
let dir_path = dir_path.to_string_lossy().to_string();
let call_id = "list-dir-call";
let arguments = serde_json::json!({
let arguments = json!({
"dir_path": dir_path,
"offset": 1,
"limit": 2,
})
.to_string();
let first_response = sse(vec![
ev_response_created("resp-1"),
ev_function_call(call_id, "list_dir", &arguments),
ev_completed("resp-1"),
]);
responses::mount_sse_once_match(&server, any(), first_response).await;
let second_response = sse(vec![
ev_assistant_message("msg-1", "done"),
ev_completed("resp-2"),
]);
responses::mount_sse_once_match(&server, any(), second_response).await;
let session_model = session_configured.model.clone();
codex
.submit(Op::UserTurn {
items: vec![UserInput::Text {
text: "list directory contents".into(),
}],
final_output_json_schema: None,
cwd: cwd.path().to_path_buf(),
approval_policy: AskForApproval::Never,
sandbox_policy: SandboxPolicy::DangerFullAccess,
model: session_model,
effort: None,
summary: ReasoningSummary::Auto,
})
.await?;
wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
let requests = server.received_requests().await.expect("recorded requests");
let request_bodies = requests
.iter()
.map(|req| req.body_json::<Value>().unwrap())
.collect::<Vec<_>>();
assert!(
!request_bodies.is_empty(),
"expected at least one request body"
);
let tool_output_item = request_bodies
.iter()
.find_map(|body| {
body.get("input")
.and_then(Value::as_array)
.and_then(|items| {
items.iter().find(|item| {
item.get("type").and_then(Value::as_str) == Some("function_call_output")
})
})
})
.unwrap_or_else(|| {
panic!("function_call_output item not found in requests: {request_bodies:#?}")
});
assert_eq!(
tool_output_item.get("call_id").and_then(Value::as_str),
Some(call_id)
);
let output_text = tool_output_item
.get("output")
.and_then(|value| match value {
Value::String(text) => Some(text.as_str()),
Value::Object(obj) => obj.get("content").and_then(Value::as_str),
_ => None,
})
.expect("output text present");
assert_eq!(output_text, "E1: [file] alpha.txt\nE2: [dir] nested");
let mocks = mount_function_call_agent_response(&server, call_id, &arguments, "list_dir").await;
test.submit_turn("list directory contents").await?;
let req = mocks.completion.single_request();
let (content_opt, _) = req
.function_call_output_content_and_success(call_id)
.expect("function_call_output present");
let output = content_opt.expect("output content present in tool output");
assert_eq!(output, "E1: [file] alpha.txt\nE2: [dir] nested");
Ok(())
}
@@ -130,15 +47,9 @@ async fn list_dir_tool_depth_one_omits_children() -> anyhow::Result<()> {
skip_if_no_network!(Ok(()));
let server = start_mock_server().await;
let test = test_codex().build(&server).await?;
let TestCodex {
codex,
cwd,
session_configured,
..
} = test_codex().build(&server).await?;
let dir_path = cwd.path().join("depth_one");
let dir_path = test.cwd.path().join("depth_one");
std::fs::create_dir(&dir_path)?;
std::fs::write(dir_path.join("alpha.txt"), "alpha")?;
std::fs::create_dir(dir_path.join("nested"))?;
@@ -146,7 +57,7 @@ async fn list_dir_tool_depth_one_omits_children() -> anyhow::Result<()> {
let dir_path = dir_path.to_string_lossy().to_string();
let call_id = "list-dir-depth1";
let arguments = serde_json::json!({
let arguments = json!({
"dir_path": dir_path,
"offset": 1,
"limit": 10,
@@ -154,77 +65,15 @@ async fn list_dir_tool_depth_one_omits_children() -> anyhow::Result<()> {
})
.to_string();
let first_response = sse(vec![
ev_response_created("resp-1"),
ev_function_call(call_id, "list_dir", &arguments),
ev_completed("resp-1"),
]);
responses::mount_sse_once_match(&server, any(), first_response).await;
let second_response = sse(vec![
ev_assistant_message("msg-1", "done"),
ev_completed("resp-2"),
]);
responses::mount_sse_once_match(&server, any(), second_response).await;
let session_model = session_configured.model.clone();
codex
.submit(Op::UserTurn {
items: vec![UserInput::Text {
text: "list directory contents depth one".into(),
}],
final_output_json_schema: None,
cwd: cwd.path().to_path_buf(),
approval_policy: AskForApproval::Never,
sandbox_policy: SandboxPolicy::DangerFullAccess,
model: session_model,
effort: None,
summary: ReasoningSummary::Auto,
})
let mocks = mount_function_call_agent_response(&server, call_id, &arguments, "list_dir").await;
test.submit_turn("list directory contents depth one")
.await?;
wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
let requests = server.received_requests().await.expect("recorded requests");
let request_bodies = requests
.iter()
.map(|req| req.body_json::<Value>().unwrap())
.collect::<Vec<_>>();
assert!(
!request_bodies.is_empty(),
"expected at least one request body"
);
let tool_output_item = request_bodies
.iter()
.find_map(|body| {
body.get("input")
.and_then(Value::as_array)
.and_then(|items| {
items.iter().find(|item| {
item.get("type").and_then(Value::as_str) == Some("function_call_output")
})
})
})
.unwrap_or_else(|| {
panic!("function_call_output item not found in requests: {request_bodies:#?}")
});
assert_eq!(
tool_output_item.get("call_id").and_then(Value::as_str),
Some(call_id)
);
let output_text = tool_output_item
.get("output")
.and_then(|value| match value {
Value::String(text) => Some(text.as_str()),
Value::Object(obj) => obj.get("content").and_then(Value::as_str),
_ => None,
})
.expect("output text present");
assert_eq!(output_text, "E1: [file] alpha.txt\nE2: [dir] nested");
let req = mocks.completion.single_request();
let (content_opt, _) = req
.function_call_output_content_and_success(call_id)
.expect("function_call_output present");
let output = content_opt.expect("output content present in tool output");
assert_eq!(output, "E1: [file] alpha.txt\nE2: [dir] nested");
Ok(())
}
@@ -235,15 +84,9 @@ async fn list_dir_tool_depth_two_includes_children_only() -> anyhow::Result<()>
skip_if_no_network!(Ok(()));
let server = start_mock_server().await;
let test = test_codex().build(&server).await?;
let TestCodex {
codex,
cwd,
session_configured,
..
} = test_codex().build(&server).await?;
let dir_path = cwd.path().join("depth_two");
let dir_path = test.cwd.path().join("depth_two");
std::fs::create_dir(&dir_path)?;
std::fs::write(dir_path.join("alpha.txt"), "alpha")?;
let nested = dir_path.join("nested");
@@ -255,7 +98,7 @@ async fn list_dir_tool_depth_two_includes_children_only() -> anyhow::Result<()>
let dir_path_string = dir_path.to_string_lossy().to_string();
let call_id = "list-dir-depth2";
let arguments = serde_json::json!({
let arguments = json!({
"dir_path": dir_path_string,
"offset": 1,
"limit": 10,
@@ -263,81 +106,16 @@ async fn list_dir_tool_depth_two_includes_children_only() -> anyhow::Result<()>
})
.to_string();
let first_response = sse(vec![
serde_json::json!({
"type": "response.created",
"response": {"id": "resp-1"}
}),
ev_function_call(call_id, "list_dir", &arguments),
ev_completed("resp-1"),
]);
responses::mount_sse_once_match(&server, any(), first_response).await;
let second_response = sse(vec![
ev_assistant_message("msg-1", "done"),
ev_completed("resp-2"),
]);
responses::mount_sse_once_match(&server, any(), second_response).await;
let session_model = session_configured.model.clone();
codex
.submit(Op::UserTurn {
items: vec![UserInput::Text {
text: "list directory contents depth two".into(),
}],
final_output_json_schema: None,
cwd: cwd.path().to_path_buf(),
approval_policy: AskForApproval::Never,
sandbox_policy: SandboxPolicy::DangerFullAccess,
model: session_model,
effort: None,
summary: ReasoningSummary::Auto,
})
let mocks = mount_function_call_agent_response(&server, call_id, &arguments, "list_dir").await;
test.submit_turn("list directory contents depth two")
.await?;
wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
let requests = server.received_requests().await.expect("recorded requests");
let request_bodies = requests
.iter()
.map(|req| req.body_json::<Value>().unwrap())
.collect::<Vec<_>>();
assert!(
!request_bodies.is_empty(),
"expected at least one request body"
);
let tool_output_item = request_bodies
.iter()
.find_map(|body| {
body.get("input")
.and_then(Value::as_array)
.and_then(|items| {
items.iter().find(|item| {
item.get("type").and_then(Value::as_str) == Some("function_call_output")
})
})
})
.unwrap_or_else(|| {
panic!("function_call_output item not found in requests: {request_bodies:#?}")
});
let req = mocks.completion.single_request();
let (content_opt, _) = req
.function_call_output_content_and_success(call_id)
.expect("function_call_output present");
let output = content_opt.expect("output content present in tool output");
assert_eq!(
tool_output_item.get("call_id").and_then(Value::as_str),
Some(call_id)
);
let output_text = tool_output_item
.get("output")
.and_then(|value| match value {
Value::String(text) => Some(text.as_str()),
Value::Object(obj) => obj.get("content").and_then(Value::as_str),
_ => None,
})
.expect("output text present");
assert_eq!(
output_text,
output,
"E1: [file] alpha.txt\nE2: [dir] nested\nE3: [file] nested/beta.txt\nE4: [dir] nested/grand"
);
@@ -350,15 +128,9 @@ async fn list_dir_tool_depth_three_includes_grandchildren() -> anyhow::Result<()
skip_if_no_network!(Ok(()));
let server = start_mock_server().await;
let test = test_codex().build(&server).await?;
let TestCodex {
codex,
cwd,
session_configured,
..
} = test_codex().build(&server).await?;
let dir_path = cwd.path().join("depth_three");
let dir_path = test.cwd.path().join("depth_three");
std::fs::create_dir(&dir_path)?;
std::fs::write(dir_path.join("alpha.txt"), "alpha")?;
let nested = dir_path.join("nested");
@@ -370,7 +142,7 @@ async fn list_dir_tool_depth_three_includes_grandchildren() -> anyhow::Result<()
let dir_path_string = dir_path.to_string_lossy().to_string();
let call_id = "list-dir-depth3";
let arguments = serde_json::json!({
let arguments = json!({
"dir_path": dir_path_string,
"offset": 1,
"limit": 10,
@@ -378,81 +150,16 @@ async fn list_dir_tool_depth_three_includes_grandchildren() -> anyhow::Result<()
})
.to_string();
let first_response = sse(vec![
serde_json::json!({
"type": "response.created",
"response": {"id": "resp-1"}
}),
ev_function_call(call_id, "list_dir", &arguments),
ev_completed("resp-1"),
]);
responses::mount_sse_once_match(&server, any(), first_response).await;
let second_response = sse(vec![
ev_assistant_message("msg-1", "done"),
ev_completed("resp-2"),
]);
responses::mount_sse_once_match(&server, any(), second_response).await;
let session_model = session_configured.model.clone();
codex
.submit(Op::UserTurn {
items: vec![UserInput::Text {
text: "list directory contents depth three".into(),
}],
final_output_json_schema: None,
cwd: cwd.path().to_path_buf(),
approval_policy: AskForApproval::Never,
sandbox_policy: SandboxPolicy::DangerFullAccess,
model: session_model,
effort: None,
summary: ReasoningSummary::Auto,
})
let mocks = mount_function_call_agent_response(&server, call_id, &arguments, "list_dir").await;
test.submit_turn("list directory contents depth three")
.await?;
wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
let requests = server.received_requests().await.expect("recorded requests");
let request_bodies = requests
.iter()
.map(|req| req.body_json::<Value>().unwrap())
.collect::<Vec<_>>();
assert!(
!request_bodies.is_empty(),
"expected at least one request body"
);
let tool_output_item = request_bodies
.iter()
.find_map(|body| {
body.get("input")
.and_then(Value::as_array)
.and_then(|items| {
items.iter().find(|item| {
item.get("type").and_then(Value::as_str) == Some("function_call_output")
})
})
})
.unwrap_or_else(|| {
panic!("function_call_output item not found in requests: {request_bodies:#?}")
});
let req = mocks.completion.single_request();
let (content_opt, _) = req
.function_call_output_content_and_success(call_id)
.expect("function_call_output present");
let output = content_opt.expect("output content present in tool output");
assert_eq!(
tool_output_item.get("call_id").and_then(Value::as_str),
Some(call_id)
);
let output_text = tool_output_item
.get("output")
.and_then(|value| match value {
Value::String(text) => Some(text.as_str()),
Value::Object(obj) => obj.get("content").and_then(Value::as_str),
_ => None,
})
.expect("output text present");
assert_eq!(
output_text,
output,
"E1: [file] alpha.txt\nE2: [dir] nested\nE3: [file] nested/beta.txt\nE4: [dir] nested/grand\nE5: [file] nested/grand/gamma.txt"
);
codex-rs/core/tests/suite/mod.rs
-2
View File
@@ -18,8 +18,6 @@ mod abort_tasks;
#[cfg(not(target_os = "windows"))]
mod apply_patch_cli;
#[cfg(not(target_os = "windows"))]
mod apply_patch_freeform;
#[cfg(not(target_os = "windows"))]
mod approvals;
mod auth_refresh;
mod cli_stream;
codex-rs/core/tests/suite/model_tools.rs
+13 -3
View File
@@ -42,7 +42,7 @@ async fn collect_tool_identifiers_for_model(model: &str) -> Vec<String> {
let server = MockServer::start().await;
let sse = sse_completed(model);
let resp_mock = responses::mount_sse_once_match(&server, wiremock::matchers::any(), sse).await;
let resp_mock = responses::mount_sse_once(&server, sse).await;
let model_provider = ModelProviderInfo {
base_url: Some(format!("{}/v1", server.uri())),
@@ -119,7 +119,12 @@ async fn model_selects_expected_tools() {
assert_eq!(
gpt5_codex_tools,
vec![
"shell".to_string(),
if cfg!(windows) {
"shell_command"
} else {
"shell"
}
.to_string(),
"list_mcp_resources".to_string(),
"list_mcp_resource_templates".to_string(),
"read_mcp_resource".to_string(),
@@ -133,7 +138,12 @@ async fn model_selects_expected_tools() {
assert_eq!(
gpt51_codex_tools,
vec![
"shell".to_string(),
if cfg!(windows) {
"shell_command"
} else {
"shell"
}
.to_string(),
"list_mcp_resources".to_string(),
"list_mcp_resource_templates".to_string(),
"read_mcp_resource".to_string(),

Some files were not shown because too many files have changed in this diff Show More