openai/codex
1
0
Fork 0
mirror of https://github.com/openai/codex.git synced 2026-10-30 19:18:49 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'main' into dev/mzeng/hot_reload_config

Browse Source
This commit is contained in:
Matthew Zeng
2026-02-27 09:38:57 -08:00
committed by GitHub
parent 13a0d98633 8cf5b00aef
commit cea21b18e2
259 changed files with 14999 additions and 6503 deletions
Download Patch File Download Diff File Expand all files Collapse all files
codex-rs/.cargo/config.toml
+6
View File
@@ -1,5 +1,11 @@
[target.'cfg(all(windows, target_env = "msvc"))']
rustflags = ["-C", "link-arg=/STACK:8388608"]
# MSVC emits a warning about code that may trip "Cortex-A53 MPCore processor bug #843419" (see
# https://developer.arm.com/documentation/epm048406/latest) which is sometimes emitted by LLVM.
# Since Arm64 Windows 10+ isn't supported on that processor, it's safe to disable the warning.
[target.aarch64-pc-windows-msvc]
rustflags = ["-C", "link-arg=/STACK:8388608", "-C", "link-arg=/arm64hazardfree"]
[target.'cfg(all(windows, target_env = "gnu"))']
rustflags = ["-C", "link-arg=-Wl,--stack,8388608"]
codex-rs/Cargo.lock
Generated
+3
View File
@@ -1791,6 +1791,7 @@ dependencies = [
"eventsource-stream",
"futures",
"http 1.4.0",
"iana-time-zone",
"image",
"indexmap 2.13.0",
"insta",
@@ -2299,6 +2300,7 @@ dependencies = [
"anyhow",
"async-trait",
"clap",
"codex-protocol",
"codex-utils-absolute-path",
"libc",
"pretty_assertions",
@@ -2399,6 +2401,7 @@ dependencies = [
"codex-utils-pty",
"codex-utils-sandbox-summary",
"codex-utils-sleep-inhibitor",
"codex-utils-string",
"codex-windows-sandbox",
"color-eyre",
"cpal",
codex-rs/Cargo.toml
+7 -5
View File
@@ -118,8 +118,8 @@ codex-shell-command = { path = "shell-command" }
codex-shell-escalation = { path = "shell-escalation" }
codex-skills = { path = "skills" }
codex-state = { path = "state" }
codex-test-macros = { path = "test-macros" }
codex-stdio-to-uds = { path = "stdio-to-uds" }
codex-test-macros = { path = "test-macros" }
codex-tui = { path = "tui" }
codex-utils-absolute-path = { path = "utils/absolute-path" }
codex-utils-approval-presets = { path = "utils/approval-presets" }
@@ -137,8 +137,8 @@ codex-utils-readiness = { path = "utils/readiness" }
codex-utils-rustls-provider = { path = "utils/rustls-provider" }
codex-utils-sandbox-summary = { path = "utils/sandbox-summary" }
codex-utils-sleep-inhibitor = { path = "utils/sleep-inhibitor" }
codex-utils-string = { path = "utils/string" }
codex-utils-stream-parser = { path = "utils/stream-parser" }
codex-utils-string = { path = "utils/string" }
codex-windows-sandbox = { path = "windows-sandbox-rs" }
core_test_support = { path = "core/tests/common" }
mcp_test_support = { path = "mcp-server/tests/common" }
@@ -165,8 +165,8 @@ clap = "4"
clap_complete = "4"
color-eyre = "0.6.3"
crossbeam-channel = "0.5.15"
csv = "1.3.1"
crossterm = "0.28.1"
csv = "1.3.1"
ctor = "0.6.3"
derive_more = "2"
diffy = "0.4.2"
@@ -178,14 +178,15 @@ env-flags = "0.1.1"
env_logger = "0.11.9"
eventsource-stream = "0.2.3"
futures = { version = "0.3", default-features = false }
globset = "0.4"
gethostname = "1.1.0"
globset = "0.4"
http = "1.3.1"
icu_decimal = "2.1"
icu_locale_core = "2.1"
icu_provider = { version = "2.1", features = ["sync"] }
ignore = "0.4.23"
image = { version = "^0.25.9", default-features = false }
iana-time-zone = "0.1.64"
include_dir = "0.7.4"
indexmap = "2.12.0"
insta = "1.46.3"
@@ -258,6 +259,7 @@ starlark = "0.13.0"
strum = "0.27.2"
strum_macros = "0.27.2"
supports-color = "3.0.2"
syntect = "5"
sys-locale = "0.3.2"
tempfile = "3.23.0"
test-log = "0.2.19"
@@ -282,7 +284,6 @@ tracing-subscriber = "0.3.22"
tracing-test = "0.2.5"
tree-sitter = "0.25.10"
tree-sitter-bash = "0.25"
syntect = "5"
ts-rs = "11"
tungstenite = { version = "0.27.0", features = ["deflate", "proxy"] }
uds_windows = "1.1.0"
@@ -352,6 +353,7 @@ ignored = [
[profile.release]
lto = "fat"
split-debuginfo = "off"
# Because we bundle some of these executables with the TypeScript CLI, we
# remove everything to make the binary as small as possible.
strip = "symbols"
codex-rs/app-server-protocol/schema/json/ApplyPatchApprovalResponse.json
+1 -1
View File
@@ -59,7 +59,7 @@
"type": "object"
},
{
"description": "User has approved this command and wants to automatically approve any future identical instances (`command` and `cwd` match exactly) for the remainder of the session.",
"description": "User has approved this request and wants future prompts in the same session-scoped approval cache to be automatically approved for the remainder of the session.",
"enum": [
"approved_for_session"
],
codex-rs/app-server-protocol/schema/json/ClientRequest.json
+1 -1
View File
@@ -1340,7 +1340,7 @@
"type": "string"
},
"output": {
"type": "string"
"$ref": "#/definitions/FunctionCallOutputPayload"
},
"type": {
"enum": [
codex-rs/app-server-protocol/schema/json/CommandExecutionRequestApprovalParams.json
+79
View File
@@ -202,6 +202,85 @@
}
]
},
"CommandExecutionApprovalDecision": {
"oneOf": [
{
"description": "User approved the command.",
"enum": [
"accept"
],
"type": "string"
},
{
"description": "User approved the command and future prompts in the same session-scoped approval cache should run without prompting.",
"enum": [
"acceptForSession"
],
"type": "string"
},
{
"additionalProperties": false,
"description": "User approved the command, and wants to apply the proposed execpolicy amendment so future matching commands can run without prompting.",
"properties": {
"acceptWithExecpolicyAmendment": {
"properties": {
"execpolicy_amendment": {
"items": {
"type": "string"
},
"type": "array"
}
},
"required": [
"execpolicy_amendment"
],
"type": "object"
}
},
"required": [
"acceptWithExecpolicyAmendment"
],
"title": "AcceptWithExecpolicyAmendmentCommandExecutionApprovalDecision",
"type": "object"
},
{
"additionalProperties": false,
"description": "User chose a persistent network policy rule (allow/deny) for this host.",
"properties": {
"applyNetworkPolicyAmendment": {
"properties": {
"network_policy_amendment": {
"$ref": "#/definitions/NetworkPolicyAmendment"
}
},
"required": [
"network_policy_amendment"
],
"type": "object"
}
},
"required": [
"applyNetworkPolicyAmendment"
],
"title": "ApplyNetworkPolicyAmendmentCommandExecutionApprovalDecision",
"type": "object"
},
{
"description": "User denied the command. The agent will continue the turn.",
"enum": [
"decline"
],
"type": "string"
},
{
"description": "User denied the command. The turn will also be immediately interrupted.",
"enum": [
"cancel"
],
"type": "string"
}
]
},
"MacOsAutomationValue": {
"anyOf": [
{
codex-rs/app-server-protocol/schema/json/CommandExecutionRequestApprovalResponse.json
+1 -1
View File
@@ -11,7 +11,7 @@
"type": "string"
},
{
"description": "User approved the command and future identical commands should run without prompting.",
"description": "User approved the command and future prompts in the same session-scoped approval cache should run without prompting.",
"enum": [
"acceptForSession"
],
codex-rs/app-server-protocol/schema/json/EventMsg.json
+101 -49
View File
@@ -1675,6 +1675,16 @@
"null"
]
},
"available_decisions": {
"description": "Ordered list of decisions the client may present for this prompt.\n\nWhen absent, clients should derive the legacy default set from the other fields on this request.",
"items": {
"$ref": "#/definitions/ReviewDecision"
},
"type": [
"array",
"null"
]
},
"call_id": {
"description": "Identifier for the associated command execution item.",
"type": "string"
@@ -1884,30 +1894,6 @@
"title": "DynamicToolCallResponseEventMsg",
"type": "object"
},
{
"properties": {
"item_id": {
"type": "string"
},
"skill_name": {
"type": "string"
},
"type": {
"enum": [
"skill_request_approval"
],
"title": "SkillRequestApprovalEventMsgType",
"type": "string"
}
},
"required": [
"item_id",
"skill_name",
"type"
],
"title": "SkillRequestApprovalEventMsg",
"type": "object"
},
{
"properties": {
"id": {
@@ -4836,7 +4822,7 @@
"type": "string"
},
"output": {
"type": "string"
"$ref": "#/definitions/FunctionCallOutputPayload"
},
"type": {
"enum": [
@@ -4995,6 +4981,86 @@
],
"type": "object"
},
"ReviewDecision": {
"description": "User's decision in response to an ExecApprovalRequest.",
"oneOf": [
{
"description": "User has approved this command and the agent should execute it.",
"enum": [
"approved"
],
"type": "string"
},
{
"additionalProperties": false,
"description": "User has approved this command and wants to apply the proposed execpolicy amendment so future matching commands are permitted.",
"properties": {
"approved_execpolicy_amendment": {
"properties": {
"proposed_execpolicy_amendment": {
"items": {
"type": "string"
},
"type": "array"
}
},
"required": [
"proposed_execpolicy_amendment"
],
"type": "object"
}
},
"required": [
"approved_execpolicy_amendment"
],
"title": "ApprovedExecpolicyAmendmentReviewDecision",
"type": "object"
},
{
"description": "User has approved this request and wants future prompts in the same session-scoped approval cache to be automatically approved for the remainder of the session.",
"enum": [
"approved_for_session"
],
"type": "string"
},
{
"additionalProperties": false,
"description": "User chose to persist a network policy rule (allow/deny) for future requests to the same host.",
"properties": {
"network_policy_amendment": {
"properties": {
"network_policy_amendment": {
"$ref": "#/definitions/NetworkPolicyAmendment"
}
},
"required": [
"network_policy_amendment"
],
"type": "object"
}
},
"required": [
"network_policy_amendment"
],
"title": "NetworkPolicyAmendmentReviewDecision",
"type": "object"
},
{
"description": "User has denied this command and the agent should not execute it, but it should continue the session and try something else.",
"enum": [
"denied"
],
"type": "string"
},
{
"description": "User has denied this command and the agent should not do anything until the user's next command.",
"enum": [
"abort"
],
"type": "string"
}
]
},
"ReviewFinding": {
"description": "A single review finding describing an observed issue or recommendation.",
"properties": {
@@ -7165,6 +7231,16 @@
"null"
]
},
"available_decisions": {
"description": "Ordered list of decisions the client may present for this prompt.\n\nWhen absent, clients should derive the legacy default set from the other fields on this request.",
"items": {
"$ref": "#/definitions/ReviewDecision"
},
"type": [
"array",
"null"
]
},
"call_id": {
"description": "Identifier for the associated command execution item.",
"type": "string"
@@ -7374,30 +7450,6 @@
"title": "DynamicToolCallResponseEventMsg",
"type": "object"
},
{
"properties": {
"item_id": {
"type": "string"
},
"skill_name": {
"type": "string"
},
"type": {
"enum": [
"skill_request_approval"
],
"title": "SkillRequestApprovalEventMsgType",
"type": "string"
}
},
"required": [
"item_id",
"skill_name",
"type"
],
"title": "SkillRequestApprovalEventMsg",
"type": "object"
},
{
"properties": {
"id": {
codex-rs/app-server-protocol/schema/json/ExecCommandApprovalResponse.json
+1 -1
View File
@@ -59,7 +59,7 @@
"type": "object"
},
{
"description": "User has approved this command and wants to automatically approve any future identical instances (`command` and `cwd` match exactly) for the remainder of the session.",
"description": "User has approved this request and wants future prompts in the same session-scoped approval cache to be automatically approved for the remainder of the session.",
"enum": [
"approved_for_session"
],
codex-rs/app-server-protocol/schema/json/ServerRequest.json
+79 -39
View File
@@ -268,6 +268,85 @@
}
]
},
"CommandExecutionApprovalDecision": {
"oneOf": [
{
"description": "User approved the command.",
"enum": [
"accept"
],
"type": "string"
},
{
"description": "User approved the command and future prompts in the same session-scoped approval cache should run without prompting.",
"enum": [
"acceptForSession"
],
"type": "string"
},
{
"additionalProperties": false,
"description": "User approved the command, and wants to apply the proposed execpolicy amendment so future matching commands can run without prompting.",
"properties": {
"acceptWithExecpolicyAmendment": {
"properties": {
"execpolicy_amendment": {
"items": {
"type": "string"
},
"type": "array"
}
},
"required": [
"execpolicy_amendment"
],
"type": "object"
}
},
"required": [
"acceptWithExecpolicyAmendment"
],
"title": "AcceptWithExecpolicyAmendmentCommandExecutionApprovalDecision",
"type": "object"
},
{
"additionalProperties": false,
"description": "User chose a persistent network policy rule (allow/deny) for this host.",
"properties": {
"applyNetworkPolicyAmendment": {
"properties": {
"network_policy_amendment": {
"$ref": "#/definitions/NetworkPolicyAmendment"
}
},
"required": [
"network_policy_amendment"
],
"type": "object"
}
},
"required": [
"applyNetworkPolicyAmendment"
],
"title": "ApplyNetworkPolicyAmendmentCommandExecutionApprovalDecision",
"type": "object"
},
{
"description": "User denied the command. The agent will continue the turn.",
"enum": [
"decline"
],
"type": "string"
},
{
"description": "User denied the command. The turn will also be immediately interrupted.",
"enum": [
"cancel"
],
"type": "string"
}
]
},
"CommandExecutionRequestApprovalParams": {
"properties": {
"approvalId": {
@@ -722,21 +801,6 @@
}
]
},
"SkillRequestApprovalParams": {
"properties": {
"itemId": {
"type": "string"
},
"skillName": {
"type": "string"
}
},
"required": [
"itemId",
"skillName"
],
"type": "object"
},
"ThreadId": {
"type": "string"
},
@@ -898,30 +962,6 @@
"title": "Item/tool/requestUserInputRequest",
"type": "object"
},
{
"properties": {
"id": {
"$ref": "#/definitions/RequestId"
},
"method": {
"enum": [
"skill/requestApproval"
],
"title": "Skill/requestApprovalRequestMethod",
"type": "string"
},
"params": {
"$ref": "#/definitions/SkillRequestApprovalParams"
}
},
"required": [
"id",
"method",
"params"
],
"title": "Skill/requestApprovalRequest",
"type": "object"
},
{
"description": "Execute a dynamic tool call on the client.",
"properties": {
codex-rs/app-server-protocol/schema/json/SkillRequestApprovalParams.json
-17
View File
@@ -1,17 +0,0 @@
{
"$schema": "http://json-schema.org/draft-07/schema#",
"properties": {
"itemId": {
"type": "string"
},
"skillName": {
"type": "string"
}
},
"required": [
"itemId",
"skillName"
],
"title": "SkillRequestApprovalParams",
"type": "object"
}
codex-rs/app-server-protocol/schema/json/SkillRequestApprovalResponse.json
-22
View File
@@ -1,22 +0,0 @@
{
"$schema": "http://json-schema.org/draft-07/schema#",
"definitions": {
"SkillApprovalDecision": {
"enum": [
"approve",
"decline"
],
"type": "string"
}
},
"properties": {
"decision": {
"$ref": "#/definitions/SkillApprovalDecision"
}
},
"required": [
"decision"
],
"title": "SkillRequestApprovalResponse",
"type": "object"
}
codex-rs/app-server-protocol/schema/json/codex_app_server_protocol.schemas.json
+73 -88
View File
@@ -1442,7 +1442,7 @@
"type": "string"
},
{
"description": "User approved the command and future identical commands should run without prompting.",
"description": "User approved the command and future prompts in the same session-scoped approval cache should run without prompting.",
"enum": [
"acceptForSession"
],
@@ -2843,6 +2843,16 @@
"null"
]
},
"available_decisions": {
"description": "Ordered list of decisions the client may present for this prompt.\n\nWhen absent, clients should derive the legacy default set from the other fields on this request.",
"items": {
"$ref": "#/definitions/ReviewDecision"
},
"type": [
"array",
"null"
]
},
"call_id": {
"description": "Identifier for the associated command execution item.",
"type": "string"
@@ -3052,30 +3062,6 @@
"title": "DynamicToolCallResponseEventMsg",
"type": "object"
},
{
"properties": {
"item_id": {
"type": "string"
},
"skill_name": {
"type": "string"
},
"type": {
"enum": [
"skill_request_approval"
],
"title": "SkillRequestApprovalEventMsgType",
"type": "string"
}
},
"required": [
"item_id",
"skill_name",
"type"
],
"title": "SkillRequestApprovalEventMsg",
"type": "object"
},
{
"properties": {
"id": {
@@ -5700,7 +5686,7 @@
"type": "object"
},
{
"description": "User has approved this command and wants to automatically approve any future identical instances (`command` and `cwd` match exactly) for the remainder of the session.",
"description": "User has approved this request and wants future prompts in the same session-scoped approval cache to be automatically approved for the remainder of the session.",
"enum": [
"approved_for_session"
],
@@ -6732,30 +6718,6 @@
"title": "Item/tool/requestUserInputRequest",
"type": "object"
},
{
"properties": {
"id": {
"$ref": "#/definitions/RequestId"
},
"method": {
"enum": [
"skill/requestApproval"
],
"title": "Skill/requestApprovalRequestMethod",
"type": "string"
},
"params": {
"$ref": "#/definitions/SkillRequestApprovalParams"
}
},
"required": [
"id",
"method",
"params"
],
"title": "Skill/requestApprovalRequest",
"type": "object"
},
{
"description": "Execute a dynamic tool call on the client.",
"properties": {
@@ -6877,43 +6839,6 @@
],
"type": "object"
},
"SkillApprovalDecision": {
"enum": [
"approve",
"decline"
],
"type": "string"
},
"SkillRequestApprovalParams": {
"$schema": "http://json-schema.org/draft-07/schema#",
"properties": {
"itemId": {
"type": "string"
},
"skillName": {
"type": "string"
}
},
"required": [
"itemId",
"skillName"
],
"title": "SkillRequestApprovalParams",
"type": "object"
},
"SkillRequestApprovalResponse": {
"$schema": "http://json-schema.org/draft-07/schema#",
"properties": {
"decision": {
"$ref": "#/definitions/SkillApprovalDecision"
}
},
"required": [
"decision"
],
"title": "SkillRequestApprovalResponse",
"type": "object"
},
"StepStatus": {
"enum": [
"pending",
@@ -10300,6 +10225,16 @@
},
"Model": {
"properties": {
"availabilityNux": {
"anyOf": [
{
"$ref": "#/definitions/v2/ModelAvailabilityNux"
},
{
"type": "null"
}
]
},
"defaultReasoningEffort": {
"$ref": "#/definitions/v2/ReasoningEffort"
},
@@ -10346,6 +10281,16 @@
"string",
"null"
]
},
"upgradeInfo": {
"anyOf": [
{
"$ref": "#/definitions/v2/ModelUpgradeInfo"
},
{
"type": "null"
}
]
}
},
"required": [
@@ -10360,6 +10305,17 @@
],
"type": "object"
},
"ModelAvailabilityNux": {
"properties": {
"message": {
"type": "string"
}
},
"required": [
"message"
],
"type": "object"
},
"ModelListParams": {
"$schema": "http://json-schema.org/draft-07/schema#",
"properties": {
@@ -10448,6 +10404,35 @@
"title": "ModelReroutedNotification",
"type": "object"
},
"ModelUpgradeInfo": {
"properties": {
"migrationMarkdown": {
"type": [
"string",
"null"
]
},
"model": {
"type": "string"
},
"modelLink": {
"type": [
"string",
"null"
]
},
"upgradeCopy": {
"type": [
"string",
"null"
]
}
},
"required": [
"model"
],
"type": "object"
},
"NetworkAccess": {
"enum": [
"restricted",
@@ -11457,7 +11442,7 @@
"type": "string"
},
"output": {
"type": "string"
"$ref": "#/definitions/v2/FunctionCallOutputPayload"
},
"type": {
"enum": [
codex-rs/app-server-protocol/schema/json/v2/ModelListResponse.json
+60
View File
@@ -22,6 +22,16 @@
},
"Model": {
"properties": {
"availabilityNux": {
"anyOf": [
{
"$ref": "#/definitions/ModelAvailabilityNux"
},
{
"type": "null"
}
]
},
"defaultReasoningEffort": {
"$ref": "#/definitions/ReasoningEffort"
},
@@ -68,6 +78,16 @@
"string",
"null"
]
},
"upgradeInfo": {
"anyOf": [
{
"$ref": "#/definitions/ModelUpgradeInfo"
},
{
"type": "null"
}
]
}
},
"required": [
@@ -82,6 +102,46 @@
],
"type": "object"
},
"ModelAvailabilityNux": {
"properties": {
"message": {
"type": "string"
}
},
"required": [
"message"
],
"type": "object"
},
"ModelUpgradeInfo": {
"properties": {
"migrationMarkdown": {
"type": [
"string",
"null"
]
},
"model": {
"type": "string"
},
"modelLink": {
"type": [
"string",
"null"
]
},
"upgradeCopy": {
"type": [
"string",
"null"
]
}
},
"required": [
"model"
],
"type": "object"
},
"ReasoningEffort": {
"description": "See https://platform.openai.com/docs/guides/reasoning?api-mode=responses#get-started-with-reasoning",
"enum": [
codex-rs/app-server-protocol/schema/json/v2/RawResponseItemCompletedNotification.json
+1 -1
View File
@@ -565,7 +565,7 @@
"type": "string"
},
"output": {
"type": "string"
"$ref": "#/definitions/FunctionCallOutputPayload"
},
"type": {
"enum": [
codex-rs/app-server-protocol/schema/json/v2/ThreadResumeParams.json
+1 -1
View File
@@ -615,7 +615,7 @@
"type": "string"
},
"output": {
"type": "string"
"$ref": "#/definitions/FunctionCallOutputPayload"
},
"type": {
"enum": [
codex-rs/app-server-protocol/schema/typescript/EventMsg.ts
+1 -2
View File
@@ -57,7 +57,6 @@ import type { RemoteSkillDownloadedEvent } from "./RemoteSkillDownloadedEvent";
import type { RequestUserInputEvent } from "./RequestUserInputEvent";
import type { ReviewRequest } from "./ReviewRequest";
import type { SessionConfiguredEvent } from "./SessionConfiguredEvent";
import type { SkillRequestApprovalEvent } from "./SkillRequestApprovalEvent";
import type { StreamErrorEvent } from "./StreamErrorEvent";
import type { TerminalInteractionEvent } from "./TerminalInteractionEvent";
import type { ThreadNameUpdatedEvent } from "./ThreadNameUpdatedEvent";
@@ -80,4 +79,4 @@ import type { WebSearchEndEvent } from "./WebSearchEndEvent";
* Response event from the agent
* NOTE: Make sure none of these values have optional types, as it will mess up the extension code-gen.
*/
export type EventMsg = { "type": "error" } & ErrorEvent | { "type": "warning" } & WarningEvent | { "type": "realtime_conversation_started" } & RealtimeConversationStartedEvent | { "type": "realtime_conversation_realtime" } & RealtimeConversationRealtimeEvent | { "type": "realtime_conversation_closed" } & RealtimeConversationClosedEvent | { "type": "model_reroute" } & ModelRerouteEvent | { "type": "context_compacted" } & ContextCompactedEvent | { "type": "thread_rolled_back" } & ThreadRolledBackEvent | { "type": "task_started" } & TurnStartedEvent | { "type": "task_complete" } & TurnCompleteEvent | { "type": "token_count" } & TokenCountEvent | { "type": "agent_message" } & AgentMessageEvent | { "type": "user_message" } & UserMessageEvent | { "type": "agent_message_delta" } & AgentMessageDeltaEvent | { "type": "agent_reasoning" } & AgentReasoningEvent | { "type": "agent_reasoning_delta" } & AgentReasoningDeltaEvent | { "type": "agent_reasoning_raw_content" } & AgentReasoningRawContentEvent | { "type": "agent_reasoning_raw_content_delta" } & AgentReasoningRawContentDeltaEvent | { "type": "agent_reasoning_section_break" } & AgentReasoningSectionBreakEvent | { "type": "session_configured" } & SessionConfiguredEvent | { "type": "thread_name_updated" } & ThreadNameUpdatedEvent | { "type": "mcp_startup_update" } & McpStartupUpdateEvent | { "type": "mcp_startup_complete" } & McpStartupCompleteEvent | { "type": "mcp_tool_call_begin" } & McpToolCallBeginEvent | { "type": "mcp_tool_call_end" } & McpToolCallEndEvent | { "type": "web_search_begin" } & WebSearchBeginEvent | { "type": "web_search_end" } & WebSearchEndEvent | { "type": "exec_command_begin" } & ExecCommandBeginEvent | { "type": "exec_command_output_delta" } & ExecCommandOutputDeltaEvent | { "type": "terminal_interaction" } & TerminalInteractionEvent | { "type": "exec_command_end" } & ExecCommandEndEvent | { "type": "view_image_tool_call" } & ViewImageToolCallEvent | { "type": "exec_approval_request" } & ExecApprovalRequestEvent | { "type": "request_user_input" } & RequestUserInputEvent | { "type": "dynamic_tool_call_request" } & DynamicToolCallRequest | { "type": "dynamic_tool_call_response" } & DynamicToolCallResponseEvent | { "type": "skill_request_approval" } & SkillRequestApprovalEvent | { "type": "elicitation_request" } & ElicitationRequestEvent | { "type": "apply_patch_approval_request" } & ApplyPatchApprovalRequestEvent | { "type": "deprecation_notice" } & DeprecationNoticeEvent | { "type": "background_event" } & BackgroundEventEvent | { "type": "undo_started" } & UndoStartedEvent | { "type": "undo_completed" } & UndoCompletedEvent | { "type": "stream_error" } & StreamErrorEvent | { "type": "patch_apply_begin" } & PatchApplyBeginEvent | { "type": "patch_apply_end" } & PatchApplyEndEvent | { "type": "turn_diff" } & TurnDiffEvent | { "type": "get_history_entry_response" } & GetHistoryEntryResponseEvent | { "type": "mcp_list_tools_response" } & McpListToolsResponseEvent | { "type": "list_custom_prompts_response" } & ListCustomPromptsResponseEvent | { "type": "list_skills_response" } & ListSkillsResponseEvent | { "type": "list_remote_skills_response" } & ListRemoteSkillsResponseEvent | { "type": "remote_skill_downloaded" } & RemoteSkillDownloadedEvent | { "type": "skills_update_available" } | { "type": "plan_update" } & UpdatePlanArgs | { "type": "turn_aborted" } & TurnAbortedEvent | { "type": "shutdown_complete" } | { "type": "entered_review_mode" } & ReviewRequest | { "type": "exited_review_mode" } & ExitedReviewModeEvent | { "type": "raw_response_item" } & RawResponseItemEvent | { "type": "item_started" } & ItemStartedEvent | { "type": "item_completed" } & ItemCompletedEvent | { "type": "agent_message_content_delta" } & AgentMessageContentDeltaEvent | { "type": "plan_delta" } & PlanDeltaEvent | { "type": "reasoning_content_delta" } & ReasoningContentDeltaEvent | { "type": "reasoning_raw_content_delta" } & ReasoningRawContentDeltaEvent | { "type": "collab_agent_spawn_begin" } & CollabAgentSpawnBeginEvent | { "type": "collab_agent_spawn_end" } & CollabAgentSpawnEndEvent | { "type": "collab_agent_interaction_begin" } & CollabAgentInteractionBeginEvent | { "type": "collab_agent_interaction_end" } & CollabAgentInteractionEndEvent | { "type": "collab_waiting_begin" } & CollabWaitingBeginEvent | { "type": "collab_waiting_end" } & CollabWaitingEndEvent | { "type": "collab_close_begin" } & CollabCloseBeginEvent | { "type": "collab_close_end" } & CollabCloseEndEvent | { "type": "collab_resume_begin" } & CollabResumeBeginEvent | { "type": "collab_resume_end" } & CollabResumeEndEvent;
export type EventMsg = { "type": "error" } & ErrorEvent | { "type": "warning" } & WarningEvent | { "type": "realtime_conversation_started" } & RealtimeConversationStartedEvent | { "type": "realtime_conversation_realtime" } & RealtimeConversationRealtimeEvent | { "type": "realtime_conversation_closed" } & RealtimeConversationClosedEvent | { "type": "model_reroute" } & ModelRerouteEvent | { "type": "context_compacted" } & ContextCompactedEvent | { "type": "thread_rolled_back" } & ThreadRolledBackEvent | { "type": "task_started" } & TurnStartedEvent | { "type": "task_complete" } & TurnCompleteEvent | { "type": "token_count" } & TokenCountEvent | { "type": "agent_message" } & AgentMessageEvent | { "type": "user_message" } & UserMessageEvent | { "type": "agent_message_delta" } & AgentMessageDeltaEvent | { "type": "agent_reasoning" } & AgentReasoningEvent | { "type": "agent_reasoning_delta" } & AgentReasoningDeltaEvent | { "type": "agent_reasoning_raw_content" } & AgentReasoningRawContentEvent | { "type": "agent_reasoning_raw_content_delta" } & AgentReasoningRawContentDeltaEvent | { "type": "agent_reasoning_section_break" } & AgentReasoningSectionBreakEvent | { "type": "session_configured" } & SessionConfiguredEvent | { "type": "thread_name_updated" } & ThreadNameUpdatedEvent | { "type": "mcp_startup_update" } & McpStartupUpdateEvent | { "type": "mcp_startup_complete" } & McpStartupCompleteEvent | { "type": "mcp_tool_call_begin" } & McpToolCallBeginEvent | { "type": "mcp_tool_call_end" } & McpToolCallEndEvent | { "type": "web_search_begin" } & WebSearchBeginEvent | { "type": "web_search_end" } & WebSearchEndEvent | { "type": "exec_command_begin" } & ExecCommandBeginEvent | { "type": "exec_command_output_delta" } & ExecCommandOutputDeltaEvent | { "type": "terminal_interaction" } & TerminalInteractionEvent | { "type": "exec_command_end" } & ExecCommandEndEvent | { "type": "view_image_tool_call" } & ViewImageToolCallEvent | { "type": "exec_approval_request" } & ExecApprovalRequestEvent | { "type": "request_user_input" } & RequestUserInputEvent | { "type": "dynamic_tool_call_request" } & DynamicToolCallRequest | { "type": "dynamic_tool_call_response" } & DynamicToolCallResponseEvent | { "type": "elicitation_request" } & ElicitationRequestEvent | { "type": "apply_patch_approval_request" } & ApplyPatchApprovalRequestEvent | { "type": "deprecation_notice" } & DeprecationNoticeEvent | { "type": "background_event" } & BackgroundEventEvent | { "type": "undo_started" } & UndoStartedEvent | { "type": "undo_completed" } & UndoCompletedEvent | { "type": "stream_error" } & StreamErrorEvent | { "type": "patch_apply_begin" } & PatchApplyBeginEvent | { "type": "patch_apply_end" } & PatchApplyEndEvent | { "type": "turn_diff" } & TurnDiffEvent | { "type": "get_history_entry_response" } & GetHistoryEntryResponseEvent | { "type": "mcp_list_tools_response" } & McpListToolsResponseEvent | { "type": "list_custom_prompts_response" } & ListCustomPromptsResponseEvent | { "type": "list_skills_response" } & ListSkillsResponseEvent | { "type": "list_remote_skills_response" } & ListRemoteSkillsResponseEvent | { "type": "remote_skill_downloaded" } & RemoteSkillDownloadedEvent | { "type": "skills_update_available" } | { "type": "plan_update" } & UpdatePlanArgs | { "type": "turn_aborted" } & TurnAbortedEvent | { "type": "shutdown_complete" } | { "type": "entered_review_mode" } & ReviewRequest | { "type": "exited_review_mode" } & ExitedReviewModeEvent | { "type": "raw_response_item" } & RawResponseItemEvent | { "type": "item_started" } & ItemStartedEvent | { "type": "item_completed" } & ItemCompletedEvent | { "type": "agent_message_content_delta" } & AgentMessageContentDeltaEvent | { "type": "plan_delta" } & PlanDeltaEvent | { "type": "reasoning_content_delta" } & ReasoningContentDeltaEvent | { "type": "reasoning_raw_content_delta" } & ReasoningRawContentDeltaEvent | { "type": "collab_agent_spawn_begin" } & CollabAgentSpawnBeginEvent | { "type": "collab_agent_spawn_end" } & CollabAgentSpawnEndEvent | { "type": "collab_agent_interaction_begin" } & CollabAgentInteractionBeginEvent | { "type": "collab_agent_interaction_end" } & CollabAgentInteractionEndEvent | { "type": "collab_waiting_begin" } & CollabWaitingBeginEvent | { "type": "collab_waiting_end" } & CollabWaitingEndEvent | { "type": "collab_close_begin" } & CollabCloseBeginEvent | { "type": "collab_close_end" } & CollabCloseEndEvent | { "type": "collab_resume_begin" } & CollabResumeBeginEvent | { "type": "collab_resume_end" } & CollabResumeEndEvent;
codex-rs/app-server-protocol/schema/typescript/ExecApprovalRequestEvent.ts
+9 -1
View File
@@ -6,6 +6,7 @@ import type { NetworkApprovalContext } from "./NetworkApprovalContext";
import type { NetworkPolicyAmendment } from "./NetworkPolicyAmendment";
import type { ParsedCommand } from "./ParsedCommand";
import type { PermissionProfile } from "./PermissionProfile";
import type { ReviewDecision } from "./ReviewDecision";
export type ExecApprovalRequestEvent = {
/**
@@ -51,4 +52,11 @@ proposed_network_policy_amendments?: Array<NetworkPolicyAmendment>,
/**
* Optional additional filesystem permissions requested for this command.
*/
additional_permissions?: PermissionProfile, parsed_cmd: Array<ParsedCommand>, };
additional_permissions?: PermissionProfile,
/**
* Ordered list of decisions the client may present for this prompt.
*
* When absent, clients should derive the legacy default set from the
* other fields on this request.
*/
available_decisions?: Array<ReviewDecision>, parsed_cmd: Array<ParsedCommand>, };
codex-rs/app-server-protocol/schema/typescript/ResponseItem.ts
+1 -1
View File
@@ -15,4 +15,4 @@ export type ResponseItem = { "type": "message", role: string, content: Array<Con
/**
* Set when using the Responses API.
*/
call_id: string | null, status: LocalShellStatus, action: LocalShellAction, } | { "type": "function_call", name: string, arguments: string, call_id: string, } | { "type": "function_call_output", call_id: string, output: FunctionCallOutputPayload, } | { "type": "custom_tool_call", status?: string, call_id: string, name: string, input: string, } | { "type": "custom_tool_call_output", call_id: string, output: string, } | { "type": "web_search_call", status?: string, action?: WebSearchAction, } | { "type": "ghost_snapshot", ghost_commit: GhostCommit, } | { "type": "compaction", encrypted_content: string, } | { "type": "other" };
call_id: string | null, status: LocalShellStatus, action: LocalShellAction, } | { "type": "function_call", name: string, arguments: string, call_id: string, } | { "type": "function_call_output", call_id: string, output: FunctionCallOutputPayload, } | { "type": "custom_tool_call", status?: string, call_id: string, name: string, input: string, } | { "type": "custom_tool_call_output", call_id: string, output: FunctionCallOutputPayload, } | { "type": "web_search_call", status?: string, action?: WebSearchAction, } | { "type": "ghost_snapshot", ghost_commit: GhostCommit, } | { "type": "compaction", encrypted_content: string, } | { "type": "other" };
codex-rs/app-server-protocol/schema/typescript/ServerRequest.ts
+1 -2
View File
@@ -8,10 +8,9 @@ import type { ChatgptAuthTokensRefreshParams } from "./v2/ChatgptAuthTokensRefre
import type { CommandExecutionRequestApprovalParams } from "./v2/CommandExecutionRequestApprovalParams";
import type { DynamicToolCallParams } from "./v2/DynamicToolCallParams";
import type { FileChangeRequestApprovalParams } from "./v2/FileChangeRequestApprovalParams";
import type { SkillRequestApprovalParams } from "./v2/SkillRequestApprovalParams";
import type { ToolRequestUserInputParams } from "./v2/ToolRequestUserInputParams";
/**
* Request initiated from the server and sent to the client.
*/
export type ServerRequest = { "method": "item/commandExecution/requestApproval", id: RequestId, params: CommandExecutionRequestApprovalParams, } | { "method": "item/fileChange/requestApproval", id: RequestId, params: FileChangeRequestApprovalParams, } | { "method": "item/tool/requestUserInput", id: RequestId, params: ToolRequestUserInputParams, } | { "method": "skill/requestApproval", id: RequestId, params: SkillRequestApprovalParams, } | { "method": "item/tool/call", id: RequestId, params: DynamicToolCallParams, } | { "method": "account/chatgptAuthTokens/refresh", id: RequestId, params: ChatgptAuthTokensRefreshParams, } | { "method": "applyPatchApproval", id: RequestId, params: ApplyPatchApprovalParams, } | { "method": "execCommandApproval", id: RequestId, params: ExecCommandApprovalParams, };
export type ServerRequest = { "method": "item/commandExecution/requestApproval", id: RequestId, params: CommandExecutionRequestApprovalParams, } | { "method": "item/fileChange/requestApproval", id: RequestId, params: FileChangeRequestApprovalParams, } | { "method": "item/tool/requestUserInput", id: RequestId, params: ToolRequestUserInputParams, } | { "method": "item/tool/call", id: RequestId, params: DynamicToolCallParams, } | { "method": "account/chatgptAuthTokens/refresh", id: RequestId, params: ChatgptAuthTokensRefreshParams, } | { "method": "applyPatchApproval", id: RequestId, params: ApplyPatchApprovalParams, } | { "method": "execCommandApproval", id: RequestId, params: ExecCommandApprovalParams, };
codex-rs/app-server-protocol/schema/typescript/SkillRequestApprovalEvent.ts
-5
View File
@@ -1,5 +0,0 @@
// GENERATED CODE! DO NOT MODIFY BY HAND!
// This file was generated by [ts-rs](https://github.com/Aleph-Alpha/ts-rs). Do not edit this file manually.
export type SkillRequestApprovalEvent = { item_id: string, skill_name: string, };
codex-rs/app-server-protocol/schema/typescript/index.ts
-1
View File
@@ -209,7 +209,6 @@ export type { SkillDependencies } from "./SkillDependencies";
export type { SkillErrorInfo } from "./SkillErrorInfo";
export type { SkillInterface } from "./SkillInterface";
export type { SkillMetadata } from "./SkillMetadata";
export type { SkillRequestApprovalEvent } from "./SkillRequestApprovalEvent";
export type { SkillScope } from "./SkillScope";
export type { SkillToolDependency } from "./SkillToolDependency";
export type { SkillsListEntry } from "./SkillsListEntry";
codex-rs/app-server-protocol/schema/typescript/v2/CommandExecutionRequestApprovalParams.ts
+6 -1
View File
@@ -3,6 +3,7 @@
// This file was generated by [ts-rs](https://github.com/Aleph-Alpha/ts-rs). Do not edit this file manually.
import type { AdditionalPermissionProfile } from "./AdditionalPermissionProfile";
import type { CommandAction } from "./CommandAction";
import type { CommandExecutionApprovalDecision } from "./CommandExecutionApprovalDecision";
import type { ExecPolicyAmendment } from "./ExecPolicyAmendment";
import type { NetworkApprovalContext } from "./NetworkApprovalContext";
import type { NetworkPolicyAmendment } from "./NetworkPolicyAmendment";
@@ -49,4 +50,8 @@ proposedExecpolicyAmendment?: ExecPolicyAmendment | null,
/**
* Optional proposed network policy amendments (allow/deny host) for future requests.
*/
proposedNetworkPolicyAmendments?: Array<NetworkPolicyAmendment> | null, };
proposedNetworkPolicyAmendments?: Array<NetworkPolicyAmendment> | null,
/**
* Ordered list of decisions the client may present for this prompt.
*/
availableDecisions?: Array<CommandExecutionApprovalDecision> | null, };
codex-rs/app-server-protocol/schema/typescript/v2/Model.ts
+3 -1
View File
@@ -3,6 +3,8 @@
// This file was generated by [ts-rs](https://github.com/Aleph-Alpha/ts-rs). Do not edit this file manually.
import type { InputModality } from "../InputModality";
import type { ReasoningEffort } from "../ReasoningEffort";
import type { ModelAvailabilityNux } from "./ModelAvailabilityNux";
import type { ModelUpgradeInfo } from "./ModelUpgradeInfo";
import type { ReasoningEffortOption } from "./ReasoningEffortOption";
export type Model = { id: string, model: string, upgrade: string | null, displayName: string, description: string, hidden: boolean, supportedReasoningEfforts: Array<ReasoningEffortOption>, defaultReasoningEffort: ReasoningEffort, inputModalities: Array<InputModality>, supportsPersonality: boolean, isDefault: boolean, };
export type Model = { id: string, model: string, upgrade: string | null, upgradeInfo: ModelUpgradeInfo | null, availabilityNux: ModelAvailabilityNux | null, displayName: string, description: string, hidden: boolean, supportedReasoningEfforts: Array<ReasoningEffortOption>, defaultReasoningEffort: ReasoningEffort, inputModalities: Array<InputModality>, supportsPersonality: boolean, isDefault: boolean, };
codex-rs/app-server-protocol/schema/typescript/v2/SkillApprovalDecision.ts → codex-rs/app-server-protocol/schema/typescript/v2/ModelAvailabilityNux.ts
+1 -1
View File
@@ -2,4 +2,4 @@
// This file was generated by [ts-rs](https://github.com/Aleph-Alpha/ts-rs). Do not edit this file manually.
export type SkillApprovalDecision = "approve" | "decline";
export type ModelAvailabilityNux = { message: string, };
codex-rs/app-server-protocol/schema/typescript/v2/SkillRequestApprovalParams.ts → codex-rs/app-server-protocol/schema/typescript/v2/ModelUpgradeInfo.ts
+1 -1
View File
@@ -2,4 +2,4 @@
// This file was generated by [ts-rs](https://github.com/Aleph-Alpha/ts-rs). Do not edit this file manually.
export type SkillRequestApprovalParams = { itemId: string, skillName: string, };
export type ModelUpgradeInfo = { model: string, upgradeCopy: string | null, modelLink: string | null, migrationMarkdown: string | null, };
codex-rs/app-server-protocol/schema/typescript/v2/SkillRequestApprovalResponse.ts
-6
View File
@@ -1,6 +0,0 @@
// GENERATED CODE! DO NOT MODIFY BY HAND!
// This file was generated by [ts-rs](https://github.com/Aleph-Alpha/ts-rs). Do not edit this file manually.
import type { SkillApprovalDecision } from "./SkillApprovalDecision";
export type SkillRequestApprovalResponse = { decision: SkillApprovalDecision, };
codex-rs/app-server-protocol/schema/typescript/v2/index.ts
+2 -3
View File
@@ -107,10 +107,12 @@ export type { McpToolCallResult } from "./McpToolCallResult";
export type { McpToolCallStatus } from "./McpToolCallStatus";
export type { MergeStrategy } from "./MergeStrategy";
export type { Model } from "./Model";
export type { ModelAvailabilityNux } from "./ModelAvailabilityNux";
export type { ModelListParams } from "./ModelListParams";
export type { ModelListResponse } from "./ModelListResponse";
export type { ModelRerouteReason } from "./ModelRerouteReason";
export type { ModelReroutedNotification } from "./ModelReroutedNotification";
export type { ModelUpgradeInfo } from "./ModelUpgradeInfo";
export type { NetworkAccess } from "./NetworkAccess";
export type { NetworkApprovalContext } from "./NetworkApprovalContext";
export type { NetworkApprovalProtocol } from "./NetworkApprovalProtocol";
@@ -141,13 +143,10 @@ export type { SandboxMode } from "./SandboxMode";
export type { SandboxPolicy } from "./SandboxPolicy";
export type { SandboxWorkspaceWrite } from "./SandboxWorkspaceWrite";
export type { SessionSource } from "./SessionSource";
export type { SkillApprovalDecision } from "./SkillApprovalDecision";
export type { SkillDependencies } from "./SkillDependencies";
export type { SkillErrorInfo } from "./SkillErrorInfo";
export type { SkillInterface } from "./SkillInterface";
export type { SkillMetadata } from "./SkillMetadata";
export type { SkillRequestApprovalParams } from "./SkillRequestApprovalParams";
export type { SkillRequestApprovalResponse } from "./SkillRequestApprovalResponse";
export type { SkillScope } from "./SkillScope";
export type { SkillToolDependency } from "./SkillToolDependency";
export type { SkillsConfigWriteParams } from "./SkillsConfigWriteParams";
codex-rs/app-server-protocol/src/protocol/common.rs
+1 -5
View File
@@ -715,11 +715,6 @@ server_request_definitions! {
response: v2::ToolRequestUserInputResponse,
},
SkillRequestApproval => "skill/requestApproval" {
params: v2::SkillRequestApprovalParams,
response: v2::SkillRequestApprovalResponse,
},
/// Execute a dynamic tool call on the client.
DynamicToolCall => "item/tool/call" {
params: v2::DynamicToolCallParams,
@@ -1545,6 +1540,7 @@ mod tests {
}),
proposed_execpolicy_amendment: None,
proposed_network_policy_amendments: None,
available_decisions: None,
};
let reason = crate::experimental_api::ExperimentalApi::experimental_reason(&params);
assert_eq!(
codex-rs/app-server-protocol/src/protocol/v1.rs
+9
View File
@@ -531,6 +531,15 @@ impl From<V1TextElement> for CoreTextElement {
}
}
impl InputItem {
pub fn text_char_count(&self) -> usize {
match self {
InputItem::Text { text, .. } => text.chars().count(),
InputItem::Image { .. } | InputItem::LocalImage { .. } => 0,
}
}
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
/// Deprecated in favor of AccountLoginCompletedNotification.
codex-rs/app-server-protocol/src/protocol/v2.rs
+69 -24
View File
@@ -31,6 +31,7 @@ use codex_protocol::models::MessagePhase;
use codex_protocol::models::PermissionProfile as CorePermissionProfile;
use codex_protocol::models::ResponseItem;
use codex_protocol::openai_models::InputModality;
use codex_protocol::openai_models::ModelAvailabilityNux as CoreModelAvailabilityNux;
use codex_protocol::openai_models::ReasoningEffort;
use codex_protocol::openai_models::default_input_modalities;
use codex_protocol::parse_command::ParsedCommand as CoreParsedCommand;
@@ -49,6 +50,7 @@ use codex_protocol::protocol::RateLimitWindow as CoreRateLimitWindow;
use codex_protocol::protocol::ReadOnlyAccess as CoreReadOnlyAccess;
use codex_protocol::protocol::RealtimeAudioFrame as CoreRealtimeAudioFrame;
use codex_protocol::protocol::RejectConfig as CoreRejectConfig;
use codex_protocol::protocol::ReviewDecision as CoreReviewDecision;
use codex_protocol::protocol::SessionSource as CoreSessionSource;
use codex_protocol::protocol::SkillDependencies as CoreSkillDependencies;
use codex_protocol::protocol::SkillErrorInfo as CoreSkillErrorInfo;
@@ -741,7 +743,8 @@ pub struct ConfigEdit {
pub enum CommandExecutionApprovalDecision {
/// User approved the command.
Accept,
/// User approved the command and future identical commands should run without prompting.
/// User approved the command and future prompts in the same session-scoped
/// approval cache should run without prompting.
AcceptForSession,
/// User approved the command, and wants to apply the proposed execpolicy amendment so future
/// matching commands can run without prompting.
@@ -758,6 +761,27 @@ pub enum CommandExecutionApprovalDecision {
Cancel,
}
impl From<CoreReviewDecision> for CommandExecutionApprovalDecision {
fn from(value: CoreReviewDecision) -> Self {
match value {
CoreReviewDecision::Approved => Self::Accept,
CoreReviewDecision::ApprovedExecpolicyAmendment {
proposed_execpolicy_amendment,
} => Self::AcceptWithExecpolicyAmendment {
execpolicy_amendment: proposed_execpolicy_amendment.into(),
},
CoreReviewDecision::ApprovedForSession => Self::AcceptForSession,
CoreReviewDecision::NetworkPolicyAmendment {
network_policy_amendment,
} => Self::ApplyNetworkPolicyAmendment {
network_policy_amendment: network_policy_amendment.into(),
},
CoreReviewDecision::Abort => Self::Cancel,
CoreReviewDecision::Denied => Self::Decline,
}
}
}
v2_enum_from_core! {
pub enum NetworkApprovalProtocol from CoreNetworkApprovalProtocol {
Http,
@@ -1366,6 +1390,21 @@ pub struct ModelListParams {
pub include_hidden: Option<bool>,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct ModelAvailabilityNux {
pub message: String,
}
impl From<CoreModelAvailabilityNux> for ModelAvailabilityNux {
fn from(value: CoreModelAvailabilityNux) -> Self {
Self {
message: value.message,
}
}
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
@@ -1373,6 +1412,8 @@ pub struct Model {
pub id: String,
pub model: String,
pub upgrade: Option<String>,
pub upgrade_info: Option<ModelUpgradeInfo>,
pub availability_nux: Option<ModelAvailabilityNux>,
pub display_name: String,
pub description: String,
pub hidden: bool,
@@ -1386,6 +1427,16 @@ pub struct Model {
pub is_default: bool,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct ModelUpgradeInfo {
pub model: String,
pub upgrade_copy: Option<String>,
pub model_link: Option<String>,
pub migration_markdown: Option<String>,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
@@ -3029,6 +3080,18 @@ impl From<CoreUserInput> for UserInput {
}
}
impl UserInput {
pub fn text_char_count(&self) -> usize {
match self {
UserInput::Text { text, .. } => text.chars().count(),
UserInput::Image { .. }
| UserInput::LocalImage { .. }
| UserInput::Skill { .. }
| UserInput::Mention { .. } => 0,
}
}
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(tag = "type", rename_all = "camelCase")]
#[ts(tag = "type")]
@@ -3802,6 +3865,11 @@ pub struct CommandExecutionRequestApprovalParams {
#[serde(default, skip_serializing_if = "Option::is_none")]
#[ts(optional = nullable)]
pub proposed_network_policy_amendments: Option<Vec<NetworkPolicyAmendment>>,
/// Ordered list of decisions the client may present for this prompt.
#[experimental("item/commandExecution/requestApproval.availableDecisions")]
#[serde(default, skip_serializing_if = "Option::is_none")]
#[ts(optional = nullable)]
pub available_decisions: Option<Vec<CommandExecutionApprovalDecision>>,
}
impl CommandExecutionRequestApprovalParams {
@@ -3842,29 +3910,6 @@ pub struct FileChangeRequestApprovalResponse {
pub decision: FileChangeApprovalDecision,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS, ExperimentalApi)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct SkillRequestApprovalParams {
pub item_id: String,
pub skill_name: String,
}
#[derive(Serialize, Deserialize, Debug, Clone, Copy, PartialEq, Eq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub enum SkillApprovalDecision {
Approve,
Decline,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
pub struct SkillRequestApprovalResponse {
pub decision: SkillApprovalDecision,
}
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
#[serde(rename_all = "camelCase")]
#[ts(export_to = "v2/")]
codex-rs/app-server-test-client/src/lib.rs
+4 -22
View File
@@ -57,9 +57,6 @@ use codex_app_server_protocol::SendUserMessageParams;
use codex_app_server_protocol::SendUserMessageResponse;
use codex_app_server_protocol::ServerNotification;
use codex_app_server_protocol::ServerRequest;
use codex_app_server_protocol::SkillApprovalDecision;
use codex_app_server_protocol::SkillRequestApprovalParams;
use codex_app_server_protocol::SkillRequestApprovalResponse;
use codex_app_server_protocol::ThreadItem;
use codex_app_server_protocol::ThreadListParams;
use codex_app_server_protocol::ThreadListResponse;
@@ -1511,9 +1508,6 @@ impl CodexClient {
ServerRequest::FileChangeRequestApproval { request_id, params } => {
self.approve_file_change_request(request_id, params)?;
}
ServerRequest::SkillRequestApproval { request_id, params } => {
self.approve_skill_request(request_id, params)?;
}
other => {
bail!("received unsupported server request: {other:?}");
}
@@ -1540,6 +1534,7 @@ impl CodexClient {
additional_permissions,
proposed_execpolicy_amendment,
proposed_network_policy_amendments,
available_decisions,
} = params;
println!(
@@ -1554,6 +1549,9 @@ impl CodexClient {
if let Some(network_approval_context) = network_approval_context.as_ref() {
println!("< network approval context: {network_approval_context:?}");
}
if let Some(available_decisions) = available_decisions.as_ref() {
println!("< available decisions: {available_decisions:?}");
}
if let Some(command) = command.as_deref() {
println!("< command: {command}");
}
@@ -1593,22 +1591,6 @@ impl CodexClient {
Ok(())
}
fn approve_skill_request(
&mut self,
request_id: RequestId,
params: SkillRequestApprovalParams,
) -> Result<()> {
println!(
"\n< skill approval requested for item {}, skill {}",
params.item_id, params.skill_name
);
let response = SkillRequestApprovalResponse {
decision: SkillApprovalDecision::Approve,
};
self.send_server_request_response(request_id, &response)?;
Ok(())
}
fn approve_file_change_request(
&mut self,
request_id: RequestId,
codex-rs/app-server/README.md
+2 -2
View File
@@ -142,7 +142,7 @@ Example with notification opt-out:
- `thread/realtime/stop` — stop the active realtime session for the thread (experimental); returns `{}`.
- `review/start` — kick off Codex’s automated reviewer for a thread; responds like `turn/start` and emits `item/started`/`item/completed` notifications with `enteredReviewMode` and `exitedReviewMode` items, plus a final assistant `agentMessage` containing the review.
- `command/exec` — run a single command under the server sandbox without starting a thread/turn (handy for utilities and validation).
- `model/list` — list available models (set `includeHidden: true` to include entries with `hidden: true`), with reasoning effort options and optional `upgrade` model ids.
- `model/list` — list available models (set `includeHidden: true` to include entries with `hidden: true`), with reasoning effort options, optional legacy `upgrade` model ids, optional `upgradeInfo` metadata (`model`, `upgradeCopy`, `modelLink`, `migrationMarkdown`), and optional `availabilityNux` metadata.
- `experimentalFeature/list` — list feature flags with stage metadata (`beta`, `underDevelopment`, `stable`, etc.), enabled/default-enabled state, and cursor pagination. For non-beta flags, `displayName`/`description`/`announcement` are `null`.
- `collaborationMode/list` — list available collaboration mode presets (experimental, no pagination). This response omits built-in developer instructions; clients should either pass `settings.developer_instructions: null` when setting a mode to use Codex's built-in instructions, or provide their own instructions explicitly.
- `skills/list` — list skills for one or more `cwd` values (optional `forceReload`).
@@ -710,7 +710,7 @@ Certain actions (shell commands or modifying files) may require explicit user ap
Order of messages:
1. `item/started` — shows the pending `commandExecution` item with `command`, `cwd`, and other fields so you can render the proposed action.
2. `item/commandExecution/requestApproval` (request) — carries the same `itemId`, `threadId`, `turnId`, optionally `approvalId` (for subcommand callbacks), and `reason`. For normal command approvals, it also includes `command`, `cwd`, and `commandActions` for friendly display. When `initialize.params.capabilities.experimentalApi = true`, it may also include experimental `additionalPermissions` describing requested per-command sandbox access. For network-only approvals, those command fields may be omitted and `networkApprovalContext` is provided instead. Optional persistence hints may also be included via `proposedExecpolicyAmendment` and `proposedNetworkPolicyAmendments`.
2. `item/commandExecution/requestApproval` (request) — carries the same `itemId`, `threadId`, `turnId`, optionally `approvalId` (for subcommand callbacks), and `reason`. For normal command approvals, it also includes `command`, `cwd`, and `commandActions` for friendly display. When `initialize.params.capabilities.experimentalApi = true`, it may also include experimental `additionalPermissions` describing requested per-command sandbox access. For network-only approvals, those command fields may be omitted and `networkApprovalContext` is provided instead. Optional persistence hints may also be included via `proposedExecpolicyAmendment` and `proposedNetworkPolicyAmendments`. Clients can prefer `availableDecisions` when present to render the exact set of choices the server wants to expose, while still falling back to the older heuristics if it is omitted.
3. Client response — for example `{ "decision": "accept" }`, `{ "decision": "acceptForSession" }`, `{ "decision": { "acceptWithExecpolicyAmendment": { "execpolicy_amendment": [...] } } }`, `{ "decision": { "applyNetworkPolicyAmendment": { "network_policy_amendment": { "host": "example.com", "action": "allow" } } } }`, `{ "decision": "decline" }`, or `{ "decision": "cancel" }`.
4. `item/completed` — final `commandExecution` item with `status: "completed" | "failed" | "declined"` and execution output. Render this as the authoritative result.
codex-rs/app-server/src/bespoke_event_handling.rs
+6 -35
View File
@@ -58,9 +58,6 @@ use codex_app_server_protocol::ReasoningSummaryTextDeltaNotification;
use codex_app_server_protocol::ReasoningTextDeltaNotification;
use codex_app_server_protocol::ServerNotification;
use codex_app_server_protocol::ServerRequestPayload;
use codex_app_server_protocol::SkillApprovalDecision as V2SkillApprovalDecision;
use codex_app_server_protocol::SkillRequestApprovalParams;
use codex_app_server_protocol::SkillRequestApprovalResponse;
use codex_app_server_protocol::TerminalInteractionNotification;
use codex_app_server_protocol::ThreadItem;
use codex_app_server_protocol::ThreadNameUpdatedNotification;
@@ -111,7 +108,6 @@ use codex_protocol::protocol::TokenCountEvent;
use codex_protocol::protocol::TurnDiffEvent;
use codex_protocol::request_user_input::RequestUserInputAnswer as CoreRequestUserInputAnswer;
use codex_protocol::request_user_input::RequestUserInputResponse as CoreRequestUserInputResponse;
use codex_protocol::skill_approval::SkillApprovalResponse as CoreSkillApprovalResponse;
use codex_shell_command::parse_command::shlex_join;
use std::collections::HashMap;
use std::convert::TryFrom;
@@ -336,6 +332,11 @@ pub(crate) async fn apply_bespoke_event_handling(
.note_permission_requested(&conversation_id.to_string())
.await;
let approval_id_for_op = ev.effective_approval_id();
let available_decisions = ev
.effective_available_decisions()
.into_iter()
.map(CommandExecutionApprovalDecision::from)
.collect::<Vec<_>>();
let ExecApprovalRequestEvent {
call_id,
approval_id,
@@ -432,6 +433,7 @@ pub(crate) async fn apply_bespoke_event_handling(
additional_permissions,
proposed_execpolicy_amendment: proposed_execpolicy_amendment_v2,
proposed_network_policy_amendments: proposed_network_policy_amendments_v2,
available_decisions: Some(available_decisions),
};
let rx = outgoing
.send_request(ServerRequestPayload::CommandExecutionRequestApproval(
@@ -518,37 +520,6 @@ pub(crate) async fn apply_bespoke_event_handling(
}
}
}
EventMsg::SkillRequestApproval(request) => {
if matches!(api_version, ApiVersion::V2) {
let item_id = request.item_id;
let skill_name = request.skill_name;
let params = SkillRequestApprovalParams {
item_id: item_id.clone(),
skill_name,
};
let rx = outgoing
.send_request(ServerRequestPayload::SkillRequestApproval(params))
.await;
tokio::spawn(async move {
let approved = match rx.await {
Ok(Ok(value)) => {
serde_json::from_value::<SkillRequestApprovalResponse>(value)
.map(|response| {
matches!(response.decision, V2SkillApprovalDecision::Approve)
})
.unwrap_or(false)
}
_ => false,
};
let _ = conversation
.submit(Op::SkillApproval {
id: item_id,
response: CoreSkillApprovalResponse { approved },
})
.await;
});
}
}
EventMsg::DynamicToolCallRequest(request) => {
if matches!(api_version, ApiVersion::V2) {
let call_id = request.call_id;
codex-rs/app-server/src/codex_message_processor.rs
+154 -14
View File
@@ -1,5 +1,7 @@
use crate::bespoke_event_handling::apply_bespoke_event_handling;
use crate::error_code::INPUT_TOO_LARGE_ERROR_CODE;
use crate::error_code::INTERNAL_ERROR_CODE;
use crate::error_code::INVALID_PARAMS_ERROR_CODE;
use crate::error_code::INVALID_REQUEST_ERROR_CODE;
use crate::fuzzy_file_search::FuzzyFileSearchSession;
use crate::fuzzy_file_search::run_fuzzy_file_search;
@@ -267,6 +269,7 @@ use codex_protocol::protocol::RolloutItem;
use codex_protocol::protocol::SessionConfiguredEvent;
use codex_protocol::protocol::SessionMetaLine;
use codex_protocol::protocol::USER_MESSAGE_BEGIN;
use codex_protocol::user_input::MAX_USER_INPUT_TEXT_CHARS;
use codex_protocol::user_input::UserInput as CoreInputItem;
use codex_rmcp_client::perform_oauth_login_return_url;
use codex_utils_json_to_toml::json_to_toml;
@@ -557,7 +560,12 @@ impl CodexMessageProcessor {
Ok((review_request, hint))
}
pub async fn process_request(&mut self, connection_id: ConnectionId, request: ClientRequest) {
pub async fn process_request(
&mut self,
connection_id: ConnectionId,
request: ClientRequest,
app_server_client_name: Option<String>,
) {
let to_connection_request_id = |request_id| ConnectionRequestId {
connection_id,
request_id,
@@ -644,8 +652,12 @@ impl CodexMessageProcessor {
.await;
}
ClientRequest::TurnStart { request_id, params } => {
self.turn_start(to_connection_request_id(request_id), params)
.await;
self.turn_start(
to_connection_request_id(request_id),
params,
app_server_client_name.clone(),
)
.await;
}
ClientRequest::TurnSteer { request_id, params } => {
self.turn_steer(to_connection_request_id(request_id), params)
@@ -764,12 +776,20 @@ impl CodexMessageProcessor {
.await;
}
ClientRequest::SendUserMessage { request_id, params } => {
self.send_user_message(to_connection_request_id(request_id), params)
.await;
self.send_user_message(
to_connection_request_id(request_id),
params,
app_server_client_name.clone(),
)
.await;
}
ClientRequest::SendUserTurn { request_id, params } => {
self.send_user_turn(to_connection_request_id(request_id), params)
.await;
self.send_user_turn(
to_connection_request_id(request_id),
params,
app_server_client_name.clone(),
)
.await;
}
ClientRequest::InterruptConversation { request_id, params } => {
self.interrupt_conversation(to_connection_request_id(request_id), params)
@@ -4146,6 +4166,7 @@ impl CodexMessageProcessor {
http_headers,
env_http_headers,
scopes.as_deref().unwrap_or_default(),
server.oauth_resource.as_deref(),
timeout_secs,
config.mcp_oauth_callback_port,
config.mcp_oauth_callback_url.as_deref(),
@@ -4732,6 +4753,36 @@ impl CodexMessageProcessor {
self.outgoing.send_error(request_id, error).await;
}
fn input_too_large_error(actual_chars: usize) -> JSONRPCErrorError {
JSONRPCErrorError {
code: INVALID_PARAMS_ERROR_CODE,
message: format!(
"Input exceeds the maximum length of {MAX_USER_INPUT_TEXT_CHARS} characters."
),
data: Some(serde_json::json!({
"input_error_code": INPUT_TOO_LARGE_ERROR_CODE,
"max_chars": MAX_USER_INPUT_TEXT_CHARS,
"actual_chars": actual_chars,
})),
}
}
fn validate_v1_input_limit(items: &[WireInputItem]) -> Result<(), JSONRPCErrorError> {
let actual_chars: usize = items.iter().map(WireInputItem::text_char_count).sum();
if actual_chars > MAX_USER_INPUT_TEXT_CHARS {
return Err(Self::input_too_large_error(actual_chars));
}
Ok(())
}
fn validate_v2_input_limit(items: &[V2UserInput]) -> Result<(), JSONRPCErrorError> {
let actual_chars: usize = items.iter().map(V2UserInput::text_char_count).sum();
if actual_chars > MAX_USER_INPUT_TEXT_CHARS {
return Err(Self::input_too_large_error(actual_chars));
}
Ok(())
}
async fn send_internal_error(&self, request_id: ConnectionRequestId, message: String) {
let error = JSONRPCErrorError {
code: INTERNAL_ERROR_CODE,
@@ -5026,11 +5077,16 @@ impl CodexMessageProcessor {
&self,
request_id: ConnectionRequestId,
params: SendUserMessageParams,
app_server_client_name: Option<String>,
) {
let SendUserMessageParams {
conversation_id,
items,
} = params;
if let Err(error) = Self::validate_v1_input_limit(&items) {
self.outgoing.send_error(request_id, error).await;
return;
}
let Ok(conversation) = self.thread_manager.get_thread(conversation_id).await else {
let error = JSONRPCErrorError {
code: INVALID_REQUEST_ERROR_CODE,
@@ -5040,6 +5096,12 @@ impl CodexMessageProcessor {
self.outgoing.send_error(request_id, error).await;
return;
};
if let Err(error) =
Self::set_app_server_client_name(conversation.as_ref(), app_server_client_name).await
{
self.outgoing.send_error(request_id, error).await;
return;
}
let mapped_items: Vec<CoreInputItem> = items
.into_iter()
@@ -5070,7 +5132,12 @@ impl CodexMessageProcessor {
.await;
}
async fn send_user_turn(&self, request_id: ConnectionRequestId, params: SendUserTurnParams) {
async fn send_user_turn(
&self,
request_id: ConnectionRequestId,
params: SendUserTurnParams,
app_server_client_name: Option<String>,
) {
let SendUserTurnParams {
conversation_id,
items,
@@ -5082,6 +5149,10 @@ impl CodexMessageProcessor {
summary,
output_schema,
} = params;
if let Err(error) = Self::validate_v1_input_limit(&items) {
self.outgoing.send_error(request_id, error).await;
return;
}
let Ok(conversation) = self.thread_manager.get_thread(conversation_id).await else {
let error = JSONRPCErrorError {
@@ -5092,6 +5163,12 @@ impl CodexMessageProcessor {
self.outgoing.send_error(request_id, error).await;
return;
};
if let Err(error) =
Self::set_app_server_client_name(conversation.as_ref(), app_server_client_name).await
{
self.outgoing.send_error(request_id, error).await;
return;
}
let mapped_items: Vec<CoreInputItem> = items
.into_iter()
@@ -5116,7 +5193,7 @@ impl CodexMessageProcessor {
sandbox_policy,
model,
effort,
summary,
summary: Some(summary),
final_output_json_schema: output_schema,
collaboration_mode: None,
personality: None,
@@ -5205,6 +5282,7 @@ impl CodexMessageProcessor {
connectors::list_cached_accessible_connectors_from_mcp_tools(&config),
connectors::list_cached_all_connectors(&config)
);
let cached_all_connectors = all_connectors.clone();
let (tx, mut rx) = tokio::sync::mpsc::unbounded_channel();
@@ -5231,6 +5309,19 @@ impl CodexMessageProcessor {
let app_list_deadline = tokio::time::Instant::now() + APP_LIST_LOAD_TIMEOUT;
let mut accessible_loaded = false;
let mut all_loaded = false;
let mut last_notified_apps = None;
if accessible_connectors.is_some() || all_connectors.is_some() {
let merged = connectors::with_app_enabled_state(
Self::merge_loaded_apps(
all_connectors.as_deref(),
accessible_connectors.as_deref(),
),
&config,
);
Self::send_app_list_updated_notification(&outgoing, merged.clone()).await;
last_notified_apps = Some(merged);
}
loop {
let result = match tokio::time::timeout_at(app_list_deadline, rx.recv()).await {
@@ -5287,14 +5378,30 @@ impl CodexMessageProcessor {
}
}
let showing_interim_force_refetch = force_refetch && !(accessible_loaded && all_loaded);
let all_connectors_for_update =
if showing_interim_force_refetch && cached_all_connectors.is_some() {
cached_all_connectors.as_deref()
} else {
all_connectors.as_deref()
};
let accessible_connectors_for_update =
if showing_interim_force_refetch && !accessible_loaded {
None
} else {
accessible_connectors.as_deref()
};
let merged = connectors::with_app_enabled_state(
Self::merge_loaded_apps(
all_connectors.as_deref(),
accessible_connectors.as_deref(),
all_connectors_for_update,
accessible_connectors_for_update,
),
&config,
);
Self::send_app_list_updated_notification(&outgoing, merged.clone()).await;
if last_notified_apps.as_ref() != Some(&merged) {
Self::send_app_list_updated_notification(&outgoing, merged.clone()).await;
last_notified_apps = Some(merged.clone());
}
if accessible_loaded && all_loaded {
match Self::paginate_apps(merged.as_slice(), start, limit) {
@@ -5563,7 +5670,16 @@ impl CodexMessageProcessor {
let _ = conversation.submit(Op::Interrupt).await;
}
async fn turn_start(&self, request_id: ConnectionRequestId, params: TurnStartParams) {
async fn turn_start(
&self,
request_id: ConnectionRequestId,
params: TurnStartParams,
app_server_client_name: Option<String>,
) {
if let Err(error) = Self::validate_v2_input_limit(&params.input) {
self.outgoing.send_error(request_id, error).await;
return;
}
let (_, thread) = match self.load_thread(&params.thread_id).await {
Ok(v) => v,
Err(error) => {
@@ -5571,6 +5687,12 @@ impl CodexMessageProcessor {
return;
}
};
if let Err(error) =
Self::set_app_server_client_name(thread.as_ref(), app_server_client_name).await
{
self.outgoing.send_error(request_id, error).await;
return;
}
let collaboration_modes_config = CollaborationModesConfig {
default_mode_request_user_input: thread.enabled(Feature::DefaultModeRequestUserInput),
@@ -5652,6 +5774,20 @@ impl CodexMessageProcessor {
}
}
async fn set_app_server_client_name(
thread: &CodexThread,
app_server_client_name: Option<String>,
) -> Result<(), JSONRPCErrorError> {
thread
.set_app_server_client_name(app_server_client_name)
.await
.map_err(|err| JSONRPCErrorError {
code: INTERNAL_ERROR_CODE,
message: format!("failed to set app server client name: {err}"),
data: None,
})
}
async fn turn_steer(&self, request_id: ConnectionRequestId, params: TurnSteerParams) {
let (_, thread) = match self.load_thread(&params.thread_id).await {
Ok(v) => v,
@@ -5669,6 +5805,10 @@ impl CodexMessageProcessor {
.await;
return;
}
if let Err(error) = Self::validate_v2_input_limit(&params.input) {
self.outgoing.send_error(request_id, error).await;
return;
}
let mapped_items: Vec<CoreInputItem> = params
.input
@@ -7575,7 +7715,7 @@ mod tests {
"role": "user",
"content": [{
"type": "input_text",
"text": "<user_instructions>\n<AGENTS.md contents>\n</user_instructions>".to_string(),
"text": "# AGENTS.md instructions for project\n\n<INSTRUCTIONS>\n<AGENTS.md contents>\n</INSTRUCTIONS>".to_string(),
}],
}),
json!({
codex-rs/app-server/src/error_code.rs
+2
View File
@@ -1,3 +1,5 @@
pub(crate) const INVALID_REQUEST_ERROR_CODE: i64 = -32600;
pub const INVALID_PARAMS_ERROR_CODE: i64 = -32602;
pub(crate) const INTERNAL_ERROR_CODE: i64 = -32603;
pub(crate) const OVERLOADED_ERROR_CODE: i64 = -32001;
pub const INPUT_TOO_LARGE_ERROR_CODE: &str = "input_too_large";
codex-rs/app-server/src/lib.rs
+2
View File
@@ -67,6 +67,8 @@ mod thread_state;
mod thread_status;
mod transport;
pub use crate::error_code::INPUT_TOO_LARGE_ERROR_CODE;
pub use crate::error_code::INVALID_PARAMS_ERROR_CODE;
pub use crate::transport::AppServerTransport;
const LOG_FORMAT_ENV_VAR: &str = "LOG_FORMAT";
codex-rs/app-server/src/message_processor.rs
+3 -1
View File
@@ -140,6 +140,7 @@ pub(crate) struct ConnectionSessionState {
pub(crate) initialized: bool,
pub(crate) experimental_api_enabled: bool,
pub(crate) opted_out_notification_methods: HashSet<String>,
pub(crate) app_server_client_name: Option<String>,
}
pub(crate) struct MessageProcessorArgs {
@@ -329,6 +330,7 @@ impl MessageProcessor {
if let Ok(mut suffix) = USER_AGENT_SUFFIX.lock() {
*suffix = Some(user_agent_suffix);
}
session.app_server_client_name = Some(name.clone());
let user_agent = get_codex_user_agent();
let response = InitializeResponse { user_agent };
@@ -430,7 +432,7 @@ impl MessageProcessor {
// inline the full `CodexMessageProcessor::process_request` future, which
// can otherwise push worker-thread stack usage over the edge.
self.codex_message_processor
.process_request(connection_id, other)
.process_request(connection_id, other, session.app_server_client_name.clone())
.boxed()
.await;
}
codex-rs/app-server/src/models.rs
+9 -1
View File
@@ -1,6 +1,7 @@
use std::sync::Arc;
use codex_app_server_protocol::Model;
use codex_app_server_protocol::ModelUpgradeInfo;
use codex_app_server_protocol::ReasoningEffortOption;
use codex_core::ThreadManager;
use codex_core::models_manager::manager::RefreshStrategy;
@@ -24,7 +25,14 @@ fn model_from_preset(preset: ModelPreset) -> Model {
Model {
id: preset.id.to_string(),
model: preset.model.to_string(),
upgrade: preset.upgrade.map(|upgrade| upgrade.id),
upgrade: preset.upgrade.as_ref().map(|upgrade| upgrade.id.clone()),
upgrade_info: preset.upgrade.as_ref().map(|upgrade| ModelUpgradeInfo {
model: upgrade.id.clone(),
upgrade_copy: upgrade.upgrade_copy.clone(),
model_link: upgrade.model_link.clone(),
migration_markdown: upgrade.migration_markdown.clone(),
}),
availability_nux: preset.availability_nux.map(Into::into),
display_name: preset.display_name.to_string(),
description: preset.description.to_string(),
hidden: !preset.show_in_picker,
codex-rs/app-server/src/transport.rs
+2
View File
@@ -986,6 +986,7 @@ mod tests {
),
proposed_execpolicy_amendment: None,
proposed_network_policy_amendments: None,
available_decisions: None,
},
}),
},
@@ -1047,6 +1048,7 @@ mod tests {
),
proposed_execpolicy_amendment: None,
proposed_network_policy_amendments: None,
available_decisions: None,
},
}),
},
codex-rs/app-server/tests/common/models_cache.rs
+3
View File
@@ -1,6 +1,7 @@
use chrono::DateTime;
use chrono::Utc;
use codex_core::test_support::all_model_presets;
use codex_protocol::config_types::ReasoningSummary;
use codex_protocol::openai_models::ConfigShellToolType;
use codex_protocol::openai_models::ModelInfo;
use codex_protocol::openai_models::ModelPreset;
@@ -30,8 +31,10 @@ fn preset_to_info(preset: &ModelPreset, priority: i32) -> ModelInfo {
base_instructions: "base instructions".to_string(),
model_messages: None,
supports_reasoning_summaries: false,
default_reasoning_summary: ReasoningSummary::Auto,
support_verbosity: false,
default_verbosity: None,
availability_nux: None,
apply_patch_tool_type: None,
truncation_policy: TruncationPolicyConfig::bytes(10_000),
supports_parallel_tool_calls: false,
codex-rs/app-server/tests/suite/output_schema.rs
+83
View File
@@ -1,8 +1,11 @@
use anyhow::Result;
use app_test_support::McpProcess;
use app_test_support::to_response;
use codex_app_server::INPUT_TOO_LARGE_ERROR_CODE;
use codex_app_server::INVALID_PARAMS_ERROR_CODE;
use codex_app_server_protocol::AddConversationListenerParams;
use codex_app_server_protocol::InputItem;
use codex_app_server_protocol::JSONRPCError;
use codex_app_server_protocol::JSONRPCResponse;
use codex_app_server_protocol::NewConversationParams;
use codex_app_server_protocol::NewConversationResponse;
@@ -13,6 +16,7 @@ use codex_protocol::config_types::ReasoningSummary;
use codex_protocol::openai_models::ReasoningEffort;
use codex_protocol::protocol::AskForApproval;
use codex_protocol::protocol::SandboxPolicy;
use codex_protocol::user_input::MAX_USER_INPUT_TEXT_CHARS;
use core_test_support::responses;
use core_test_support::skip_if_no_network;
use pretty_assertions::assert_eq;
@@ -124,6 +128,85 @@ async fn send_user_turn_accepts_output_schema_v1() -> Result<()> {
Ok(())
}
#[tokio::test]
async fn send_user_turn_rejects_oversized_input_v1() -> Result<()> {
let server = responses::start_mock_server().await;
let body = responses::sse(vec![
responses::ev_response_created("resp-1"),
responses::ev_assistant_message("msg-1", "Done"),
responses::ev_completed("resp-1"),
]);
let _response_mock = responses::mount_sse_once(&server, body).await;
let codex_home = TempDir::new()?;
create_config_toml(codex_home.path(), &server.uri())?;
let mut mcp = McpProcess::new(codex_home.path()).await?;
timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
let new_conv_id = mcp
.send_new_conversation_request(NewConversationParams {
..Default::default()
})
.await?;
let new_conv_resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(new_conv_id)),
)
.await??;
let NewConversationResponse {
conversation_id, ..
} = to_response::<NewConversationResponse>(new_conv_resp)?;
let listener_id = mcp
.send_add_conversation_listener_request(AddConversationListenerParams {
conversation_id,
experimental_raw_events: false,
})
.await?;
timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(listener_id)),
)
.await??;
let oversized_input = "x".repeat(MAX_USER_INPUT_TEXT_CHARS + 1);
let send_turn_id = mcp
.send_send_user_turn_request(SendUserTurnParams {
conversation_id,
items: vec![InputItem::Text {
text: oversized_input.clone(),
text_elements: Vec::new(),
}],
cwd: codex_home.path().to_path_buf(),
approval_policy: AskForApproval::Never,
sandbox_policy: SandboxPolicy::DangerFullAccess,
model: "mock-model".to_string(),
effort: Some(ReasoningEffort::Low),
summary: ReasoningSummary::Auto,
output_schema: None,
})
.await?;
let err: JSONRPCError = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_error_message(RequestId::Integer(send_turn_id)),
)
.await??;
assert_eq!(err.error.code, INVALID_PARAMS_ERROR_CODE);
assert_eq!(
err.error.message,
format!("Input exceeds the maximum length of {MAX_USER_INPUT_TEXT_CHARS} characters.")
);
let data = err.error.data.expect("expected structured error data");
assert_eq!(data["input_error_code"], INPUT_TOO_LARGE_ERROR_CODE);
assert_eq!(data["max_chars"], MAX_USER_INPUT_TEXT_CHARS);
assert_eq!(data["actual_chars"], oversized_input.chars().count());
Ok(())
}
#[tokio::test]
async fn send_user_turn_output_schema_is_per_turn_v1() -> Result<()> {
skip_if_no_network!(Ok(()));
codex-rs/app-server/tests/suite/send_message.rs
+74 -14
View File
@@ -3,9 +3,12 @@ use app_test_support::McpProcess;
use app_test_support::create_fake_rollout;
use app_test_support::rollout_path;
use app_test_support::to_response;
use codex_app_server::INPUT_TOO_LARGE_ERROR_CODE;
use codex_app_server::INVALID_PARAMS_ERROR_CODE;
use codex_app_server_protocol::AddConversationListenerParams;
use codex_app_server_protocol::AddConversationSubscriptionResponse;
use codex_app_server_protocol::InputItem;
use codex_app_server_protocol::JSONRPCError;
use codex_app_server_protocol::JSONRPCNotification;
use codex_app_server_protocol::JSONRPCResponse;
use codex_app_server_protocol::NewConversationParams;
@@ -27,6 +30,7 @@ use codex_protocol::protocol::RolloutItem;
use codex_protocol::protocol::RolloutLine;
use codex_protocol::protocol::SandboxPolicy;
use codex_protocol::protocol::TurnContextItem;
use codex_protocol::user_input::MAX_USER_INPUT_TEXT_CHARS;
use core_test_support::responses;
use pretty_assertions::assert_eq;
use std::io::Write;
@@ -210,17 +214,13 @@ async fn test_send_message_raw_notifications_opt_in() -> Result<()> {
})
.await?;
let permissions = read_raw_response_item(&mut mcp, conversation_id).await;
assert_permissions_message(&permissions);
let developer = read_raw_response_item(&mut mcp, conversation_id).await;
assert_permissions_message(&developer);
assert_developer_message(&developer, "Use the test harness tools.");
let instructions = read_raw_response_item(&mut mcp, conversation_id).await;
assert_instructions_message(&instructions);
let environment = read_raw_response_item(&mut mcp, conversation_id).await;
assert_environment_message(&environment);
let contextual_user = read_raw_response_item(&mut mcp, conversation_id).await;
assert_instructions_message(&contextual_user);
assert_environment_message(&contextual_user);
let response: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
@@ -272,6 +272,66 @@ async fn test_send_message_session_not_found() -> Result<()> {
Ok(())
}
#[tokio::test]
async fn test_send_message_rejects_oversized_input() -> Result<()> {
let server = responses::start_mock_server().await;
let body = responses::sse(vec![
responses::ev_response_created("resp-1"),
responses::ev_assistant_message("msg-1", "Done"),
responses::ev_completed("resp-1"),
]);
let _response_mock = responses::mount_sse_once(&server, body).await;
let codex_home = TempDir::new()?;
create_config_toml(codex_home.path(), &server.uri())?;
let mut mcp = McpProcess::new(codex_home.path()).await?;
timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
let new_conv_id = mcp
.send_new_conversation_request(NewConversationParams {
..Default::default()
})
.await?;
let new_conv_resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(new_conv_id)),
)
.await??;
let NewConversationResponse {
conversation_id, ..
} = to_response::<_>(new_conv_resp)?;
let oversized_input = "x".repeat(MAX_USER_INPUT_TEXT_CHARS + 1);
let req_id = mcp
.send_send_user_message_request(SendUserMessageParams {
conversation_id,
items: vec![InputItem::Text {
text: oversized_input.clone(),
text_elements: Vec::new(),
}],
})
.await?;
let err: JSONRPCError = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_error_message(RequestId::Integer(req_id)),
)
.await??;
assert_eq!(err.error.code, INVALID_PARAMS_ERROR_CODE);
assert_eq!(
err.error.message,
format!("Input exceeds the maximum length of {MAX_USER_INPUT_TEXT_CHARS} characters.")
);
let data = err.error.data.expect("expected structured error data");
assert_eq!(data["input_error_code"], INPUT_TOO_LARGE_ERROR_CODE);
assert_eq!(data["max_chars"], MAX_USER_INPUT_TEXT_CHARS);
assert_eq!(data["actual_chars"], oversized_input.chars().count());
Ok(())
}
#[tokio::test]
async fn resume_with_model_mismatch_appends_model_switch_once() -> Result<()> {
let server = responses::start_mock_server().await;
@@ -481,9 +541,8 @@ fn assert_permissions_message(item: &ResponseItem) {
false,
)
.into_text();
assert_eq!(
texts,
vec![expected.as_str()],
assert!(
texts.iter().any(|text| *text == expected),
"expected permissions developer message, got {texts:?}"
);
}
@@ -496,9 +555,8 @@ fn assert_developer_message(item: &ResponseItem, expected_text: &str) {
ResponseItem::Message { role, content, .. } => {
assert_eq!(role, "developer");
let texts = content_texts(content);
assert_eq!(
texts,
vec![expected_text],
assert!(
texts.contains(&expected_text),
"expected developer instructions message, got {texts:?}"
);
}
@@ -562,6 +620,8 @@ fn append_rollout_turn_context(path: &Path, timestamp: &str, model: &str) -> std
item: RolloutItem::TurnContext(TurnContextItem {
turn_id: None,
cwd: PathBuf::from("/"),
current_date: None,
timezone: None,
approval_policy: AskForApproval::Never,
sandbox_policy: SandboxPolicy::DangerFullAccess,
network: None,
codex-rs/app-server/tests/suite/v2/app_list.rs
+37 -2
View File
@@ -1027,6 +1027,41 @@ async fn list_apps_force_refetch_patches_updates_from_cached_snapshots() -> Resu
let first_update = read_app_list_updated_notification(&mut mcp).await?;
assert_eq!(
first_update.data,
vec![
AppInfo {
id: "beta".to_string(),
name: "Beta App".to_string(),
description: Some("Beta v1".to_string()),
logo_url: None,
logo_url_dark: None,
distribution_channel: None,
branding: None,
app_metadata: None,
labels: None,
install_url: Some("https://chatgpt.com/apps/beta-app/beta".to_string()),
is_accessible: true,
is_enabled: true,
},
AppInfo {
id: "alpha".to_string(),
name: "Alpha".to_string(),
description: Some("Alpha v1".to_string()),
logo_url: None,
logo_url_dark: None,
distribution_channel: None,
branding: None,
app_metadata: None,
labels: None,
install_url: Some("https://chatgpt.com/apps/alpha/alpha".to_string()),
is_accessible: false,
is_enabled: true,
},
]
);
let second_update = read_app_list_updated_notification(&mut mcp).await?;
assert_eq!(
second_update.data,
vec![
AppInfo {
id: "alpha".to_string(),
@@ -1073,8 +1108,8 @@ async fn list_apps_force_refetch_patches_updates_from_cached_snapshots() -> Resu
is_accessible: false,
is_enabled: true,
}];
let second_update = read_app_list_updated_notification(&mut mcp).await?;
assert_eq!(second_update.data, expected_final);
let third_update = read_app_list_updated_notification(&mut mcp).await?;
assert_eq!(third_update.data, expected_final);
let refetch_response: JSONRPCResponse = timeout(
DEFAULT_TIMEOUT,
codex-rs/app-server/tests/suite/v2/initialize.rs
+106
View File
@@ -1,16 +1,24 @@
use anyhow::Result;
use app_test_support::McpProcess;
use app_test_support::create_final_assistant_message_sse_response;
use app_test_support::create_mock_responses_server_sequence_unchecked;
use app_test_support::to_response;
use codex_app_server_protocol::ClientInfo;
use codex_app_server_protocol::InitializeCapabilities;
use codex_app_server_protocol::InitializeResponse;
use codex_app_server_protocol::JSONRPCMessage;
use codex_app_server_protocol::JSONRPCResponse;
use codex_app_server_protocol::RequestId;
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 core_test_support::fs_wait;
use pretty_assertions::assert_eq;
use serde_json::Value;
use std::path::Path;
use std::time::Duration;
use tempfile::TempDir;
use tokio::time::timeout;
@@ -178,11 +186,103 @@ async fn initialize_opt_out_notification_methods_filters_notifications() -> Resu
Ok(())
}
#[tokio::test]
async fn turn_start_notify_payload_includes_initialize_client_name() -> Result<()> {
let responses = vec![create_final_assistant_message_sse_response("Done")?];
let server = create_mock_responses_server_sequence_unchecked(responses).await;
let codex_home = TempDir::new()?;
let notify_script = codex_home.path().join("notify.py");
std::fs::write(
&notify_script,
r#"from pathlib import Path
import sys
payload_path = Path(__file__).with_name("notify.json")
tmp_path = payload_path.with_suffix(".json.tmp")
tmp_path.write_text(sys.argv[-1], encoding="utf-8")
tmp_path.replace(payload_path)
"#,
)?;
let notify_file = codex_home.path().join("notify.json");
let notify_script = notify_script
.to_str()
.expect("notify script path should be valid UTF-8");
create_config_toml_with_extra(
codex_home.path(),
&server.uri(),
"never",
&format!(
"notify = [\"python3\", {}]",
toml_basic_string(notify_script)
),
)?;
let mut mcp = McpProcess::new(codex_home.path()).await?;
timeout(
DEFAULT_READ_TIMEOUT,
mcp.initialize_with_client_info(ClientInfo {
name: "xcode".to_string(),
title: Some("Xcode".to_string()),
version: "1.0.0".to_string(),
}),
)
.await??;
let thread_req = mcp
.send_thread_start_request(ThreadStartParams::default())
.await?;
let thread_resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(thread_req)),
)
.await??;
let ThreadStartResponse { thread, .. } = to_response(thread_resp)?;
let turn_req = mcp
.send_turn_start_request(TurnStartParams {
thread_id: thread.id,
input: vec![V2UserInput::Text {
text: "Hello".to_string(),
text_elements: Vec::new(),
}],
..Default::default()
})
.await?;
let turn_resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(turn_req)),
)
.await??;
let _: TurnStartResponse = to_response(turn_resp)?;
timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_notification_message("turn/completed"),
)
.await??;
fs_wait::wait_for_path_exists(&notify_file, Duration::from_secs(5)).await?;
let payload_raw = tokio::fs::read_to_string(&notify_file).await?;
let payload: Value = serde_json::from_str(&payload_raw)?;
assert_eq!(payload["client"], "xcode");
Ok(())
}
// Helper to create a config.toml pointing at the mock model server.
fn create_config_toml(
codex_home: &Path,
server_uri: &str,
approval_policy: &str,
) -> std::io::Result<()> {
create_config_toml_with_extra(codex_home, server_uri, approval_policy, "")
}
fn create_config_toml_with_extra(
codex_home: &Path,
server_uri: &str,
approval_policy: &str,
extra: &str,
) -> std::io::Result<()> {
let config_toml = codex_home.join("config.toml");
std::fs::write(
@@ -195,6 +295,8 @@ sandbox_mode = "read-only"
model_provider = "mock_provider"
{extra}
[model_providers.mock_provider]
name = "Mock provider for test"
base_url = "{server_uri}/v1"
@@ -205,3 +307,7 @@ stream_max_retries = 0
),
)
}
fn toml_basic_string(value: &str) -> String {
format!("\"{}\"", value.replace('\\', "\\\\").replace('"', "\\\""))
}
codex-rs/app-server/tests/suite/v2/mod.rs
-1
View File
@@ -19,7 +19,6 @@ mod realtime_conversation;
mod request_user_input;
mod review;
mod safety_check_downgrade;
mod skill_approval;
mod skills_list;
mod thread_archive;
mod thread_fork;
codex-rs/app-server/tests/suite/v2/model_list.rs
+8
View File
@@ -9,6 +9,7 @@ use codex_app_server_protocol::JSONRPCResponse;
use codex_app_server_protocol::Model;
use codex_app_server_protocol::ModelListParams;
use codex_app_server_protocol::ModelListResponse;
use codex_app_server_protocol::ModelUpgradeInfo;
use codex_app_server_protocol::ReasoningEffortOption;
use codex_app_server_protocol::RequestId;
use codex_protocol::openai_models::ModelPreset;
@@ -24,6 +25,13 @@ fn model_from_preset(preset: &ModelPreset) -> Model {
id: preset.id.clone(),
model: preset.model.clone(),
upgrade: preset.upgrade.as_ref().map(|upgrade| upgrade.id.clone()),
upgrade_info: preset.upgrade.as_ref().map(|upgrade| ModelUpgradeInfo {
model: upgrade.id.clone(),
upgrade_copy: upgrade.upgrade_copy.clone(),
model_link: upgrade.model_link.clone(),
migration_markdown: upgrade.migration_markdown.clone(),
}),
availability_nux: preset.availability_nux.clone().map(Into::into),
display_name: preset.display_name.clone(),
description: preset.description.clone(),
hidden: !preset.show_in_picker,
codex-rs/app-server/tests/suite/v2/skill_approval.rs
-109
View File
@@ -1,109 +0,0 @@
use anyhow::Result;
use app_test_support::McpProcess;
use app_test_support::create_final_assistant_message_sse_response;
use app_test_support::create_mock_responses_server_sequence;
use app_test_support::to_response;
use codex_app_server_protocol::JSONRPCResponse;
use codex_app_server_protocol::RequestId;
use codex_app_server_protocol::ServerRequest;
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 tokio::time::timeout;
const DEFAULT_READ_TIMEOUT: std::time::Duration = std::time::Duration::from_secs(10);
#[tokio::test(flavor = "multi_thread", worker_threads = 4)]
async fn skill_request_approval_round_trip() -> Result<()> {
let codex_home = tempfile::TempDir::new()?;
let server =
create_mock_responses_server_sequence(vec![create_final_assistant_message_sse_response(
"done",
)?])
.await;
create_config_toml(codex_home.path(), &server.uri())?;
let mut mcp = McpProcess::new(codex_home.path()).await?;
timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
let thread_start_id = mcp
.send_thread_start_request(ThreadStartParams {
model: Some("mock-model".to_string()),
..Default::default()
})
.await?;
let thread_start_resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(thread_start_id)),
)
.await??;
let ThreadStartResponse { thread, .. } = to_response(thread_start_resp)?;
let turn_start_id = mcp
.send_turn_start_request(TurnStartParams {
thread_id: thread.id.clone(),
input: vec![V2UserInput::Text {
text: "ask something".to_string(),
text_elements: Vec::new(),
}],
model: Some("mock-model".to_string()),
..Default::default()
})
.await?;
let turn_start_resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(turn_start_id)),
)
.await??;
let TurnStartResponse { turn, .. } = to_response(turn_start_resp)?;
let server_req = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_request_message(),
)
.await??;
let ServerRequest::SkillRequestApproval { request_id, params } = server_req else {
panic!("expected SkillRequestApproval request, got: {server_req:?}");
};
assert_eq!(params.item_id, turn.id);
assert_eq!(params.skill_name, "test-skill");
mcp.send_response(request_id, serde_json::json!({ "decision": "approve" }))
.await?;
timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_notification_message("turn/completed"),
)
.await??;
Ok(())
}
fn create_config_toml(codex_home: &std::path::Path, server_uri: &str) -> std::io::Result<()> {
let config_toml = codex_home.join("config.toml");
std::fs::write(
config_toml,
format!(
r#"
model = "mock-model"
approval_policy = "never"
sandbox_mode = "read-only"
model_provider = "mock_provider"
[features]
skill_approval = true
[model_providers.mock_provider]
name = "Mock provider for test"
base_url = "{server_uri}/v1"
request_max_retries = 0
stream_max_retries = 0
"#
),
)
}
codex-rs/app-server/tests/suite/v2/turn_start.rs
+141
View File
@@ -9,6 +9,8 @@ use app_test_support::create_mock_responses_server_sequence_unchecked;
use app_test_support::create_shell_command_sse_response;
use app_test_support::format_with_current_shell_display;
use app_test_support::to_response;
use codex_app_server::INPUT_TOO_LARGE_ERROR_CODE;
use codex_app_server::INVALID_PARAMS_ERROR_CODE;
use codex_app_server_protocol::ByteRange;
use codex_app_server_protocol::ClientInfo;
use codex_app_server_protocol::CommandExecutionApprovalDecision;
@@ -19,6 +21,7 @@ use codex_app_server_protocol::FileChangeOutputDeltaNotification;
use codex_app_server_protocol::FileChangeRequestApprovalResponse;
use codex_app_server_protocol::ItemCompletedNotification;
use codex_app_server_protocol::ItemStartedNotification;
use codex_app_server_protocol::JSONRPCError;
use codex_app_server_protocol::JSONRPCNotification;
use codex_app_server_protocol::JSONRPCResponse;
use codex_app_server_protocol::PatchApplyStatus;
@@ -45,6 +48,7 @@ use codex_protocol::config_types::Personality;
use codex_protocol::config_types::ReasoningSummary;
use codex_protocol::config_types::Settings;
use codex_protocol::openai_models::ReasoningEffort;
use codex_protocol::user_input::MAX_USER_INPUT_TEXT_CHARS;
use core_test_support::responses;
use core_test_support::skip_if_no_network;
use pretty_assertions::assert_eq;
@@ -223,6 +227,143 @@ async fn turn_start_emits_user_message_item_with_text_elements() -> Result<()> {
Ok(())
}
#[tokio::test]
async fn turn_start_accepts_text_at_limit_with_mention_item() -> Result<()> {
let responses = vec![create_final_assistant_message_sse_response("Done")?];
let server = create_mock_responses_server_sequence_unchecked(responses).await;
let codex_home = TempDir::new()?;
create_config_toml(
codex_home.path(),
&server.uri(),
"never",
&BTreeMap::from([(Feature::Personality, true)]),
)?;
let mut mcp = McpProcess::new(codex_home.path()).await?;
timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
let thread_req = mcp
.send_thread_start_request(ThreadStartParams {
model: Some("mock-model".to_string()),
..Default::default()
})
.await?;
let thread_resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(thread_req)),
)
.await??;
let ThreadStartResponse { thread, .. } = to_response::<ThreadStartResponse>(thread_resp)?;
let turn_req = mcp
.send_turn_start_request(TurnStartParams {
thread_id: thread.id,
input: vec![
V2UserInput::Text {
text: "x".repeat(MAX_USER_INPUT_TEXT_CHARS),
text_elements: Vec::new(),
},
V2UserInput::Mention {
name: "Demo App".to_string(),
path: "app://demo-app".to_string(),
},
],
..Default::default()
})
.await?;
let turn_resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(turn_req)),
)
.await??;
let TurnStartResponse { turn } = to_response::<TurnStartResponse>(turn_resp)?;
assert_eq!(turn.status, TurnStatus::InProgress);
timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_notification_message("turn/completed"),
)
.await??;
Ok(())
}
#[tokio::test]
async fn turn_start_rejects_combined_oversized_text_input() -> Result<()> {
let codex_home = TempDir::new()?;
create_config_toml(
codex_home.path(),
"http://localhost/unused",
"never",
&BTreeMap::from([(Feature::Personality, true)]),
)?;
let mut mcp = McpProcess::new(codex_home.path()).await?;
timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
let thread_req = mcp
.send_thread_start_request(ThreadStartParams {
model: Some("mock-model".to_string()),
..Default::default()
})
.await?;
let thread_resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(thread_req)),
)
.await??;
let ThreadStartResponse { thread, .. } = to_response::<ThreadStartResponse>(thread_resp)?;
let first = "x".repeat(MAX_USER_INPUT_TEXT_CHARS / 2);
let second = "y".repeat(MAX_USER_INPUT_TEXT_CHARS / 2 + 1);
let actual_chars = first.chars().count() + second.chars().count();
let turn_req = mcp
.send_turn_start_request(TurnStartParams {
thread_id: thread.id,
input: vec![
V2UserInput::Text {
text: first,
text_elements: Vec::new(),
},
V2UserInput::Text {
text: second,
text_elements: Vec::new(),
},
],
..Default::default()
})
.await?;
let err: JSONRPCError = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_error_message(RequestId::Integer(turn_req)),
)
.await??;
assert_eq!(err.error.code, INVALID_PARAMS_ERROR_CODE);
assert_eq!(
err.error.message,
format!("Input exceeds the maximum length of {MAX_USER_INPUT_TEXT_CHARS} characters.")
);
let data = err.error.data.expect("expected structured error data");
assert_eq!(data["input_error_code"], INPUT_TOO_LARGE_ERROR_CODE);
assert_eq!(data["max_chars"], MAX_USER_INPUT_TEXT_CHARS);
assert_eq!(data["actual_chars"], actual_chars);
let turn_started = tokio::time::timeout(
std::time::Duration::from_millis(250),
mcp.read_stream_until_notification_message("turn/started"),
)
.await;
assert!(
turn_started.is_err(),
"did not expect a turn/started notification for rejected input"
);
Ok(())
}
#[tokio::test]
async fn turn_start_emits_notifications_and_accepts_model_override() -> Result<()> {
// Provide a mock server and config so model wiring is valid.
codex-rs/app-server/tests/suite/v2/turn_steer.rs
+106
View File
@@ -6,6 +6,8 @@ use app_test_support::create_mock_responses_server_sequence;
use app_test_support::create_mock_responses_server_sequence_unchecked;
use app_test_support::create_shell_command_sse_response;
use app_test_support::to_response;
use codex_app_server::INPUT_TOO_LARGE_ERROR_CODE;
use codex_app_server::INVALID_PARAMS_ERROR_CODE;
use codex_app_server_protocol::JSONRPCError;
use codex_app_server_protocol::JSONRPCNotification;
use codex_app_server_protocol::JSONRPCResponse;
@@ -17,6 +19,7 @@ use codex_app_server_protocol::TurnStartResponse;
use codex_app_server_protocol::TurnSteerParams;
use codex_app_server_protocol::TurnSteerResponse;
use codex_app_server_protocol::UserInput as V2UserInput;
use codex_protocol::user_input::MAX_USER_INPUT_TEXT_CHARS;
use tempfile::TempDir;
use tokio::time::timeout;
@@ -67,6 +70,109 @@ async fn turn_steer_requires_active_turn() -> Result<()> {
Ok(())
}
#[tokio::test]
async fn turn_steer_rejects_oversized_text_input() -> Result<()> {
#[cfg(target_os = "windows")]
let shell_command = vec![
"powershell".to_string(),
"-Command".to_string(),
"Start-Sleep -Seconds 10".to_string(),
];
#[cfg(not(target_os = "windows"))]
let shell_command = vec!["sleep".to_string(), "10".to_string()];
let tmp = TempDir::new()?;
let codex_home = tmp.path().join("codex_home");
std::fs::create_dir(&codex_home)?;
let working_directory = tmp.path().join("workdir");
std::fs::create_dir(&working_directory)?;
let server =
create_mock_responses_server_sequence_unchecked(vec![create_shell_command_sse_response(
shell_command.clone(),
Some(&working_directory),
Some(10_000),
"call_sleep",
)?])
.await;
create_config_toml(&codex_home, &server.uri())?;
let mut mcp = McpProcess::new(&codex_home).await?;
timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
let thread_req = mcp
.send_thread_start_request(ThreadStartParams {
model: Some("mock-model".to_string()),
..Default::default()
})
.await?;
let thread_resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(thread_req)),
)
.await??;
let ThreadStartResponse { thread, .. } = to_response::<ThreadStartResponse>(thread_resp)?;
let turn_req = mcp
.send_turn_start_request(TurnStartParams {
thread_id: thread.id.clone(),
input: vec![V2UserInput::Text {
text: "run sleep".to_string(),
text_elements: Vec::new(),
}],
cwd: Some(working_directory.clone()),
..Default::default()
})
.await?;
let turn_resp: JSONRPCResponse = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_response_message(RequestId::Integer(turn_req)),
)
.await??;
let TurnStartResponse { turn } = to_response::<TurnStartResponse>(turn_resp)?;
let _task_started: JSONRPCNotification = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_notification_message("codex/event/task_started"),
)
.await??;
let oversized_input = "x".repeat(MAX_USER_INPUT_TEXT_CHARS + 1);
let steer_req = mcp
.send_turn_steer_request(TurnSteerParams {
thread_id: thread.id.clone(),
input: vec![V2UserInput::Text {
text: oversized_input.clone(),
text_elements: Vec::new(),
}],
expected_turn_id: turn.id.clone(),
})
.await?;
let steer_err: JSONRPCError = timeout(
DEFAULT_READ_TIMEOUT,
mcp.read_stream_until_error_message(RequestId::Integer(steer_req)),
)
.await??;
assert_eq!(steer_err.error.code, INVALID_PARAMS_ERROR_CODE);
assert_eq!(
steer_err.error.message,
format!("Input exceeds the maximum length of {MAX_USER_INPUT_TEXT_CHARS} characters.")
);
let data = steer_err
.error
.data
.expect("expected structured error data");
assert_eq!(data["input_error_code"], INPUT_TOO_LARGE_ERROR_CODE);
assert_eq!(data["max_chars"], MAX_USER_INPUT_TEXT_CHARS);
assert_eq!(data["actual_chars"], oversized_input.chars().count());
mcp.interrupt_turn_and_wait_for_aborted(thread.id, turn.id, DEFAULT_READ_TIMEOUT)
.await?;
Ok(())
}
#[tokio::test]
async fn turn_steer_returns_active_turn_id() -> Result<()> {
#[cfg(target_os = "windows")]
codex-rs/chatgpt/src/connectors.rs
+1 -6
View File
@@ -298,12 +298,7 @@ fn merge_directory_app(existing: &mut DirectoryApp, incoming: DirectoryApp) {
.as_deref()
.map(|value| !value.trim().is_empty())
.unwrap_or(false);
let existing_description_present = existing
.description
.as_deref()
.map(|value| !value.trim().is_empty())
.unwrap_or(false);
if !existing_description_present && incoming_description_present {
if incoming_description_present {
existing.description = description;
}
codex-rs/cli/src/login.rs
+1 -1
View File
@@ -22,7 +22,7 @@ const LOGIN_SUCCESS_MESSAGE: &str = "Successfully logged in";
fn print_login_server_start(actual_port: u16, auth_url: &str) {
eprintln!(
"Starting local login server on http://localhost:{actual_port}.\nIf your browser did not open, navigate to this URL to authenticate:\n\n{auth_url}"
"Starting local login server on http://localhost:{actual_port}.\nIf your browser did not open, navigate to this URL to authenticate:\n\n{auth_url}\n\nOn a remote or headless machine? Use `codex login --device-auth` instead."
);
}
codex-rs/cli/src/main.rs
+1
View File
@@ -808,6 +808,7 @@ async fn cli_main(arg0_paths: Arg0DispatchPaths) -> anyhow::Result<()> {
stage_width = stage_width.max(stage.len());
rows.push((name, stage, enabled));
}
rows.sort_unstable_by_key(|(name, _, _)| *name);
for (name, stage, enabled) in rows {
println!("{name:<name_width$} {stage:<stage_width$} {enabled}");
codex-rs/cli/src/mcp_cmd.rs
+3
View File
@@ -250,6 +250,7 @@ async fn run_add(config_overrides: &CliConfigOverrides, add_args: AddArgs) -> Re
enabled_tools: None,
disabled_tools: None,
scopes: None,
oauth_resource: None,
};
servers.insert(name.clone(), new_entry);
@@ -272,6 +273,7 @@ async fn run_add(config_overrides: &CliConfigOverrides, add_args: AddArgs) -> Re
oauth_config.http_headers,
oauth_config.env_http_headers,
&Vec::new(),
None,
config.mcp_oauth_callback_port,
config.mcp_oauth_callback_url.as_deref(),
)
@@ -356,6 +358,7 @@ async fn run_login(config_overrides: &CliConfigOverrides, login_args: LoginArgs)
http_headers,
env_http_headers,
&scopes,
server.oauth_resource.as_deref(),
config.mcp_oauth_callback_port,
config.mcp_oauth_callback_url.as_deref(),
)
codex-rs/cli/tests/features.rs
+31
View File
@@ -2,6 +2,7 @@ use std::path::Path;
use anyhow::Result;
use predicates::str::contains;
use pretty_assertions::assert_eq;
use tempfile::TempDir;
fn codex_command(codex_home: &Path) -> Result<assert_cmd::Command> {
@@ -58,3 +59,33 @@ async fn features_enable_under_development_feature_prints_warning() -> Result<()
Ok(())
}
#[tokio::test]
async fn features_list_is_sorted_alphabetically_by_feature_name() -> Result<()> {
let codex_home = TempDir::new()?;
let mut cmd = codex_command(codex_home.path())?;
let output = cmd
.args(["features", "list"])
.assert()
.success()
.get_output()
.stdout
.clone();
let stdout = String::from_utf8(output)?;
let actual_names = stdout
.lines()
.map(|line| {
line.split_once(" ")
.map(|(name, _)| name.trim_end().to_string())
.expect("feature list output should contain aligned columns")
})
.collect::<Vec<_>>();
let mut expected_names = actual_names.clone();
expected_names.sort();
assert_eq!(actual_names, expected_names);
Ok(())
}
codex-rs/cloud-requirements/src/lib.rs
+127 -2
View File
@@ -28,17 +28,21 @@ use serde::Serialize;
use sha2::Sha256;
use std::path::PathBuf;
use std::sync::Arc;
use std::sync::Mutex;
use std::sync::OnceLock;
use std::time::Duration;
use std::time::Instant;
use thiserror::Error;
use tokio::fs;
use tokio::task::JoinHandle;
use tokio::time::sleep;
use tokio::time::timeout;
const CLOUD_REQUIREMENTS_TIMEOUT: Duration = Duration::from_secs(15);
const CLOUD_REQUIREMENTS_MAX_ATTEMPTS: usize = 5;
const CLOUD_REQUIREMENTS_CACHE_FILENAME: &str = "cloud-requirements-cache.json";
const CLOUD_REQUIREMENTS_CACHE_TTL: Duration = Duration::from_secs(60 * 60);
const CLOUD_REQUIREMENTS_CACHE_REFRESH_INTERVAL: Duration = Duration::from_secs(5 * 60);
const CLOUD_REQUIREMENTS_CACHE_TTL: Duration = Duration::from_secs(30 * 60);
const CLOUD_REQUIREMENTS_CACHE_WRITE_HMAC_KEY: &[u8] =
b"codex-cloud-requirements-cache-v3-064f8542-75b4-494c-a294-97d3ce597271";
const CLOUD_REQUIREMENTS_CACHE_READ_HMAC_KEYS: &[&[u8]] =
@@ -46,6 +50,11 @@ const CLOUD_REQUIREMENTS_CACHE_READ_HMAC_KEYS: &[&[u8]] =
type HmacSha256 = Hmac<Sha256>;
fn refresher_task_slot() -> &'static Mutex<Option<JoinHandle<()>>> {
static REFRESHER_TASK: OnceLock<Mutex<Option<JoinHandle<()>>>> = OnceLock::new();
REFRESHER_TASK.get_or_init(|| Mutex::new(None))
}
#[derive(Clone, Copy, Debug, Eq, PartialEq)]
enum FetchCloudRequirementsStatus {
BackendClientInit,
@@ -188,6 +197,7 @@ impl RequirementsFetcher for BackendRequirementsFetcher {
}
}
#[derive(Clone)]
struct CloudRequirementsService {
auth_manager: Arc<AuthManager>,
fetcher: Arc<dyn RequirementsFetcher>,
@@ -325,6 +335,54 @@ impl CloudRequirementsService {
None
}
async fn refresh_cache_in_background(&self) {
loop {
sleep(CLOUD_REQUIREMENTS_CACHE_REFRESH_INTERVAL).await;
match timeout(self.timeout, self.refresh_cache()).await {
Ok(true) => {}
Ok(false) => break,
Err(_) => {
tracing::warn!(
"Timed out refreshing cloud requirements cache from remote; keeping existing cache"
);
}
}
}
}
async fn refresh_cache(&self) -> bool {
let Some(auth) = self.auth_manager.auth().await else {
return false;
};
if !auth.is_chatgpt_auth()
|| !matches!(
auth.account_plan_type(),
Some(PlanType::Business | PlanType::Enterprise)
)
{
return false;
}
let token_data = auth.get_token_data().ok();
let chatgpt_user_id = token_data
.as_ref()
.and_then(|token_data| token_data.id_token.chatgpt_user_id.as_deref());
let account_id = auth.get_account_id();
let account_id = account_id.as_deref();
if self
.fetch_with_retries(&auth, chatgpt_user_id, account_id)
.await
.is_none()
{
tracing::warn!(
path = %self.cache_path.display(),
"Failed to refresh cloud requirements cache from remote"
);
}
true
}
async fn load_cache(
&self,
chatgpt_user_id: Option<&str>,
@@ -452,7 +510,17 @@ pub fn cloud_requirements_loader(
codex_home,
CLOUD_REQUIREMENTS_TIMEOUT,
);
let refresh_service = service.clone();
let task = tokio::spawn(async move { service.fetch_with_timeout().await });
let refresh_task =
tokio::spawn(async move { refresh_service.refresh_cache_in_background().await });
let mut refresher_guard = refresher_task_slot().lock().unwrap_or_else(|err| {
tracing::warn!("cloud requirements refresher task slot was poisoned");
err.into_inner()
});
if let Some(existing_task) = refresher_guard.replace(refresh_task) {
existing_task.abort();
}
CloudRequirementsLoader::new(async move {
task.await
.inspect_err(|err| tracing::warn!(error = %err, "Cloud requirements task failed"))
@@ -1052,7 +1120,11 @@ mod tests {
let cache_file: CloudRequirementsCacheFile =
serde_json::from_str(&std::fs::read_to_string(path).expect("read cache"))
.expect("parse cache");
assert!(cache_file.signed_payload.expires_at > Utc::now());
assert!(
cache_file.signed_payload.expires_at
<= cache_file.signed_payload.cached_at + ChronoDuration::minutes(30)
);
assert!(cache_file.signed_payload.expires_at > cache_file.signed_payload.cached_at);
assert!(cache_file.signed_payload.cached_at <= Utc::now());
assert_eq!(
cache_file.signed_payload.chatgpt_user_id,
@@ -1130,4 +1202,57 @@ mod tests {
CLOUD_REQUIREMENTS_MAX_ATTEMPTS
);
}
#[tokio::test]
async fn refresh_from_remote_updates_cached_cloud_requirements() {
let codex_home = tempdir().expect("tempdir");
let fetcher = Arc::new(SequenceFetcher::new(vec![
Ok(Some("allowed_approval_policies = [\"never\"]".to_string())),
Ok(Some(
"allowed_approval_policies = [\"on-request\"]".to_string(),
)),
]));
let service = CloudRequirementsService::new(
auth_manager_with_plan("business"),
fetcher,
codex_home.path().to_path_buf(),
CLOUD_REQUIREMENTS_TIMEOUT,
);
assert_eq!(
service.fetch().await,
Some(ConfigRequirementsToml {
allowed_approval_policies: Some(vec![AskForApproval::Never]),
allowed_sandbox_modes: None,
allowed_web_search_modes: None,
mcp_servers: None,
rules: None,
enforce_residency: None,
network: None,
})
);
service.refresh_cache().await;
let path = codex_home.path().join(CLOUD_REQUIREMENTS_CACHE_FILENAME);
let cache_file: CloudRequirementsCacheFile =
serde_json::from_str(&std::fs::read_to_string(path).expect("read cache"))
.expect("parse cache");
assert_eq!(
cache_file
.signed_payload
.contents
.as_deref()
.and_then(|contents| parse_cloud_requirements(contents).ok().flatten()),
Some(ConfigRequirementsToml {
allowed_approval_policies: Some(vec![AskForApproval::OnRequest]),
allowed_sandbox_modes: None,
allowed_web_search_modes: None,
mcp_servers: None,
rules: None,
enforce_residency: None,
network: None,
})
);
}
}
codex-rs/codex-api/src/endpoint/realtime_websocket/methods.rs
+53 -3
View File
@@ -25,6 +25,8 @@ use tokio_tungstenite::WebSocketStream;
use tokio_tungstenite::tungstenite::Error as WsError;
use tokio_tungstenite::tungstenite::Message;
use tokio_tungstenite::tungstenite::client::IntoClientRequest;
use tracing::debug;
use tracing::error;
use tracing::info;
use tracing::trace;
use tungstenite::protocol::WebSocketConfig;
@@ -62,15 +64,23 @@ impl WsStream {
};
match command {
WsCommand::Send { message, tx_result } => {
debug!("realtime websocket sending message");
let result = inner.send(message).await;
let should_break = result.is_err();
if let Err(err) = &result {
error!("realtime websocket send failed: {err}");
}
let _ = tx_result.send(result);
if should_break {
break;
}
}
WsCommand::Close { tx_result } => {
info!("realtime websocket sending close");
let result = inner.close(None).await;
if let Err(err) = &result {
error!("realtime websocket close failed: {err}");
}
let _ = tx_result.send(result);
break;
}
@@ -82,7 +92,9 @@ impl WsStream {
};
match message {
Ok(Message::Ping(payload)) => {
trace!(payload_len = payload.len(), "realtime websocket received ping");
if let Err(err) = inner.send(Message::Pong(payload)).await {
error!("realtime websocket failed to send pong: {err}");
let _ = tx_message.send(Err(err));
break;
}
@@ -93,6 +105,24 @@ impl WsStream {
| Message::Close(_)
| Message::Frame(_))) => {
let is_close = matches!(message, Message::Close(_));
match &message {
Message::Text(_) => trace!("realtime websocket received text frame"),
Message::Binary(binary) => {
error!(
payload_len = binary.len(),
"realtime websocket received unexpected binary frame"
);
}
Message::Close(frame) => info!(
"realtime websocket received close frame: code={:?} reason={:?}",
frame.as_ref().map(|frame| frame.code),
frame.as_ref().map(|frame| frame.reason.as_str())
),
Message::Frame(_) => {
trace!("realtime websocket received raw frame");
}
Message::Ping(_) | Message::Pong(_) => {}
}
if tx_message.send(Ok(message)).is_err() {
break;
}
@@ -101,6 +131,7 @@ impl WsStream {
}
}
Err(err) => {
error!("realtime websocket receive failed: {err}");
let _ = tx_message.send(Err(err));
break;
}
@@ -108,6 +139,7 @@ impl WsStream {
}
}
}
info!("realtime websocket pump exiting");
});
(
@@ -298,7 +330,7 @@ impl RealtimeWebsocketWriter {
async fn send_json(&self, message: RealtimeOutboundMessage) -> Result<(), ApiError> {
let payload = serde_json::to_string(&message)
.map_err(|err| ApiError::Stream(format!("failed to encode realtime request: {err}")))?;
trace!("realtime websocket request: {payload}");
debug!(?message, "realtime websocket request");
if self.is_closed.load(Ordering::SeqCst) {
return Err(ApiError::Stream(
@@ -325,12 +357,14 @@ impl RealtimeWebsocketEvents {
Some(Ok(msg)) => msg,
Some(Err(err)) => {
self.is_closed.store(true, Ordering::SeqCst);
error!("realtime websocket read failed: {err}");
return Err(ApiError::Stream(format!(
"failed to read websocket message: {err}"
)));
}
None => {
self.is_closed.store(true, Ordering::SeqCst);
info!("realtime websocket event stream ended");
return Ok(None);
}
};
@@ -338,11 +372,18 @@ impl RealtimeWebsocketEvents {
match msg {
Message::Text(text) => {
if let Some(event) = parse_realtime_event(&text) {
debug!(?event, "realtime websocket parsed event");
return Ok(Some(event));
}
debug!("realtime websocket ignored unsupported text frame");
}
Message::Close(_) => {
Message::Close(frame) => {
self.is_closed.store(true, Ordering::SeqCst);
info!(
"realtime websocket closed: code={:?} reason={:?}",
frame.as_ref().map(|frame| frame.code),
frame.as_ref().map(|frame| frame.reason.as_str())
);
return Ok(None);
}
Message::Binary(_) => {
@@ -383,15 +424,24 @@ impl RealtimeWebsocketClient {
request.headers_mut().extend(headers);
info!("connecting realtime websocket: {ws_url}");
let (stream, _) =
let (stream, response) =
tokio_tungstenite::connect_async_with_config(request, Some(websocket_config()), false)
.await
.map_err(|err| {
ApiError::Stream(format!("failed to connect realtime websocket: {err}"))
})?;
info!(
ws_url = %ws_url,
status = %response.status(),
"realtime websocket connected"
);
let (stream, rx_message) = WsStream::new(stream);
let connection = RealtimeWebsocketConnection::new(stream, rx_message);
debug!(
conversation_id = config.session_id.as_deref().unwrap_or("<none>"),
"realtime websocket sending session.create"
);
connection
.send_session_create(config.prompt, config.session_id)
.await?;
codex-rs/codex-api/tests/models_integration.rs
+3
View File
@@ -3,6 +3,7 @@ use codex_api::ModelsClient;
use codex_api::provider::Provider;
use codex_api::provider::RetryConfig;
use codex_client::ReqwestTransport;
use codex_protocol::config_types::ReasoningSummary;
use codex_protocol::openai_models::ConfigShellToolType;
use codex_protocol::openai_models::ModelInfo;
use codex_protocol::openai_models::ModelVisibility;
@@ -78,8 +79,10 @@ async fn models_client_hits_models_endpoint() {
base_instructions: "base instructions".to_string(),
model_messages: None,
supports_reasoning_summaries: false,
default_reasoning_summary: ReasoningSummary::Auto,
support_verbosity: false,
default_verbosity: None,
availability_nux: None,
apply_patch_tool_type: None,
truncation_policy: TruncationPolicyConfig::bytes(10_000),
supports_parallel_tool_calls: false,
codex-rs/core/Cargo.toml
+1
View File
@@ -61,6 +61,7 @@ env-flags = { workspace = true }
eventsource-stream = { workspace = true }
futures = { workspace = true }
http = { workspace = true }
iana-time-zone = { workspace = true }
indexmap = { workspace = true }
keyring = { workspace = true, features = ["crypto-rust"] }
libc = { workspace = true }
codex-rs/core/config.schema.json
+38 -6
View File
@@ -409,9 +409,6 @@
"shell_zsh_fork": {
"type": "boolean"
},
"skill_approval": {
"type": "boolean"
},
"skill_env_var_dependency_prompt": {
"type": "boolean"
},
@@ -616,6 +613,10 @@
"additionalProperties": false,
"description": "Memories settings loaded from config.toml.",
"properties": {
"generate_memories": {
"description": "When `false`, newly created threads are stored with `memory_mode = \"disabled\"` in the state DB.",
"type": "boolean"
},
"max_raw_memories_for_global": {
"description": "Maximum number of recent raw memories retained for global consolidation.",
"format": "uint",
@@ -633,6 +634,11 @@
"minimum": 0.0,
"type": "integer"
},
"max_unused_days": {
"description": "Maximum number of days since a memory was last used before it becomes ineligible for phase 2 selection.",
"format": "int64",
"type": "integer"
},
"min_rollout_idle_hours": {
"description": "Minimum idle time between last thread activity and memory creation (hours). > 12h recommended.",
"format": "int64",
@@ -645,6 +651,10 @@
"phase_2_model": {
"description": "Model used for memory consolidation.",
"type": "string"
},
"use_memories": {
"description": "When `false`, skip injecting memory usage instructions into developer prompts.",
"type": "boolean"
}
},
"type": "object"
@@ -1144,6 +1154,10 @@
},
"type": "object"
},
"oauth_resource": {
"default": null,
"type": "string"
},
"required": {
"default": null,
"type": "boolean"
@@ -1177,6 +1191,18 @@
},
"type": "object"
},
"RealtimeAudioToml": {
"additionalProperties": false,
"properties": {
"microphone": {
"type": "string"
},
"speaker": {
"type": "string"
}
},
"type": "object"
},
"ReasoningEffort": {
"description": "See https://platform.openai.com/docs/guides/reasoning?api-mode=responses#get-started-with-reasoning",
"enum": [
@@ -1538,6 +1564,15 @@
"default": null,
"description": "Settings for app-specific controls."
},
"audio": {
"allOf": [
{
"$ref": "#/definitions/RealtimeAudioToml"
}
],
"default": null,
"description": "Machine-local realtime audio device preferences used by realtime voice."
},
"background_terminal_max_timeout": {
"description": "Maximum poll window for background terminal output (`write_stdin`), in milliseconds. Default: `300000` (5 minutes).",
"format": "uint64",
@@ -1705,9 +1740,6 @@
"shell_zsh_fork": {
"type": "boolean"
},
"skill_approval": {
"type": "boolean"
},
"skill_env_var_dependency_prompt": {
"type": "boolean"
},
codex-rs/core/src/agent/agent_names.txt
+100 -87
View File
@@ -1,87 +1,100 @@
Ash
Elm
Yew
Fir
Oak
Pine
Spruce
Cedar
Birch
Maple
Beech
Alder
Willow
Poplar
Aspen
Larch
Juniper
Cypress
Hemlock
Sequoia
Baobab
Acacia
Banyan
Teak
Olive
Cherry
Apple
Pear
Plum
Peach
Mango
Guava
Fig
Date
Palm
Holly
Ivy
Fern
Moss
Reed
Cane
Rush
Sage
Mint
Thyme
Basil
Dill
Fennel
Parsley
Laurel
Myrtle
Lotus
Tulip
Rose
Daisy
Lily
Iris
Poppy
Aster
Peony
Camellia
Azalea
Begonia
Petunia
Verbena
Marigold
Daffodil
Orchid
Violet
Clover
Bamboo
Eucalypt
Oleander
Yucca
Agave
Aloe
Cactus
Sedum
Boxwood
Hawthorn
Redwood
Ironwood
Sandal
Kapok
Mahonia
Spirea
Wisteria
Euclid
Archimedes
Ptolemy
Hypatia
Avicenna
Averroes
Aquinas
Copernicus
Kepler
Galileo
Bacon
Descartes
Pascal
Fermat
Huygens
Leibniz
Newton
Halley
Euler
Lagrange
Laplace
Volta
Gauss
Ampere
Faraday
Darwin
Lovelace
Boole
Pasteur
Maxwell
Mendel
Curie
Planck
Tesla
Poincare
Noether
Hilbert
Einstein
Raman
Bohr
Turing
Hubble
Feynman
Franklin
McClintock
Meitner
Herschel
Linnaeus
Wegener
Chandrasekhar
Sagan
Goodall
Carson
Carver
Socrates
Plato
Aristotle
Epicurus
Cicero
Confucius
Mencius
Zeno
Locke
Hume
Kant
Hegel
Kierkegaard
Mill
Nietzsche
Peirce
James
Dewey
Russell
Popper
Sartre
Beauvoir
Arendt
Rawls
Singer
Anscombe
Parfit
Kuhn
Boyle
Hooke
Harvey
Dalton
Ohm
Helmholtz
Gibbs
Lorentz
Schrodinger
Heisenberg
Pauli
Dirac
Bernoulli
Godel
Nash
Banach
Ramanujan
Erdos
codex-rs/core/src/agent/control.rs
+378 -4
View File
@@ -4,11 +4,17 @@ use crate::agent::status::is_final;
use crate::error::CodexErr;
use crate::error::Result as CodexResult;
use crate::find_thread_path_by_id_str;
use crate::rollout::RolloutRecorder;
use crate::session_prefix::format_subagent_context_line;
use crate::session_prefix::format_subagent_notification_message;
use crate::state_db;
use crate::thread_manager::ThreadManagerState;
use codex_protocol::ThreadId;
use codex_protocol::models::FunctionCallOutputPayload;
use codex_protocol::models::ResponseItem;
use codex_protocol::protocol::InitialHistory;
use codex_protocol::protocol::Op;
use codex_protocol::protocol::RolloutItem;
use codex_protocol::protocol::SessionSource;
use codex_protocol::protocol::SubAgentSource;
use codex_protocol::protocol::TokenUsage;
@@ -18,6 +24,12 @@ use std::sync::Weak;
use tokio::sync::watch;
const AGENT_NAMES: &str = include_str!("agent_names.txt");
const FORKED_SPAWN_AGENT_OUTPUT_MESSAGE: &str = "You are the newly spawned agent. The prior conversation history was forked from your parent agent. Treat the next user message as your new task, and use the forked history only as background context.";
#[derive(Clone, Debug, Default)]
pub(crate) struct SpawnAgentOptions {
pub(crate) fork_parent_spawn_call_id: Option<String>,
}
fn agent_nickname_list() -> Vec<&'static str> {
AGENT_NAMES
@@ -57,6 +69,17 @@ impl AgentControl {
config: crate::config::Config,
items: Vec<UserInput>,
session_source: Option<SessionSource>,
) -> CodexResult<ThreadId> {
self.spawn_agent_with_options(config, items, session_source, SpawnAgentOptions::default())
.await
}
pub(crate) async fn spawn_agent_with_options(
&self,
config: crate::config::Config,
items: Vec<UserInput>,
session_source: Option<SessionSource>,
options: SpawnAgentOptions,
) -> CodexResult<ThreadId> {
let state = self.upgrade()?;
let mut reservation = self.state.reserve_spawn_slot(config.agent_max_threads)?;
@@ -82,9 +105,75 @@ impl AgentControl {
// The same `AgentControl` is sent to spawn the thread.
let new_thread = match session_source {
Some(session_source) => {
state
.spawn_new_thread_with_source(config, self.clone(), session_source, false, None)
.await?
if let Some(call_id) = options.fork_parent_spawn_call_id.as_ref() {
let SessionSource::SubAgent(SubAgentSource::ThreadSpawn {
parent_thread_id,
..
}) = session_source.clone()
else {
return Err(CodexErr::Fatal(
"spawn_agent fork requires a thread-spawn session source".to_string(),
));
};
let parent_thread = state.get_thread(parent_thread_id).await.ok();
if let Some(parent_thread) = parent_thread.as_ref() {
// `record_conversation_items` only queues rollout writes asynchronously.
// Flush/materialize the live parent before snapshotting JSONL for a fork.
parent_thread
.codex
.session
.ensure_rollout_materialized()
.await;
parent_thread.codex.session.flush_rollout().await;
}
let rollout_path = parent_thread
.as_ref()
.and_then(|parent_thread| parent_thread.rollout_path())
.or(find_thread_path_by_id_str(
config.codex_home.as_path(),
&parent_thread_id.to_string(),
)
.await?)
.ok_or_else(|| {
CodexErr::Fatal(format!(
"parent thread rollout unavailable for fork: {parent_thread_id}"
))
})?;
let mut forked_rollout_items =
RolloutRecorder::get_rollout_history(&rollout_path)
.await?
.get_rollout_items();
let mut output = FunctionCallOutputPayload::from_text(
FORKED_SPAWN_AGENT_OUTPUT_MESSAGE.to_string(),
);
output.success = Some(true);
forked_rollout_items.push(RolloutItem::ResponseItem(
ResponseItem::FunctionCallOutput {
call_id: call_id.clone(),
output,
},
));
let initial_history = InitialHistory::Forked(forked_rollout_items);
state
.fork_thread_with_source(
config,
initial_history,
self.clone(),
session_source,
false,
)
.await?
} else {
state
.spawn_new_thread_with_source(
config,
self.clone(),
session_source,
false,
None,
)
.await?
}
}
None => state.spawn_new_thread(config, self.clone()).await?,
};
@@ -255,6 +344,40 @@ impl AgentControl {
thread.total_token_usage().await
}
pub(crate) async fn format_environment_context_subagents(
&self,
parent_thread_id: ThreadId,
) -> String {
let Ok(state) = self.upgrade() else {
return String::new();
};
let mut agents = Vec::new();
for thread_id in state.list_thread_ids().await {
let Ok(thread) = state.get_thread(thread_id).await else {
continue;
};
let snapshot = thread.config_snapshot().await;
let SessionSource::SubAgent(SubAgentSource::ThreadSpawn {
parent_thread_id: agent_parent_thread_id,
agent_nickname,
..
}) = snapshot.session_source
else {
continue;
};
if agent_parent_thread_id != parent_thread_id {
continue;
}
agents.push(format_subagent_context_line(
&thread_id.to_string(),
agent_nickname.as_deref(),
));
}
agents.sort();
agents.join("\n")
}
/// Starts a detached watcher for sub-agents spawned from another thread.
///
/// This is only enabled for `SubAgentSource::ThreadSpawn`, where a parent thread exists and
@@ -319,8 +442,8 @@ mod tests {
use crate::config::Config;
use crate::config::ConfigBuilder;
use crate::config_loader::LoaderOverrides;
use crate::contextual_user_message::SUBAGENT_NOTIFICATION_OPEN_TAG;
use crate::features::Feature;
use crate::session_prefix::SUBAGENT_NOTIFICATION_OPEN_TAG;
use assert_matches::assert_matches;
use codex_protocol::config_types::ModeKind;
use codex_protocol::models::ContentItem;
@@ -421,6 +544,21 @@ mod tests {
})
}
/// Returns true when any message item contains `needle` in a text span.
fn history_contains_text(history_items: &[ResponseItem], needle: &str) -> bool {
history_items.iter().any(|item| {
let ResponseItem::Message { content, .. } = item else {
return false;
};
content.iter().any(|content_item| match content_item {
ContentItem::InputText { text } | ContentItem::OutputText { text } => {
text.contains(needle)
}
ContentItem::InputImage { .. } => false,
})
})
}
async fn wait_for_subagent_notification(parent_thread: &Arc<CodexThread>) -> bool {
let wait = async {
loop {
@@ -673,6 +811,242 @@ mod tests {
assert_eq!(captured, Some(expected));
}
#[tokio::test]
async fn spawn_agent_can_fork_parent_thread_history() {
let harness = AgentControlHarness::new().await;
let (parent_thread_id, parent_thread) = harness.start_thread().await;
parent_thread
.inject_user_message_without_turn("parent seed context".to_string())
.await;
let turn_context = parent_thread.codex.session.new_default_turn().await;
let parent_spawn_call_id = "spawn-call-history".to_string();
let parent_spawn_call = ResponseItem::FunctionCall {
id: None,
name: "spawn_agent".to_string(),
arguments: "{}".to_string(),
call_id: parent_spawn_call_id.clone(),
};
parent_thread
.codex
.session
.record_conversation_items(turn_context.as_ref(), &[parent_spawn_call])
.await;
parent_thread
.codex
.session
.ensure_rollout_materialized()
.await;
parent_thread.codex.session.flush_rollout().await;
let child_thread_id = harness
.control
.spawn_agent_with_options(
harness.config.clone(),
text_input("child task"),
Some(SessionSource::SubAgent(SubAgentSource::ThreadSpawn {
parent_thread_id,
depth: 1,
agent_nickname: None,
agent_role: None,
})),
SpawnAgentOptions {
fork_parent_spawn_call_id: Some(parent_spawn_call_id),
},
)
.await
.expect("forked spawn should succeed");
let child_thread = harness
.manager
.get_thread(child_thread_id)
.await
.expect("child thread should be registered");
assert_ne!(child_thread_id, parent_thread_id);
let history = child_thread.codex.session.clone_history().await;
assert!(history_contains_text(
history.raw_items(),
"parent seed context"
));
let expected = (
child_thread_id,
Op::UserInput {
items: vec![UserInput::Text {
text: "child task".to_string(),
text_elements: Vec::new(),
}],
final_output_json_schema: None,
},
);
let captured = harness
.manager
.captured_ops()
.into_iter()
.find(|entry| *entry == expected);
assert_eq!(captured, Some(expected));
let _ = harness
.control
.shutdown_agent(child_thread_id)
.await
.expect("child shutdown should submit");
let _ = parent_thread
.submit(Op::Shutdown {})
.await
.expect("parent shutdown should submit");
}
#[tokio::test]
async fn spawn_agent_fork_injects_output_for_parent_spawn_call() {
let harness = AgentControlHarness::new().await;
let (parent_thread_id, parent_thread) = harness.start_thread().await;
let turn_context = parent_thread.codex.session.new_default_turn().await;
let parent_spawn_call_id = "spawn-call-1".to_string();
let parent_spawn_call = ResponseItem::FunctionCall {
id: None,
name: "spawn_agent".to_string(),
arguments: "{}".to_string(),
call_id: parent_spawn_call_id.clone(),
};
parent_thread
.codex
.session
.record_conversation_items(turn_context.as_ref(), &[parent_spawn_call])
.await;
parent_thread
.codex
.session
.ensure_rollout_materialized()
.await;
parent_thread.codex.session.flush_rollout().await;
let child_thread_id = harness
.control
.spawn_agent_with_options(
harness.config.clone(),
text_input("child task"),
Some(SessionSource::SubAgent(SubAgentSource::ThreadSpawn {
parent_thread_id,
depth: 1,
agent_nickname: None,
agent_role: None,
})),
SpawnAgentOptions {
fork_parent_spawn_call_id: Some(parent_spawn_call_id.clone()),
},
)
.await
.expect("forked spawn should succeed");
let child_thread = harness
.manager
.get_thread(child_thread_id)
.await
.expect("child thread should be registered");
let history = child_thread.codex.session.clone_history().await;
let injected_output = history.raw_items().iter().find_map(|item| match item {
ResponseItem::FunctionCallOutput { call_id, output }
if call_id == &parent_spawn_call_id =>
{
Some(output)
}
_ => None,
});
let injected_output =
injected_output.expect("forked child should contain synthetic tool output");
assert_eq!(
injected_output.text_content(),
Some(FORKED_SPAWN_AGENT_OUTPUT_MESSAGE)
);
assert_eq!(injected_output.success, Some(true));
let _ = harness
.control
.shutdown_agent(child_thread_id)
.await
.expect("child shutdown should submit");
let _ = parent_thread
.submit(Op::Shutdown {})
.await
.expect("parent shutdown should submit");
}
#[tokio::test]
async fn spawn_agent_fork_flushes_parent_rollout_before_loading_history() {
let harness = AgentControlHarness::new().await;
let (parent_thread_id, parent_thread) = harness.start_thread().await;
let turn_context = parent_thread.codex.session.new_default_turn().await;
let parent_spawn_call_id = "spawn-call-unflushed".to_string();
let parent_spawn_call = ResponseItem::FunctionCall {
id: None,
name: "spawn_agent".to_string(),
arguments: "{}".to_string(),
call_id: parent_spawn_call_id.clone(),
};
parent_thread
.codex
.session
.record_conversation_items(turn_context.as_ref(), &[parent_spawn_call])
.await;
let child_thread_id = harness
.control
.spawn_agent_with_options(
harness.config.clone(),
text_input("child task"),
Some(SessionSource::SubAgent(SubAgentSource::ThreadSpawn {
parent_thread_id,
depth: 1,
agent_nickname: None,
agent_role: None,
})),
SpawnAgentOptions {
fork_parent_spawn_call_id: Some(parent_spawn_call_id.clone()),
},
)
.await
.expect("forked spawn should flush parent rollout before loading history");
let child_thread = harness
.manager
.get_thread(child_thread_id)
.await
.expect("child thread should be registered");
let history = child_thread.codex.session.clone_history().await;
let mut parent_call_index = None;
let mut injected_output_index = None;
for (idx, item) in history.raw_items().iter().enumerate() {
match item {
ResponseItem::FunctionCall { call_id, .. } if call_id == &parent_spawn_call_id => {
parent_call_index = Some(idx);
}
ResponseItem::FunctionCallOutput { call_id, .. }
if call_id == &parent_spawn_call_id =>
{
injected_output_index = Some(idx);
}
_ => {}
}
}
let parent_call_index =
parent_call_index.expect("forked child should include the parent spawn_agent call");
let injected_output_index = injected_output_index
.expect("forked child should include synthetic output for the parent spawn_agent call");
assert!(parent_call_index < injected_output_index);
let _ = harness
.control
.shutdown_agent(child_thread_id)
.await
.expect("child shutdown should submit");
let _ = parent_thread
.submit(Op::Shutdown {})
.await
.expect("parent shutdown should submit");
}
#[tokio::test]
async fn spawn_agent_respects_max_threads_limit() {
let max_threads = 1usize;
codex-rs/core/src/agent/role.rs
+18 -13
View File
@@ -187,19 +187,24 @@ Rules:
config_file: None,
}
),
(
"awaiter".to_string(),
AgentRoleConfig {
description: Some(r#"Use an `awaiter` agent EVERY TIME you must run a command that might take some time.
This includes, but not only:
* testing
* monitoring of a long running process
* explicit ask to wait for something
When YOU wait for the `awaiter` agent to be done, use the largest possible timeout."#.to_string()),
config_file: Some("awaiter.toml".to_string().parse().unwrap_or_default()),
}
)
// Awaiter is temp removed
// (
// "awaiter".to_string(),
// AgentRoleConfig {
// description: Some(r#"Use an `awaiter` agent EVERY TIME you must run a command that will take some very long time.
// This includes, but not only:
// * testing
// * monitoring of a long running process
// * explicit ask to wait for something
//
// Rules:
// - When an awaiter is running, you can work on something else. If you need to wait for its completion, use the largest possible timeout.
// - Be patient with the `awaiter`.
// - Do not use an awaiter for every compilation/test if it won't take time. Only use if for long running commands.
// - Close the awaiter when you're done with it."#.to_string()),
// config_file: Some("awaiter.toml".to_string().parse().unwrap_or_default()),
// }
// )
])
});
&CONFIG
codex-rs/core/src/client_common.rs
+61 -8
View File
@@ -84,19 +84,13 @@ fn reserialize_shell_outputs(items: &mut [ResponseItem]) {
shell_call_ids.insert(call_id.clone());
}
}
ResponseItem::CustomToolCallOutput { call_id, output } => {
if shell_call_ids.remove(call_id)
&& let Some(structured) = parse_structured_shell_output(output)
{
*output = structured
}
}
ResponseItem::FunctionCall { name, call_id, .. }
if is_shell_tool_name(name) || name == "apply_patch" =>
{
shell_call_ids.insert(call_id.clone());
}
ResponseItem::FunctionCallOutput { call_id, output } => {
ResponseItem::FunctionCallOutput { call_id, output }
| ResponseItem::CustomToolCallOutput { call_id, output } => {
if shell_call_ids.remove(call_id)
&& let Some(structured) = output
.text_content()
@@ -240,6 +234,7 @@ mod tests {
use codex_api::common::OpenAiVerbosity;
use codex_api::common::TextControls;
use codex_api::create_text_param_for_request;
use codex_protocol::models::FunctionCallOutputPayload;
use pretty_assertions::assert_eq;
use super::*;
@@ -343,4 +338,62 @@ mod tests {
let v = serde_json::to_value(&req).expect("json");
assert!(v.get("text").is_none());
}
#[test]
fn reserializes_shell_outputs_for_function_and_custom_tool_calls() {
let raw_output = r#"{"output":"hello","metadata":{"exit_code":0,"duration_seconds":0.5}}"#;
let expected_output = "Exit code: 0\nWall time: 0.5 seconds\nOutput:\nhello";
let mut items = vec![
ResponseItem::FunctionCall {
id: None,
name: "shell".to_string(),
arguments: "{}".to_string(),
call_id: "call-1".to_string(),
},
ResponseItem::FunctionCallOutput {
call_id: "call-1".to_string(),
output: FunctionCallOutputPayload::from_text(raw_output.to_string()),
},
ResponseItem::CustomToolCall {
id: None,
status: None,
call_id: "call-2".to_string(),
name: "apply_patch".to_string(),
input: "*** Begin Patch".to_string(),
},
ResponseItem::CustomToolCallOutput {
call_id: "call-2".to_string(),
output: FunctionCallOutputPayload::from_text(raw_output.to_string()),
},
];
reserialize_shell_outputs(&mut items);
assert_eq!(
items,
vec![
ResponseItem::FunctionCall {
id: None,
name: "shell".to_string(),
arguments: "{}".to_string(),
call_id: "call-1".to_string(),
},
ResponseItem::FunctionCallOutput {
call_id: "call-1".to_string(),
output: FunctionCallOutputPayload::from_text(expected_output.to_string()),
},
ResponseItem::CustomToolCall {
id: None,
status: None,
call_id: "call-2".to_string(),
name: "apply_patch".to_string(),
input: "*** Begin Patch".to_string(),
},
ResponseItem::CustomToolCallOutput {
call_id: "call-2".to_string(),
output: FunctionCallOutputPayload::from_text(expected_output.to_string()),
},
]
);
}
}
codex-rs/core/src/codex.rs
+255 -149
View File
@@ -53,6 +53,8 @@ use crate::util::error_or_panic;
use crate::ws_version_from_features;
use async_channel::Receiver;
use async_channel::Sender;
use chrono::Local;
use chrono::Utc;
use codex_hooks::HookEvent;
use codex_hooks::HookEventAfterAgent;
use codex_hooks::HookPayload;
@@ -94,7 +96,6 @@ use codex_protocol::protocol::TurnContextNetworkItem;
use codex_protocol::protocol::TurnStartedEvent;
use codex_protocol::request_user_input::RequestUserInputArgs;
use codex_protocol::request_user_input::RequestUserInputResponse;
use codex_protocol::skill_approval::SkillApprovalResponse;
use codex_rmcp_client::ElicitationResponse;
use codex_rmcp_client::OAuthCredentialsStoreMode;
use codex_utils_stream_parser::AssistantTextChunk;
@@ -252,6 +253,7 @@ use crate::tools::ToolRouter;
use crate::tools::context::SharedTurnDiffTracker;
use crate::tools::handlers::SEARCH_TOOL_BM25_TOOL_NAME;
use crate::tools::js_repl::JsReplHandle;
use crate::tools::js_repl::resolve_compatible_node;
use crate::tools::network_approval::NetworkApprovalService;
use crate::tools::network_approval::build_blocked_request_observer;
use crate::tools::network_approval::build_network_policy_decider;
@@ -342,6 +344,18 @@ impl Codex {
config.features.disable(Feature::Collab);
}
if config.features.enabled(Feature::JsRepl)
&& let Err(err) = resolve_compatible_node(config.js_repl_node_path.as_deref()).await
{
let message = format!(
"Disabled `js_repl` for this session because the configured Node runtime is unavailable or incompatible. {err}"
);
warn!("{message}");
config.features.disable(Feature::JsRepl);
config.features.disable(Feature::JsReplToolsOnly);
config.startup_warnings.push(message);
}
let allowed_skills_for_implicit_invocation =
loaded_skills.allowed_skills_for_implicit_invocation();
let user_instructions =
@@ -429,6 +443,7 @@ impl Codex {
thread_name: None,
original_config_do_not_use: Arc::clone(&config),
metrics_service_name,
app_server_client_name: None,
session_source,
dynamic_tools,
persist_extended_history,
@@ -516,6 +531,18 @@ impl Codex {
self.session.steer_input(input, expected_turn_id).await
}
pub(crate) async fn set_app_server_client_name(
&self,
app_server_client_name: Option<String>,
) -> ConstraintResult<()> {
self.session
.update_settings(SessionSettingsUpdate {
app_server_client_name,
..Default::default()
})
.await
}
pub(crate) async fn agent_status(&self) -> AgentStatus {
self.agent_status.borrow().clone()
}
@@ -583,6 +610,9 @@ pub(crate) struct TurnContext {
/// the model as well as sandbox policies are resolved against this path
/// instead of `std::env::current_dir()`.
pub(crate) cwd: PathBuf,
pub(crate) current_date: Option<String>,
pub(crate) timezone: Option<String>,
pub(crate) app_server_client_name: Option<String>,
pub(crate) developer_instructions: Option<String>,
pub(crate) compact_prompt: Option<String>,
pub(crate) user_instructions: Option<String>,
@@ -667,6 +697,9 @@ impl TurnContext {
reasoning_summary: self.reasoning_summary,
session_source: self.session_source.clone(),
cwd: self.cwd.clone(),
current_date: self.current_date.clone(),
timezone: self.timezone.clone(),
app_server_client_name: self.app_server_client_name.clone(),
developer_instructions: self.developer_instructions.clone(),
compact_prompt: self.compact_prompt.clone(),
user_instructions: self.user_instructions.clone(),
@@ -707,6 +740,8 @@ impl TurnContext {
TurnContextItem {
turn_id: Some(self.sub_id.clone()),
cwd: self.cwd.clone(),
current_date: self.current_date.clone(),
timezone: self.timezone.clone(),
approval_policy: self.approval_policy.value(),
sandbox_policy: self.sandbox_policy.get().clone(),
network: self.turn_context_network_item(),
@@ -736,13 +771,23 @@ impl TurnContext {
}
}
fn local_time_context() -> (String, String) {
match iana_time_zone::get_timezone() {
Ok(timezone) => (Local::now().format("%Y-%m-%d").to_string(), timezone),
Err(_) => (
Utc::now().format("%Y-%m-%d").to_string(),
"Etc/UTC".to_string(),
),
}
}
#[derive(Clone)]
pub(crate) struct SessionConfiguration {
/// Provider identifier ("openai", "openrouter", ...).
provider: ModelProviderInfo,
collaboration_mode: CollaborationMode,
model_reasoning_summary: ReasoningSummaryConfig,
model_reasoning_summary: Option<ReasoningSummaryConfig>,
/// Developer instructions that supplement the base instructions.
developer_instructions: Option<String>,
@@ -782,6 +827,7 @@ pub(crate) struct SessionConfiguration {
original_config_do_not_use: Arc<Config>,
/// Optional service name tag for session metrics.
metrics_service_name: Option<String>,
app_server_client_name: Option<String>,
/// Source of the session (cli, vscode, exec, mcp, ...)
session_source: SessionSource,
dynamic_tools: Vec<DynamicToolSpec>,
@@ -812,7 +858,7 @@ impl SessionConfiguration {
next_configuration.collaboration_mode = collaboration_mode;
}
if let Some(summary) = updates.reasoning_summary {
next_configuration.model_reasoning_summary = summary;
next_configuration.model_reasoning_summary = Some(summary);
}
if let Some(personality) = updates.personality {
next_configuration.personality = Some(personality);
@@ -829,6 +875,9 @@ impl SessionConfiguration {
if let Some(cwd) = updates.cwd.clone() {
next_configuration.cwd = cwd;
}
if let Some(app_server_client_name) = updates.app_server_client_name.clone() {
next_configuration.app_server_client_name = Some(app_server_client_name);
}
Ok(next_configuration)
}
}
@@ -843,6 +892,7 @@ pub(crate) struct SessionSettingsUpdate {
pub(crate) reasoning_summary: Option<ReasoningSummaryConfig>,
pub(crate) final_output_json_schema: Option<Option<Value>>,
pub(crate) personality: Option<Personality>,
pub(crate) app_server_client_name: Option<String>,
}
impl Session {
@@ -973,7 +1023,9 @@ impl Session {
skills_outcome: Arc<SkillLoadOutcome>,
) -> TurnContext {
let reasoning_effort = session_configuration.collaboration_mode.reasoning_effort();
let reasoning_summary = session_configuration.model_reasoning_summary;
let reasoning_summary = session_configuration
.model_reasoning_summary
.unwrap_or(model_info.default_reasoning_summary);
let otel_manager = otel_manager.clone().with_model(
session_configuration.collaboration_mode.model(),
model_info.slug.as_str(),
@@ -1003,6 +1055,7 @@ impl Session {
.features
.enabled(Feature::UseLinuxSandboxBwrap),
));
let (current_date, timezone) = local_time_context();
TurnContext {
sub_id,
config: per_turn_config.clone(),
@@ -1014,6 +1067,9 @@ impl Session {
reasoning_summary,
session_source,
cwd,
current_date: Some(current_date),
timezone: Some(timezone),
app_server_client_name: session_configuration.app_server_client_name.clone(),
developer_instructions: session_configuration.developer_instructions.clone(),
compact_prompt: session_configuration.compact_prompt.clone(),
user_instructions: session_configuration.user_instructions.clone(),
@@ -1112,7 +1168,7 @@ impl Session {
//
// - initialize RolloutRecorder with new or resumed session info
// - perform default shell discovery
// - load history metadata
// - load history metadata (skipped for subagents)
let rollout_fut = async {
if config.ephemeral {
Ok::<_, anyhow::Error>((None, None))
@@ -1129,7 +1185,16 @@ impl Session {
}
};
let history_meta_fut = crate::message_history::history_metadata(&config);
let history_meta_fut = async {
if matches!(
session_configuration.session_source,
SessionSource::SubAgent(_)
) {
(0, 0)
} else {
crate::message_history::history_metadata(&config).await
}
};
let auth_manager_clone = Arc::clone(&auth_manager);
let config_for_mcp = Arc::clone(&config);
let auth_and_mcp_fut = async move {
@@ -1250,7 +1315,9 @@ impl Session {
otel_manager.conversation_starts(
config.model_provider.name.as_str(),
session_configuration.collaboration_mode.reasoning_effort(),
config.model_reasoning_summary,
config
.model_reasoning_summary
.unwrap_or(ReasoningSummaryConfig::Auto),
config.model_context_window,
config.model_auto_compact_token_limit,
config.permissions.approval_policy.value(),
@@ -1657,7 +1724,10 @@ impl Session {
match conversation_history {
InitialHistory::New => {
// Build and record initial items (user instructions + environment context)
let items = self.build_initial_context(&turn_context).await;
// TODO(ccunningham): Defer initial context insertion until the first real turn
// starts so it reflects the actual first-turn settings (permissions, etc.) and
// we do not emit model-visible "diff" updates before the first user message.
let items = self.build_initial_context(&turn_context, None).await;
self.record_conversation_items(&turn_context, &items).await;
{
let mut state = self.state.lock().await;
@@ -1761,7 +1831,7 @@ impl Session {
}
// Append the current session's initial context after the reconstructed history.
let initial_context = self.build_initial_context(&turn_context).await;
let initial_context = self.build_initial_context(&turn_context, None).await;
self.record_conversation_items(&turn_context, &initial_context)
.await;
{
@@ -2590,9 +2660,13 @@ impl Session {
/// Emit an exec approval request event and await the user's decision.
///
/// The request is keyed by `call_id` + `approval_id` so matching responses are delivered
/// to the correct in-flight turn. If the task is aborted, this returns the
/// default `ReviewDecision` (`Denied`).
/// The request is keyed by `call_id` + `approval_id` so matching responses
/// are delivered to the correct in-flight turn. If the task is aborted,
/// this returns the default `ReviewDecision` (`Denied`).
///
/// Note that if `available_decisions` is `None`, then the other fields will
/// be used to derive the available decisions via
/// [ExecApprovalRequestEvent::default_available_decisions].
#[allow(clippy::too_many_arguments)]
pub async fn request_command_approval(
&self,
@@ -2605,6 +2679,7 @@ impl Session {
network_approval_context: Option<NetworkApprovalContext>,
proposed_execpolicy_amendment: Option<ExecPolicyAmendment>,
additional_permissions: Option<PermissionProfile>,
available_decisions: Option<Vec<ReviewDecision>>,
) -> ReviewDecision {
// command-level approvals use `call_id`.
// `approval_id` is only present for subcommand callbacks (execve intercept)
@@ -2638,6 +2713,14 @@ impl Session {
},
]
});
let available_decisions = available_decisions.unwrap_or_else(|| {
ExecApprovalRequestEvent::default_available_decisions(
network_approval_context.as_ref(),
proposed_execpolicy_amendment.as_ref(),
proposed_network_policy_amendments.as_deref(),
additional_permissions.as_ref(),
)
});
let event = EventMsg::ExecApprovalRequest(ExecApprovalRequestEvent {
call_id,
approval_id,
@@ -2649,6 +2732,7 @@ impl Session {
proposed_execpolicy_amendment,
proposed_network_policy_amendments,
additional_permissions,
available_decisions: Some(available_decisions),
parsed_cmd,
});
self.send_event(turn_context, event).await;
@@ -2723,40 +2807,6 @@ impl Session {
rx_response.await.ok()
}
pub async fn request_skill_approval(
&self,
turn_context: &TurnContext,
item_id: String,
skill_name: String,
) -> Option<SkillApprovalResponse> {
let (tx_response, rx_response) = oneshot::channel();
let prev_entry = {
let mut active = self.active_turn.lock().await;
match active.as_mut() {
Some(at) => {
let mut ts = at.turn_state.lock().await;
ts.insert_pending_skill_approval(item_id.clone(), tx_response)
}
None => None,
}
};
if prev_entry.is_some() {
warn!("Overwriting existing pending skill approval for item_id: {item_id}");
}
self.send_event(
turn_context,
EventMsg::SkillRequestApproval(
codex_protocol::skill_approval::SkillRequestApprovalEvent {
item_id,
skill_name,
},
),
)
.await;
rx_response.await.ok()
}
pub async fn notify_user_input_response(
&self,
sub_id: &str,
@@ -2803,31 +2853,6 @@ impl Session {
}
}
pub async fn notify_skill_approval_response(
&self,
item_id: &str,
response: SkillApprovalResponse,
) {
let entry = {
let mut active = self.active_turn.lock().await;
match active.as_mut() {
Some(at) => {
let mut ts = at.turn_state.lock().await;
ts.remove_pending_skill_approval(item_id)
}
None => None,
}
};
match entry {
Some(tx_response) => {
tx_response.send(response).ok();
}
None => {
warn!("No pending skill approval found for item_id: {item_id}");
}
}
}
pub async fn notify_approval(&self, approval_id: &str, decision: ReviewDecision) {
let entry = {
let mut active = self.active_turn.lock().await;
@@ -2895,7 +2920,7 @@ impl Session {
} else {
let user_messages = collect_user_messages(history.raw_items());
let rebuilt = compact::build_compacted_history(
self.build_initial_context(turn_context).await,
self.build_initial_context(turn_context, None).await,
&user_messages,
&compacted.message,
);
@@ -3023,10 +3048,20 @@ impl Session {
pub(crate) async fn build_initial_context(
&self,
turn_context: &TurnContext,
previous_user_turn_model: Option<&str>,
) -> Vec<ResponseItem> {
let mut items = Vec::<ResponseItem>::with_capacity(4);
let mut developer_sections = Vec::<String>::with_capacity(8);
let mut contextual_user_sections = Vec::<String>::with_capacity(2);
let shell = self.user_shell();
items.push(
if let Some(model_switch_message) =
crate::context_manager::updates::build_model_instructions_update_item(
previous_user_turn_model,
turn_context,
)
{
developer_sections.push(model_switch_message.into_text());
}
developer_sections.push(
DeveloperInstructions::from_policy(
turn_context.sandbox_policy.get(),
turn_context.approval_policy.value(),
@@ -3034,17 +3069,18 @@ impl Session {
&turn_context.cwd,
turn_context.features.enabled(Feature::RequestPermissions),
)
.into(),
.into_text(),
);
if let Some(developer_instructions) = turn_context.developer_instructions.as_deref() {
items.push(DeveloperInstructions::new(developer_instructions.to_string()).into());
developer_sections.push(developer_instructions.to_string());
}
// Add developer instructions for memories.
if let Some(memory_prompt) =
build_memory_tool_developer_instructions(&turn_context.config.codex_home).await
&& turn_context.features.enabled(Feature::MemoryTool)
if turn_context.features.enabled(Feature::MemoryTool)
&& turn_context.config.memories.use_memories
&& let Some(memory_prompt) =
build_memory_tool_developer_instructions(&turn_context.config.codex_home).await
{
items.push(DeveloperInstructions::new(memory_prompt).into());
developer_sections.push(memory_prompt);
}
// Add developer instructions from collaboration_mode if they exist and are non-empty
let (collaboration_mode, base_instructions) = {
@@ -3057,7 +3093,7 @@ impl Session {
if let Some(collab_instructions) =
DeveloperInstructions::from_collaboration_mode(&collaboration_mode)
{
items.push(collab_instructions.into());
developer_sections.push(collab_instructions.into_text());
}
if self.features.enabled(Feature::Personality)
&& let Some(personality) = turn_context.personality
@@ -3072,34 +3108,53 @@ impl Session {
personality,
)
{
items.push(
DeveloperInstructions::personality_spec_message(personality_message).into(),
developer_sections.push(
DeveloperInstructions::personality_spec_message(personality_message)
.into_text(),
);
}
}
if turn_context.features.enabled(Feature::Apps) {
items.push(DeveloperInstructions::new(render_apps_section()).into());
developer_sections.push(render_apps_section());
}
if turn_context.features.enabled(Feature::CodexGitCommit)
&& let Some(commit_message_instruction) = commit_message_trailer_instruction(
turn_context.config.commit_attribution.as_deref(),
)
{
items.push(DeveloperInstructions::new(commit_message_instruction).into());
developer_sections.push(commit_message_instruction);
}
if let Some(user_instructions) = turn_context.user_instructions.as_deref() {
items.push(
contextual_user_sections.push(
UserInstructions {
text: user_instructions.to_string(),
directory: turn_context.cwd.to_string_lossy().into_owned(),
}
.into(),
.serialize_to_text(),
);
}
items.push(ResponseItem::from(EnvironmentContext::from_turn_context(
turn_context,
shell.as_ref(),
)));
let subagents = self
.services
.agent_control
.format_environment_context_subagents(self.conversation_id)
.await;
contextual_user_sections.push(
EnvironmentContext::from_turn_context(turn_context, shell.as_ref())
.with_subagents(subagents)
.serialize_to_xml(),
);
let mut items = Vec::with_capacity(2);
if let Some(developer_message) =
crate::context_manager::updates::build_developer_update_item(developer_sections)
{
items.push(developer_message);
}
if let Some(contextual_user_message) =
crate::context_manager::updates::build_contextual_user_message(contextual_user_sections)
{
items.push(contextual_user_message);
}
items
}
@@ -3144,22 +3199,8 @@ impl Session {
let reference_context_item = self.reference_context_item().await;
let should_inject_full_context = reference_context_item.is_none();
let context_items = if should_inject_full_context {
let mut initial_context = self.build_initial_context(turn_context).await;
// Full reinjection bypasses the settings-diff path, so add the model-switch
// instruction explicitly when needed. Keep it before the rest of full context so
// model-specific guidance is read first.
if let Some(model_switch_item) =
crate::context_manager::updates::build_model_instructions_update_item(
previous_user_turn_model,
turn_context,
)
{
// TODO(ccunningham): When a model switch changes the effective personality
// instructions, inject the updated personality spec alongside <model_switch>
// here so resume/model-switch paths can avoid forcing full reinjection.
initial_context.insert(0, model_switch_item);
}
initial_context
self.build_initial_context(turn_context, previous_user_turn_model)
.await
} else {
// Steady-state path: append only context diffs to minimize token overhead.
self.build_settings_update_items(
@@ -3549,9 +3590,9 @@ impl Session {
use_linux_sandbox_bwrap: turn_context.features.enabled(Feature::UseLinuxSandboxBwrap),
};
{
let mut cancel_guard = self.services.mcp_startup_cancellation_token.lock().await;
cancel_guard.cancel();
*cancel_guard = CancellationToken::new();
let mut guard = self.services.mcp_startup_cancellation_token.lock().await;
guard.cancel();
*guard = CancellationToken::new();
}
let (refreshed_manager, cancel_token) = McpConnectionManager::new(
&mcp_servers,
@@ -3565,11 +3606,11 @@ impl Session {
)
.await;
{
let mut cancel_guard = self.services.mcp_startup_cancellation_token.lock().await;
if cancel_guard.is_cancelled() {
let mut guard = self.services.mcp_startup_cancellation_token.lock().await;
if guard.is_cancelled() {
cancel_token.cancel();
}
*cancel_guard = cancel_token;
*guard = cancel_token;
}
let mut manager = self.services.mcp_connection_manager.write().await;
@@ -3731,9 +3772,6 @@ async fn submission_loop(sess: Arc<Session>, config: Arc<Config>, rx_sub: Receiv
Op::DynamicToolResponse { id, response } => {
handlers::dynamic_tool_response(&sess, id, response).await;
}
Op::SkillApproval { id, response } => {
handlers::skill_approval_response(&sess, id, response).await;
}
Op::AddToHistory { text } => {
handlers::add_to_history(&sess, &config, text).await;
}
@@ -3857,7 +3895,6 @@ mod handlers {
use codex_protocol::protocol::TurnAbortReason;
use codex_protocol::protocol::WarningEvent;
use codex_protocol::request_user_input::RequestUserInputResponse;
use codex_protocol::skill_approval::SkillApprovalResponse;
use crate::context_manager::is_user_turn_boundary;
use codex_protocol::config_types::CollaborationMode;
@@ -3930,9 +3967,10 @@ mod handlers {
sandbox_policy: Some(sandbox_policy),
windows_sandbox_level: None,
collaboration_mode,
reasoning_summary: Some(summary),
reasoning_summary: summary,
final_output_json_schema: Some(final_output_json_schema),
personality,
app_server_client_name: None,
},
)
}
@@ -4098,14 +4136,6 @@ mod handlers {
sess.notify_dynamic_tool_response(&id, response).await;
}
pub async fn skill_approval_response(
sess: &Arc<Session>,
id: String,
response: SkillApprovalResponse,
) {
sess.notify_skill_approval_response(&id, response).await;
}
pub async fn add_to_history(sess: &Arc<Session>, config: &Arc<Config>, text: String) {
let id = sess.conversation_id;
let config = Arc::clone(config);
@@ -4660,7 +4690,9 @@ async fn spawn_review_thread(
let provider_for_context = provider.clone();
let otel_manager_for_context = otel_manager.clone();
let reasoning_effort = per_turn_config.model_reasoning_effort;
let reasoning_summary = per_turn_config.model_reasoning_summary;
let reasoning_summary = per_turn_config
.model_reasoning_summary
.unwrap_or(model_info.default_reasoning_summary);
let session_source = parent_turn_context.session_source.clone();
let per_turn_config = Arc::new(per_turn_config);
@@ -4688,6 +4720,9 @@ async fn spawn_review_thread(
tools_config,
features: parent_turn_context.features.clone(),
ghost_snapshot: parent_turn_context.ghost_snapshot.clone(),
current_date: parent_turn_context.current_date.clone(),
timezone: parent_turn_context.timezone.clone(),
app_server_client_name: parent_turn_context.app_server_client_name.clone(),
developer_instructions: None,
user_instructions: None,
compact_prompt: parent_turn_context.compact_prompt.clone(),
@@ -4819,15 +4854,6 @@ pub(crate) async fn run_turn(
collaboration_mode_kind: turn_context.collaboration_mode.mode,
});
sess.send_event(&turn_context, event).await;
if turn_context.config.features.enabled(Feature::SkillApproval) {
let _ = sess
.request_skill_approval(
turn_context.as_ref(),
turn_context.sub_id.clone(),
"test-skill".to_string(),
)
.await;
}
// TODO(ccunningham): Pre-turn compaction runs before context updates and the
// new user message are recorded. Estimate pending incoming items (context
// diffs/full reinjection + user input) and trigger compaction preemptively
@@ -5065,6 +5091,7 @@ pub(crate) async fn run_turn(
&sess,
&turn_context,
InitialContextInjection::BeforeLastUserMessage,
previous_model.as_deref(),
)
.await
.is_err()
@@ -5081,6 +5108,7 @@ pub(crate) async fn run_turn(
.dispatch(HookPayload {
session_id: sess.conversation_id,
cwd: turn_context.cwd.clone(),
client: turn_context.app_server_client_name.clone(),
triggered_at: chrono::Utc::now(),
hook_event: HookEvent::AfterAgent {
event: HookEventAfterAgent {
@@ -5188,7 +5216,13 @@ async fn run_pre_sampling_compact(
.unwrap_or(i64::MAX);
// Compact if the total usage tokens are greater than the auto compact limit
if total_usage_tokens >= auto_compact_limit {
run_auto_compact(sess, turn_context, InitialContextInjection::DoNotInject).await?;
run_auto_compact(
sess,
turn_context,
InitialContextInjection::DoNotInject,
None,
)
.await?;
}
Ok(())
}
@@ -5231,6 +5265,7 @@ async fn maybe_run_previous_model_inline_compact(
sess,
&previous_model_turn_context,
InitialContextInjection::DoNotInject,
None,
)
.await?;
return Ok(true);
@@ -5242,12 +5277,14 @@ async fn run_auto_compact(
sess: &Arc<Session>,
turn_context: &Arc<TurnContext>,
initial_context_injection: InitialContextInjection,
previous_user_turn_model: Option<&str>,
) -> CodexResult<()> {
if should_use_remote_compact_task(&turn_context.provider) {
run_inline_remote_auto_compact_task(
Arc::clone(sess),
Arc::clone(turn_context),
initial_context_injection,
previous_user_turn_model,
)
.await?;
} else {
@@ -5255,6 +5292,7 @@ async fn run_auto_compact(
Arc::clone(sess),
Arc::clone(turn_context),
initial_context_injection,
previous_user_turn_model,
)
.await?;
}
@@ -5856,7 +5894,6 @@ fn realtime_text_for_event(msg: &EventMsg) -> Option<String> {
| EventMsg::RequestUserInput(_)
| EventMsg::DynamicToolCallRequest(_)
| EventMsg::DynamicToolCallResponse(_)
| EventMsg::SkillRequestApproval(_)
| EventMsg::ElicitationRequest(_)
| EventMsg::ApplyPatchApprovalRequest(_)
| EventMsg::DeprecationNotice(_)
@@ -7270,6 +7307,8 @@ mod tests {
let previous_context_item = TurnContextItem {
turn_id: Some(turn_context.sub_id.clone()),
cwd: turn_context.cwd.clone(),
current_date: turn_context.current_date.clone(),
timezone: turn_context.timezone.clone(),
approval_policy: turn_context.approval_policy.value(),
sandbox_policy: turn_context.sandbox_policy.get().clone(),
network: None,
@@ -7307,6 +7346,8 @@ mod tests {
let mut previous_context_item = TurnContextItem {
turn_id: Some(turn_context.sub_id.clone()),
cwd: turn_context.cwd.clone(),
current_date: turn_context.current_date.clone(),
timezone: turn_context.timezone.clone(),
approval_policy: turn_context.approval_policy.value(),
sandbox_policy: turn_context.sandbox_policy.get().clone(),
network: None,
@@ -7493,7 +7534,7 @@ mod tests {
session
.record_context_updates_and_set_reference_context_item(&turn_context, None)
.await;
expected.extend(session.build_initial_context(&turn_context).await);
expected.extend(session.build_initial_context(&turn_context, None).await);
let history_after_seed = session.clone_history().await;
assert_eq!(expected, history_after_seed.raw_items());
@@ -7501,7 +7542,10 @@ mod tests {
.record_context_updates_and_set_reference_context_item(&turn_context, None)
.await;
let history_after_second_seed = session.clone_history().await;
assert_eq!(expected, history_after_second_seed.raw_items());
assert_eq!(
history_after_seed.raw_items(),
history_after_second_seed.raw_items()
);
}
#[tokio::test]
@@ -7655,7 +7699,7 @@ mod tests {
let reconstruction_turn = session.new_default_turn().await;
expected.extend(
session
.build_initial_context(reconstruction_turn.as_ref())
.build_initial_context(reconstruction_turn.as_ref(), None)
.await,
);
let history = session.state.lock().await.clone_history();
@@ -7669,6 +7713,8 @@ mod tests {
let previous_context_item = TurnContextItem {
turn_id: Some(turn_context.sub_id.clone()),
cwd: turn_context.cwd.clone(),
current_date: turn_context.current_date.clone(),
timezone: turn_context.timezone.clone(),
approval_policy: turn_context.approval_policy.value(),
sandbox_policy: turn_context.sandbox_policy.get().clone(),
network: None,
@@ -7698,7 +7744,7 @@ mod tests {
async fn thread_rollback_drops_last_turn_from_history() {
let (sess, tc, rx) = make_session_and_context_with_rx().await;
let initial_context = sess.build_initial_context(tc.as_ref()).await;
let initial_context = sess.build_initial_context(tc.as_ref(), None).await;
sess.record_into_history(&initial_context, tc.as_ref())
.await;
@@ -7769,7 +7815,7 @@ mod tests {
async fn thread_rollback_clears_history_when_num_turns_exceeds_existing_turns() {
let (sess, tc, rx) = make_session_and_context_with_rx().await;
let initial_context = sess.build_initial_context(tc.as_ref()).await;
let initial_context = sess.build_initial_context(tc.as_ref(), None).await;
sess.record_into_history(&initial_context, tc.as_ref())
.await;
@@ -7797,7 +7843,7 @@ mod tests {
async fn thread_rollback_fails_when_turn_in_progress() {
let (sess, tc, rx) = make_session_and_context_with_rx().await;
let initial_context = sess.build_initial_context(tc.as_ref()).await;
let initial_context = sess.build_initial_context(tc.as_ref(), None).await;
sess.record_into_history(&initial_context, tc.as_ref())
.await;
@@ -7818,7 +7864,7 @@ mod tests {
async fn thread_rollback_fails_when_num_turns_is_zero() {
let (sess, tc, rx) = make_session_and_context_with_rx().await;
let initial_context = sess.build_initial_context(tc.as_ref()).await;
let initial_context = sess.build_initial_context(tc.as_ref(), None).await;
sess.record_into_history(&initial_context, tc.as_ref())
.await;
@@ -7872,6 +7918,7 @@ mod tests {
thread_name: None,
original_config_do_not_use: Arc::clone(&config),
metrics_service_name: None,
app_server_client_name: None,
session_source: SessionSource::Exec,
dynamic_tools: Vec::new(),
persist_extended_history: false,
@@ -7964,6 +8011,7 @@ mod tests {
thread_name: None,
original_config_do_not_use: Arc::clone(&config),
metrics_service_name: None,
app_server_client_name: None,
session_source: SessionSource::Exec,
dynamic_tools: Vec::new(),
persist_extended_history: false,
@@ -8275,6 +8323,7 @@ mod tests {
thread_name: None,
original_config_do_not_use: Arc::clone(&config),
metrics_service_name: None,
app_server_client_name: None,
session_source: SessionSource::Exec,
dynamic_tools: Vec::new(),
persist_extended_history: false,
@@ -8328,6 +8377,7 @@ mod tests {
thread_name: None,
original_config_do_not_use: Arc::clone(&config),
metrics_service_name: None,
app_server_client_name: None,
session_source: SessionSource::Exec,
dynamic_tools: Vec::new(),
persist_extended_history: false,
@@ -8409,6 +8459,7 @@ mod tests {
thread_name: None,
original_config_do_not_use: Arc::clone(&config),
metrics_service_name: None,
app_server_client_name: None,
session_source: SessionSource::Exec,
dynamic_tools: Vec::new(),
persist_extended_history: false,
@@ -8568,6 +8619,7 @@ mod tests {
thread_name: None,
original_config_do_not_use: Arc::clone(&config),
metrics_service_name: None,
app_server_client_name: None,
session_source: SessionSource::Exec,
dynamic_tools,
persist_extended_history: false,
@@ -8834,6 +8886,42 @@ mod tests {
assert!(environment_update.contains("<denied>blocked.example.com</denied>"));
}
#[tokio::test]
async fn build_settings_update_items_emits_environment_item_for_time_changes() {
let (session, previous_context) = make_session_and_context().await;
let previous_context = Arc::new(previous_context);
let mut current_context = previous_context
.with_model(
previous_context.model_info.slug.clone(),
&session.services.models_manager,
)
.await;
current_context.current_date = Some("2026-02-27".to_string());
current_context.timezone = Some("Europe/Berlin".to_string());
let reference_context_item = previous_context.to_turn_context_item();
let update_items = session.build_settings_update_items(
Some(&reference_context_item),
None,
&current_context,
);
let environment_update = update_items
.iter()
.find_map(|item| match item {
ResponseItem::Message { role, content, .. } if role == "user" => {
let [ContentItem::InputText { text }] = content.as_slice() else {
return None;
};
text.contains("<environment_context>").then_some(text)
}
_ => None,
})
.expect("environment update item should be emitted");
assert!(environment_update.contains("<current_date>2026-02-27</current_date>"));
assert!(environment_update.contains("<timezone>Europe/Berlin</timezone>"));
}
#[tokio::test]
async fn record_context_updates_and_set_reference_context_item_injects_full_context_when_baseline_missing()
{
@@ -8842,7 +8930,7 @@ mod tests {
.record_context_updates_and_set_reference_context_item(&turn_context, None)
.await;
let history = session.clone_history().await;
let initial_context = session.build_initial_context(&turn_context).await;
let initial_context = session.build_initial_context(&turn_context, None).await;
assert_eq!(history.raw_items().to_vec(), initial_context);
let current_context = session.reference_context_item().await;
@@ -8886,10 +8974,28 @@ mod tests {
let history = session.clone_history().await;
let mut expected_history = vec![compacted_summary];
expected_history.extend(session.build_initial_context(&turn_context).await);
expected_history.extend(session.build_initial_context(&turn_context, None).await);
assert_eq!(history.raw_items().to_vec(), expected_history);
}
#[tokio::test]
async fn build_initial_context_prepends_model_switch_message() {
let (session, turn_context) = make_session_and_context().await;
let initial_context = session
.build_initial_context(&turn_context, Some("previous-regular-model"))
.await;
let ResponseItem::Message { role, content, .. } = &initial_context[0] else {
panic!("expected developer message");
};
assert_eq!(role, "developer");
let [ContentItem::InputText { text }, ..] = content.as_slice() else {
panic!("expected developer text");
};
assert!(text.contains("<model_switch>"));
}
#[tokio::test]
async fn run_user_shell_command_does_not_set_reference_context_item() {
let (session, _turn_context, rx) = make_session_and_context_with_rx().await;
@@ -9211,7 +9317,7 @@ mod tests {
let ContentItem::InputText { text } = content_item else {
return false;
};
text.contains(crate::session_prefix::TURN_ABORTED_OPEN_TAG)
text.contains(crate::contextual_user_message::TURN_ABORTED_OPEN_TAG)
})
}),
"expected a model-visible turn aborted marker in history after interrupt"
@@ -9285,7 +9391,7 @@ mod tests {
// personality_spec) matches reconstruction.
let reconstruction_turn = session.new_default_turn().await;
let mut initial_context = session
.build_initial_context(reconstruction_turn.as_ref())
.build_initial_context(reconstruction_turn.as_ref(), None)
.await;
// Ensure personality_spec is present when Personality is enabled, so expected matches
// what reconstruction produces (build_initial_context may omit it when baked into model).
codex-rs/core/src/codex_delegate.rs
+2
View File
@@ -322,6 +322,7 @@ async fn handle_exec_approval(
network_approval_context,
proposed_execpolicy_amendment,
additional_permissions,
available_decisions,
..
} = event;
// Race approval with cancellation and timeout to avoid hangs.
@@ -335,6 +336,7 @@ async fn handle_exec_approval(
network_approval_context,
proposed_execpolicy_amendment,
additional_permissions,
available_decisions,
);
let decision = await_approval_with_cancel(
approval_fut,
codex-rs/core/src/codex_thread.rs
+10
View File
@@ -1,6 +1,7 @@
use crate::agent::AgentStatus;
use crate::codex::Codex;
use crate::codex::SteerInputError;
use crate::config::ConstraintResult;
use crate::error::Result as CodexResult;
use crate::features::Feature;
use crate::file_watcher::WatchRegistration;
@@ -67,6 +68,15 @@ impl CodexThread {
self.codex.steer_input(input, expected_turn_id).await
}
pub async fn set_app_server_client_name(
&self,
app_server_client_name: Option<String>,
) -> ConstraintResult<()> {
self.codex
.set_app_server_client_name(app_server_client_name)
.await
}
/// Use sparingly: this is intended to be removed soon.
pub async fn submit_with_id(&self, sub: Submission) -> CodexResult<()> {
self.codex.submit_with_id(sub).await
codex-rs/core/src/compact.rs
+63 -155
View File
@@ -7,7 +7,6 @@ use crate::client_common::ResponseEvent;
use crate::codex::Session;
use crate::codex::TurnContext;
use crate::codex::get_last_assistant_message_from_turn;
use crate::context_manager::ContextManager;
use crate::error::CodexErr;
use crate::error::Result as CodexResult;
use crate::protocol::CompactedItem;
@@ -51,47 +50,11 @@ pub(crate) fn should_use_remote_compact_task(provider: &ModelProviderInfo) -> bo
provider.is_openai()
}
fn is_model_switch_developer_message(item: &ResponseItem) -> bool {
match item {
ResponseItem::Message { role, content, .. } if role == "developer" => {
matches!(
content.as_slice(),
[ContentItem::InputText { text }] if text.starts_with("<model_switch>\n")
)
}
_ => false,
}
}
pub(crate) fn extract_trailing_model_switch_update_for_compaction_request(
history: &mut ContextManager,
) -> Option<ResponseItem> {
let history_items = history.raw_items();
let last_user_turn_boundary_index = history_items
.iter()
.rposition(crate::context_manager::is_user_turn_boundary);
let model_switch_index = history_items
.iter()
.enumerate()
.rev()
.find_map(|(i, item)| {
let is_trailing = last_user_turn_boundary_index.is_none_or(|boundary| i > boundary);
if is_trailing && is_model_switch_developer_message(item) {
Some(i)
} else {
None
}
})?;
let mut replacement = history_items.to_vec();
let model_switch_item = replacement.remove(model_switch_index);
history.replace(replacement);
Some(model_switch_item)
}
pub(crate) async fn run_inline_auto_compact_task(
sess: Arc<Session>,
turn_context: Arc<TurnContext>,
initial_context_injection: InitialContextInjection,
previous_user_turn_model: Option<&str>,
) -> CodexResult<()> {
let prompt = turn_context.compact_prompt().to_string();
let input = vec![UserInput::Text {
@@ -100,7 +63,14 @@ pub(crate) async fn run_inline_auto_compact_task(
text_elements: Vec::new(),
}];
run_compact_task_inner(sess, turn_context, input, initial_context_injection).await?;
run_compact_task_inner(
sess,
turn_context,
input,
initial_context_injection,
previous_user_turn_model,
)
.await?;
Ok(())
}
@@ -120,6 +90,7 @@ pub(crate) async fn run_compact_task(
turn_context,
input,
InitialContextInjection::DoNotInject,
None,
)
.await
}
@@ -129,6 +100,7 @@ async fn run_compact_task_inner(
turn_context: Arc<TurnContext>,
input: Vec<UserInput>,
initial_context_injection: InitialContextInjection,
previous_user_turn_model: Option<&str>,
) -> CodexResult<()> {
let compaction_item = TurnItem::ContextCompaction(ContextCompactionItem::new());
sess.emit_turn_item_started(&turn_context, &compaction_item)
@@ -136,10 +108,6 @@ async fn run_compact_task_inner(
let initial_input_for_turn: ResponseInputItem = ResponseInputItem::from(input);
let mut history = sess.clone_history().await;
// Keep compaction prompts in-distribution: if a model-switch update was injected at the
// tail of history (between turns), exclude it from the compaction request payload.
let stripped_model_switch_item =
extract_trailing_model_switch_update_for_compaction_request(&mut history);
history.record_items(
&[initial_input_for_turn.into()],
turn_context.truncation_policy,
@@ -240,15 +208,12 @@ async fn run_compact_task_inner(
initial_context_injection,
InitialContextInjection::BeforeLastUserMessage
) {
let initial_context = sess.build_initial_context(turn_context.as_ref()).await;
let initial_context = sess
.build_initial_context(turn_context.as_ref(), previous_user_turn_model)
.await;
new_history =
insert_initial_context_before_last_real_user_or_summary(new_history, initial_context);
}
// Reattach the stripped model-switch update only after successful compaction so the model
// still sees the switch instructions on the next real sampling request.
if let Some(model_switch_item) = stripped_model_switch_item {
new_history.push(model_switch_item);
}
let ghost_snapshots: Vec<ResponseItem> = history_items
.iter()
.filter(|item| matches!(item, ResponseItem::GhostSnapshot { .. }))
@@ -491,14 +456,18 @@ mod tests {
async fn process_compacted_history_with_test_session(
compacted_history: Vec<ResponseItem>,
previous_user_turn_model: Option<&str>,
) -> (Vec<ResponseItem>, Vec<ResponseItem>) {
let (session, turn_context) = crate::codex::make_session_and_context().await;
let initial_context = session.build_initial_context(&turn_context).await;
let initial_context = session
.build_initial_context(&turn_context, previous_user_turn_model)
.await;
let refreshed = crate::compact_remote::process_compacted_history(
&session,
&turn_context,
compacted_history,
InitialContextInjection::BeforeLastUserMessage,
previous_user_turn_model,
)
.await;
(refreshed, initial_context)
@@ -534,107 +503,6 @@ mod tests {
assert_eq!(None, joined);
}
#[test]
fn extract_trailing_model_switch_update_for_compaction_request_removes_trailing_item() {
let mut history = ContextManager::new();
history.replace(vec![
ResponseItem::Message {
id: None,
role: "user".to_string(),
content: vec![ContentItem::InputText {
text: "USER_MESSAGE".to_string(),
}],
end_turn: None,
phase: None,
},
ResponseItem::Message {
id: None,
role: "assistant".to_string(),
content: vec![ContentItem::OutputText {
text: "ASSISTANT_REPLY".to_string(),
}],
end_turn: None,
phase: None,
},
ResponseItem::Message {
id: None,
role: "developer".to_string(),
content: vec![ContentItem::InputText {
text: "<model_switch>\nNEW_MODEL_INSTRUCTIONS".to_string(),
}],
end_turn: None,
phase: None,
},
]);
let model_switch_item =
extract_trailing_model_switch_update_for_compaction_request(&mut history);
assert_eq!(history.raw_items().len(), 2);
assert!(model_switch_item.is_some());
assert!(
history
.raw_items()
.iter()
.all(|item| !is_model_switch_developer_message(item))
);
}
#[test]
fn extract_trailing_model_switch_update_for_compaction_request_keeps_historical_item() {
let mut history = ContextManager::new();
history.replace(vec![
ResponseItem::Message {
id: None,
role: "user".to_string(),
content: vec![ContentItem::InputText {
text: "FIRST_USER_MESSAGE".to_string(),
}],
end_turn: None,
phase: None,
},
ResponseItem::Message {
id: None,
role: "developer".to_string(),
content: vec![ContentItem::InputText {
text: "<model_switch>\nOLDER_MODEL_INSTRUCTIONS".to_string(),
}],
end_turn: None,
phase: None,
},
ResponseItem::Message {
id: None,
role: "assistant".to_string(),
content: vec![ContentItem::OutputText {
text: "ASSISTANT_REPLY".to_string(),
}],
end_turn: None,
phase: None,
},
ResponseItem::Message {
id: None,
role: "user".to_string(),
content: vec![ContentItem::InputText {
text: "SECOND_USER_MESSAGE".to_string(),
}],
end_turn: None,
phase: None,
},
]);
let model_switch_item =
extract_trailing_model_switch_update_for_compaction_request(&mut history);
assert_eq!(history.raw_items().len(), 4);
assert!(model_switch_item.is_none());
assert!(
history
.raw_items()
.iter()
.any(is_model_switch_developer_message)
);
}
#[test]
fn collect_user_messages_extracts_user_text_only() {
let items = vec![
@@ -802,7 +670,7 @@ do things
},
];
let (refreshed, mut expected) =
process_compacted_history_with_test_session(compacted_history).await;
process_compacted_history_with_test_session(compacted_history, None).await;
expected.push(ResponseItem::Message {
id: None,
role: "user".to_string(),
@@ -827,7 +695,7 @@ do things
phase: None,
}];
let (refreshed, mut expected) =
process_compacted_history_with_test_session(compacted_history).await;
process_compacted_history_with_test_session(compacted_history, None).await;
expected.push(ResponseItem::Message {
id: None,
role: "user".to_string(),
@@ -903,7 +771,7 @@ keep me updated
},
];
let (refreshed, mut expected) =
process_compacted_history_with_test_session(compacted_history).await;
process_compacted_history_with_test_session(compacted_history, None).await;
expected.push(ResponseItem::Message {
id: None,
role: "user".to_string(),
@@ -949,7 +817,7 @@ keep me updated
];
let (refreshed, initial_context) =
process_compacted_history_with_test_session(compacted_history).await;
process_compacted_history_with_test_session(compacted_history, None).await;
let mut expected = vec![
ResponseItem::Message {
id: None,
@@ -983,6 +851,46 @@ keep me updated
assert_eq!(refreshed, expected);
}
#[tokio::test]
async fn process_compacted_history_reinjects_model_switch_message() {
let compacted_history = vec![ResponseItem::Message {
id: None,
role: "user".to_string(),
content: vec![ContentItem::InputText {
text: "summary".to_string(),
}],
end_turn: None,
phase: None,
}];
let (refreshed, initial_context) = process_compacted_history_with_test_session(
compacted_history,
Some("previous-regular-model"),
)
.await;
let ResponseItem::Message { role, content, .. } = &initial_context[0] else {
panic!("expected developer message");
};
assert_eq!(role, "developer");
let [ContentItem::InputText { text }, ..] = content.as_slice() else {
panic!("expected developer text");
};
assert!(text.contains("<model_switch>"));
let mut expected = initial_context;
expected.push(ResponseItem::Message {
id: None,
role: "user".to_string(),
content: vec![ContentItem::InputText {
text: "summary".to_string(),
}],
end_turn: None,
phase: None,
});
assert_eq!(refreshed, expected);
}
#[test]
fn insert_initial_context_before_last_real_user_or_summary_keeps_summary_last() {
let compacted_history = vec![
codex-rs/core/src/compact_remote.rs
+34 -20
View File
@@ -4,7 +4,6 @@ use crate::Prompt;
use crate::codex::Session;
use crate::codex::TurnContext;
use crate::compact::InitialContextInjection;
use crate::compact::extract_trailing_model_switch_update_for_compaction_request;
use crate::compact::insert_initial_context_before_last_real_user_or_summary;
use crate::context_manager::ContextManager;
use crate::context_manager::TotalTokenUsageBreakdown;
@@ -28,8 +27,15 @@ pub(crate) async fn run_inline_remote_auto_compact_task(
sess: Arc<Session>,
turn_context: Arc<TurnContext>,
initial_context_injection: InitialContextInjection,
previous_user_turn_model: Option<&str>,
) -> CodexResult<()> {
run_remote_compact_task_inner(&sess, &turn_context, initial_context_injection).await?;
run_remote_compact_task_inner(
&sess,
&turn_context,
initial_context_injection,
previous_user_turn_model,
)
.await?;
Ok(())
}
@@ -44,16 +50,28 @@ pub(crate) async fn run_remote_compact_task(
});
sess.send_event(&turn_context, start_event).await;
run_remote_compact_task_inner(&sess, &turn_context, InitialContextInjection::DoNotInject).await
run_remote_compact_task_inner(
&sess,
&turn_context,
InitialContextInjection::DoNotInject,
None,
)
.await
}
async fn run_remote_compact_task_inner(
sess: &Arc<Session>,
turn_context: &Arc<TurnContext>,
initial_context_injection: InitialContextInjection,
previous_user_turn_model: Option<&str>,
) -> CodexResult<()> {
if let Err(err) =
run_remote_compact_task_inner_impl(sess, turn_context, initial_context_injection).await
if let Err(err) = run_remote_compact_task_inner_impl(
sess,
turn_context,
initial_context_injection,
previous_user_turn_model,
)
.await
{
let event = EventMsg::Error(
err.to_error_event(Some("Error running remote compact task".to_string())),
@@ -68,15 +86,12 @@ async fn run_remote_compact_task_inner_impl(
sess: &Arc<Session>,
turn_context: &Arc<TurnContext>,
initial_context_injection: InitialContextInjection,
previous_user_turn_model: Option<&str>,
) -> CodexResult<()> {
let compaction_item = TurnItem::ContextCompaction(ContextCompactionItem::new());
sess.emit_turn_item_started(turn_context, &compaction_item)
.await;
let mut history = sess.clone_history().await;
// Keep compaction prompts in-distribution: if a model-switch update was injected at the
// tail of history (between turns), exclude it from the compaction request payload.
let stripped_model_switch_item =
extract_trailing_model_switch_update_for_compaction_request(&mut history);
let base_instructions = sess.get_base_instructions().await;
let deleted_items = trim_function_call_history_to_fit_context_window(
&mut history,
@@ -90,7 +105,6 @@ async fn run_remote_compact_task_inner_impl(
"trimmed history items before remote compaction"
);
}
// Required to keep `/undo` available after compaction
let ghost_snapshots: Vec<ResponseItem> = history
.raw_items()
@@ -134,13 +148,9 @@ async fn run_remote_compact_task_inner_impl(
turn_context.as_ref(),
new_history,
initial_context_injection,
previous_user_turn_model,
)
.await;
// Reattach the stripped model-switch update only after successful compaction so the model
// still sees the switch instructions on the next real sampling request.
if let Some(model_switch_item) = stripped_model_switch_item {
new_history.push(model_switch_item);
}
if !ghost_snapshots.is_empty() {
new_history.extend(ghost_snapshots);
@@ -170,6 +180,7 @@ pub(crate) async fn process_compacted_history(
turn_context: &TurnContext,
mut compacted_history: Vec<ResponseItem>,
initial_context_injection: InitialContextInjection,
previous_user_turn_model: Option<&str>,
) -> Vec<ResponseItem> {
// Mid-turn compaction is the only path that must inject initial context above the last user
// message in the replacement history. Pre-turn compaction instead injects context after the
@@ -178,7 +189,8 @@ pub(crate) async fn process_compacted_history(
initial_context_injection,
InitialContextInjection::BeforeLastUserMessage
) {
sess.build_initial_context(turn_context).await
sess.build_initial_context(turn_context, previous_user_turn_model)
.await
} else {
Vec::new()
};
@@ -205,10 +217,12 @@ pub(crate) async fn process_compacted_history(
fn should_keep_compacted_history_item(item: &ResponseItem) -> bool {
match item {
ResponseItem::Message { role, .. } if role == "developer" => false,
ResponseItem::Message { role, .. } if role == "user" => matches!(
crate::event_mapping::parse_turn_item(item),
Some(TurnItem::UserMessage(_))
),
ResponseItem::Message { role, .. } if role == "user" => {
matches!(
crate::event_mapping::parse_turn_item(item),
Some(TurnItem::UserMessage(_))
)
}
ResponseItem::Message { role, .. } if role == "assistant" => true,
ResponseItem::Message { .. } => false,
ResponseItem::Compaction { .. } => true,
codex-rs/core/src/config/edit.rs
+80
View File
@@ -195,6 +195,11 @@ mod document_helpers {
{
entry["scopes"] = array_from_iter(scopes.iter().cloned());
}
if let Some(resource) = &config.oauth_resource
&& !resource.is_empty()
{
entry["oauth_resource"] = value(resource.clone());
}
entry
}
@@ -839,6 +844,30 @@ impl ConfigEditsBuilder {
self
}
pub fn set_realtime_microphone(mut self, microphone: Option<&str>) -> Self {
let segments = vec!["audio".to_string(), "microphone".to_string()];
match microphone {
Some(microphone) => self.edits.push(ConfigEdit::SetPath {
segments,
value: value(microphone),
}),
None => self.edits.push(ConfigEdit::ClearPath { segments }),
}
self
}
pub fn set_realtime_speaker(mut self, speaker: Option<&str>) -> Self {
let segments = vec!["audio".to_string(), "speaker".to_string()];
match speaker {
Some(speaker) => self.edits.push(ConfigEdit::SetPath {
segments,
value: value(speaker),
}),
None => self.edits.push(ConfigEdit::ClearPath { segments }),
}
self
}
pub fn clear_legacy_windows_sandbox_keys(mut self) -> Self {
for key in [
"experimental_windows_sandbox",
@@ -1441,6 +1470,7 @@ gpt-5 = "gpt-5.1"
enabled_tools: Some(vec!["one".to_string(), "two".to_string()]),
disabled_tools: None,
scopes: None,
oauth_resource: None,
},
);
@@ -1465,6 +1495,7 @@ gpt-5 = "gpt-5.1"
enabled_tools: None,
disabled_tools: Some(vec!["forbidden".to_string()]),
scopes: None,
oauth_resource: Some("https://resource.example.com".to_string()),
},
);
@@ -1483,6 +1514,7 @@ bearer_token_env_var = \"TOKEN\"
enabled = false
startup_timeout_sec = 5.0
disabled_tools = [\"forbidden\"]
oauth_resource = \"https://resource.example.com\"
[mcp_servers.http.http_headers]
Z-Header = \"z\"
@@ -1532,6 +1564,7 @@ foo = { command = "cmd" }
enabled_tools: None,
disabled_tools: None,
scopes: None,
oauth_resource: None,
},
);
@@ -1578,6 +1611,7 @@ foo = { command = "cmd" } # keep me
enabled_tools: None,
disabled_tools: None,
scopes: None,
oauth_resource: None,
},
);
@@ -1623,6 +1657,7 @@ foo = { command = "cmd", args = ["--flag"] } # keep me
enabled_tools: None,
disabled_tools: None,
scopes: None,
oauth_resource: None,
},
);
@@ -1669,6 +1704,7 @@ foo = { command = "cmd" }
enabled_tools: None,
disabled_tools: None,
scopes: None,
oauth_resource: None,
},
);
@@ -1804,6 +1840,50 @@ model_reasoning_effort = "high"
assert_eq!(notice, Some(true));
}
#[test]
fn blocking_builder_set_realtime_audio_persists_and_clears() {
let tmp = tempdir().expect("tmpdir");
let codex_home = tmp.path();
ConfigEditsBuilder::new(codex_home)
.set_realtime_microphone(Some("USB Mic"))
.set_realtime_speaker(Some("Desk Speakers"))
.apply_blocking()
.expect("persist realtime audio");
let raw = std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
let config: TomlValue = toml::from_str(&raw).expect("parse config");
let realtime_audio = config
.get("audio")
.and_then(TomlValue::as_table)
.expect("audio table should exist");
assert_eq!(
realtime_audio.get("microphone").and_then(TomlValue::as_str),
Some("USB Mic")
);
assert_eq!(
realtime_audio.get("speaker").and_then(TomlValue::as_str),
Some("Desk Speakers")
);
ConfigEditsBuilder::new(codex_home)
.set_realtime_microphone(None)
.apply_blocking()
.expect("clear realtime microphone");
let raw = std::fs::read_to_string(codex_home.join(CONFIG_TOML_FILE)).expect("read config");
let config: TomlValue = toml::from_str(&raw).expect("parse config");
let realtime_audio = config
.get("audio")
.and_then(TomlValue::as_table)
.expect("audio table should exist");
assert_eq!(realtime_audio.get("microphone"), None);
assert_eq!(
realtime_audio.get("speaker").and_then(TomlValue::as_str),
Some("Desk Speakers")
);
}
#[test]
fn replace_mcp_servers_blocking_clears_table_when_empty() {
let tmp = tempdir().expect("tmpdir");
codex-rs/core/src/config/mod.rs
+141 -13
View File
@@ -49,8 +49,6 @@ use crate::project_doc::LOCAL_PROJECT_DOC_FILENAME;
use crate::protocol::AskForApproval;
use crate::protocol::ReadOnlyAccess;
use crate::protocol::SandboxPolicy;
#[cfg(target_os = "macos")]
use crate::seatbelt_permissions::MacOsSeatbeltProfileExtensions;
use crate::unified_exec::DEFAULT_MAX_BACKGROUND_TERMINAL_TIMEOUT_MS;
use crate::unified_exec::MIN_EMPTY_YIELD_TIME_MS;
use crate::windows_sandbox::WindowsSandboxLevelExt;
@@ -66,6 +64,7 @@ use codex_protocol::config_types::TrustLevel;
use codex_protocol::config_types::Verbosity;
use codex_protocol::config_types::WebSearchMode;
use codex_protocol::config_types::WindowsSandboxLevel;
use codex_protocol::models::MacOsSeatbeltProfileExtensions;
use codex_protocol::openai_models::ModelsResponse;
use codex_protocol::openai_models::ReasoningEffort;
use codex_rmcp_client::OAuthCredentialsStoreMode;
@@ -82,8 +81,6 @@ use std::path::Path;
use std::path::PathBuf;
#[cfg(test)]
use tempfile::tempdir;
#[cfg(not(target_os = "macos"))]
type MacOsSeatbeltProfileExtensions = ();
use crate::config::permissions::network_proxy_config_from_permissions;
use crate::config::profile::ConfigProfile;
@@ -412,9 +409,9 @@ pub struct Config {
/// global default").
pub plan_mode_reasoning_effort: Option<ReasoningEffort>,
/// If not "none", the value to use for `reasoning.summary` when making a
/// request using the Responses API.
pub model_reasoning_summary: ReasoningSummary,
/// Optional value to use for `reasoning.summary` when making a request
/// using the Responses API. When unset, the model catalog default is used.
pub model_reasoning_summary: Option<ReasoningSummary>,
/// Optional override to force-enable reasoning summaries for the configured model.
pub model_supports_reasoning_summaries: Option<bool>,
@@ -429,6 +426,9 @@ pub struct Config {
/// Base URL for requests to ChatGPT (as opposed to the OpenAI API).
pub chatgpt_base_url: String,
/// Machine-local realtime audio device preferences used by realtime voice.
pub realtime_audio: RealtimeAudioConfig,
/// Experimental / do not use. Overrides only the realtime conversation
/// websocket transport base URL (the `Op::RealtimeConversation` `/ws`
/// connection) without changing normal provider HTTP requests.
@@ -1178,6 +1178,10 @@ pub struct ConfigToml {
/// Base URL for requests to ChatGPT (as opposed to the OpenAI API).
pub chatgpt_base_url: Option<String>,
/// Machine-local realtime audio device preferences used by realtime voice.
#[serde(default)]
pub audio: Option<RealtimeAudioToml>,
/// Experimental / do not use. Overrides only the realtime conversation
/// websocket transport base URL (the `Op::RealtimeConversation` `/ws`
/// connection) without changing normal provider HTTP requests.
@@ -1309,6 +1313,19 @@ impl ProjectConfig {
}
}
#[derive(Debug, Clone, Default, PartialEq, Eq)]
pub struct RealtimeAudioConfig {
pub microphone: Option<String>,
pub speaker: Option<String>,
}
#[derive(Serialize, Deserialize, Debug, Clone, Default, PartialEq, Eq, JsonSchema)]
#[schemars(deny_unknown_fields)]
pub struct RealtimeAudioToml {
pub microphone: Option<String>,
pub speaker: Option<String>,
}
#[derive(Serialize, Deserialize, Debug, Clone, Default, PartialEq, JsonSchema)]
#[schemars(deny_unknown_fields)]
pub struct ToolsToml {
@@ -2141,8 +2158,7 @@ impl Config {
.or(cfg.plan_mode_reasoning_effort),
model_reasoning_summary: config_profile
.model_reasoning_summary
.or(cfg.model_reasoning_summary)
.unwrap_or_default(),
.or(cfg.model_reasoning_summary),
model_supports_reasoning_summaries: cfg.model_supports_reasoning_summaries,
model_catalog,
model_verbosity: config_profile.model_verbosity.or(cfg.model_verbosity),
@@ -2150,6 +2166,12 @@ impl Config {
.chatgpt_base_url
.or(cfg.chatgpt_base_url)
.unwrap_or("https://chatgpt.com/backend-api/".to_string()),
realtime_audio: cfg
.audio
.map_or_else(RealtimeAudioConfig::default, |audio| RealtimeAudioConfig {
microphone: audio.microphone,
speaker: audio.speaker,
}),
experimental_realtime_ws_base_url: cfg.experimental_realtime_ws_base_url,
experimental_realtime_ws_backend_prompt: cfg.experimental_realtime_ws_backend_prompt,
forced_chatgpt_workspace_id,
@@ -2411,6 +2433,7 @@ mod tests {
enabled_tools: None,
disabled_tools: None,
scopes: None,
oauth_resource: None,
}
}
@@ -2430,6 +2453,7 @@ mod tests {
enabled_tools: None,
disabled_tools: None,
scopes: None,
oauth_resource: None,
}
}
@@ -2466,7 +2490,10 @@ persistence = "none"
let memories = r#"
[memories]
generate_memories = false
use_memories = false
max_raw_memories_for_global = 512
max_unused_days = 21
max_rollout_age_days = 42
max_rollouts_per_startup = 9
min_rollout_idle_hours = 24
@@ -2477,7 +2504,10 @@ phase_2_model = "gpt-5"
toml::from_str::<ConfigToml>(memories).expect("TOML deserialization should succeed");
assert_eq!(
Some(MemoriesToml {
generate_memories: Some(false),
use_memories: Some(false),
max_raw_memories_for_global: Some(512),
max_unused_days: Some(21),
max_rollout_age_days: Some(42),
max_rollouts_per_startup: Some(9),
min_rollout_idle_hours: Some(24),
@@ -2496,7 +2526,10 @@ phase_2_model = "gpt-5"
assert_eq!(
config.memories,
MemoriesConfig {
generate_memories: false,
use_memories: false,
max_raw_memories_for_global: 512,
max_unused_days: 21,
max_rollout_age_days: 42,
max_rollouts_per_startup: 9,
min_rollout_idle_hours: 24,
@@ -3464,6 +3497,7 @@ profile = "project"
enabled_tools: None,
disabled_tools: None,
scopes: None,
oauth_resource: None,
},
);
@@ -3621,6 +3655,7 @@ bearer_token = "secret"
enabled_tools: None,
disabled_tools: None,
scopes: None,
oauth_resource: None,
},
)]);
@@ -3692,6 +3727,7 @@ ZIG_VAR = "3"
enabled_tools: None,
disabled_tools: None,
scopes: None,
oauth_resource: None,
},
)]);
@@ -3743,6 +3779,7 @@ ZIG_VAR = "3"
enabled_tools: None,
disabled_tools: None,
scopes: None,
oauth_resource: None,
},
)]);
@@ -3792,6 +3829,7 @@ ZIG_VAR = "3"
enabled_tools: None,
disabled_tools: None,
scopes: None,
oauth_resource: None,
},
)]);
@@ -3857,6 +3895,7 @@ startup_timeout_sec = 2.0
enabled_tools: None,
disabled_tools: None,
scopes: None,
oauth_resource: None,
},
)]);
apply_blocking(
@@ -3934,6 +3973,7 @@ X-Auth = "DOCS_AUTH"
enabled_tools: None,
disabled_tools: None,
scopes: None,
oauth_resource: None,
},
)]);
@@ -3964,6 +4004,7 @@ X-Auth = "DOCS_AUTH"
enabled_tools: None,
disabled_tools: None,
scopes: None,
oauth_resource: None,
},
);
apply_blocking(
@@ -4032,6 +4073,7 @@ url = "https://example.com/mcp"
enabled_tools: None,
disabled_tools: None,
scopes: None,
oauth_resource: None,
},
),
(
@@ -4052,6 +4094,7 @@ url = "https://example.com/mcp"
enabled_tools: None,
disabled_tools: None,
scopes: None,
oauth_resource: None,
},
),
]);
@@ -4135,6 +4178,7 @@ url = "https://example.com/mcp"
enabled_tools: None,
disabled_tools: None,
scopes: None,
oauth_resource: None,
},
)]);
@@ -4180,6 +4224,7 @@ url = "https://example.com/mcp"
enabled_tools: None,
disabled_tools: None,
scopes: None,
oauth_resource: None,
},
)]);
@@ -4225,6 +4270,7 @@ url = "https://example.com/mcp"
enabled_tools: Some(vec!["allowed".to_string()]),
disabled_tools: Some(vec!["blocked".to_string()]),
scopes: None,
oauth_resource: None,
},
)]);
@@ -4253,6 +4299,51 @@ url = "https://example.com/mcp"
Ok(())
}
#[tokio::test]
async fn replace_mcp_servers_streamable_http_serializes_oauth_resource() -> anyhow::Result<()> {
let codex_home = TempDir::new()?;
let servers = BTreeMap::from([(
"docs".to_string(),
McpServerConfig {
transport: McpServerTransportConfig::StreamableHttp {
url: "https://example.com/mcp".to_string(),
bearer_token_env_var: None,
http_headers: None,
env_http_headers: None,
},
enabled: true,
required: false,
disabled_reason: None,
startup_timeout_sec: None,
tool_timeout_sec: None,
enabled_tools: None,
disabled_tools: None,
scopes: None,
oauth_resource: Some("https://resource.example.com".to_string()),
},
)]);
apply_blocking(
codex_home.path(),
None,
&[ConfigEdit::ReplaceMcpServers(servers.clone())],
)?;
let config_path = codex_home.path().join(CONFIG_TOML_FILE);
let serialized = std::fs::read_to_string(&config_path)?;
assert!(serialized.contains(r#"oauth_resource = "https://resource.example.com""#));
let loaded = load_global_mcp_servers(codex_home.path()).await?;
let docs = loaded.get("docs").expect("docs entry");
assert_eq!(
docs.oauth_resource.as_deref(),
Some("https://resource.example.com")
);
Ok(())
}
#[tokio::test]
async fn set_model_updates_defaults() -> anyhow::Result<()> {
let codex_home = TempDir::new()?;
@@ -4761,12 +4852,13 @@ model_verbosity = "high"
show_raw_agent_reasoning: false,
model_reasoning_effort: Some(ReasoningEffort::High),
plan_mode_reasoning_effort: None,
model_reasoning_summary: ReasoningSummary::Detailed,
model_reasoning_summary: Some(ReasoningSummary::Detailed),
model_supports_reasoning_summaries: None,
model_catalog: None,
model_verbosity: None,
personality: Some(Personality::Pragmatic),
chatgpt_base_url: "https://chatgpt.com/backend-api/".to_string(),
realtime_audio: RealtimeAudioConfig::default(),
experimental_realtime_ws_base_url: None,
experimental_realtime_ws_backend_prompt: None,
base_instructions: None,
@@ -4887,12 +4979,13 @@ model_verbosity = "high"
show_raw_agent_reasoning: false,
model_reasoning_effort: None,
plan_mode_reasoning_effort: None,
model_reasoning_summary: ReasoningSummary::default(),
model_reasoning_summary: None,
model_supports_reasoning_summaries: None,
model_catalog: None,
model_verbosity: None,
personality: Some(Personality::Pragmatic),
chatgpt_base_url: "https://chatgpt.com/backend-api/".to_string(),
realtime_audio: RealtimeAudioConfig::default(),
experimental_realtime_ws_base_url: None,
experimental_realtime_ws_backend_prompt: None,
base_instructions: None,
@@ -5011,12 +5104,13 @@ model_verbosity = "high"
show_raw_agent_reasoning: false,
model_reasoning_effort: None,
plan_mode_reasoning_effort: None,
model_reasoning_summary: ReasoningSummary::default(),
model_reasoning_summary: None,
model_supports_reasoning_summaries: None,
model_catalog: None,
model_verbosity: None,
personality: Some(Personality::Pragmatic),
chatgpt_base_url: "https://chatgpt.com/backend-api/".to_string(),
realtime_audio: RealtimeAudioConfig::default(),
experimental_realtime_ws_base_url: None,
experimental_realtime_ws_backend_prompt: None,
base_instructions: None,
@@ -5121,12 +5215,13 @@ model_verbosity = "high"
show_raw_agent_reasoning: false,
model_reasoning_effort: Some(ReasoningEffort::High),
plan_mode_reasoning_effort: None,
model_reasoning_summary: ReasoningSummary::Detailed,
model_reasoning_summary: Some(ReasoningSummary::Detailed),
model_supports_reasoning_summaries: None,
model_catalog: None,
model_verbosity: Some(Verbosity::High),
personality: Some(Personality::Pragmatic),
chatgpt_base_url: "https://chatgpt.com/backend-api/".to_string(),
realtime_audio: RealtimeAudioConfig::default(),
experimental_realtime_ws_base_url: None,
experimental_realtime_ws_backend_prompt: None,
base_instructions: None,
@@ -5971,6 +6066,39 @@ experimental_realtime_ws_backend_prompt = "prompt from config"
);
Ok(())
}
#[test]
fn realtime_audio_loads_from_config_toml() -> std::io::Result<()> {
let cfg: ConfigToml = toml::from_str(
r#"
[audio]
microphone = "USB Mic"
speaker = "Desk Speakers"
"#,
)
.expect("TOML deserialization should succeed");
let realtime_audio = cfg
.audio
.as_ref()
.expect("realtime audio config should be present");
assert_eq!(realtime_audio.microphone.as_deref(), Some("USB Mic"));
assert_eq!(realtime_audio.speaker.as_deref(), Some("Desk Speakers"));
let codex_home = TempDir::new()?;
let config = Config::load_from_base_config_with_overrides(
cfg,
ConfigOverrides::default(),
codex_home.path().to_path_buf(),
)?;
assert_eq!(config.realtime_audio.microphone.as_deref(), Some("USB Mic"));
assert_eq!(
config.realtime_audio.speaker.as_deref(),
Some("Desk Speakers")
);
Ok(())
}
}
#[cfg(test)]
codex-rs/core/src/config/types.rs
+58
View File
@@ -27,6 +27,7 @@ pub const DEFAULT_MEMORIES_MAX_ROLLOUTS_PER_STARTUP: usize = 16;
pub const DEFAULT_MEMORIES_MAX_ROLLOUT_AGE_DAYS: i64 = 30;
pub const DEFAULT_MEMORIES_MIN_ROLLOUT_IDLE_HOURS: i64 = 6;
pub const DEFAULT_MEMORIES_MAX_RAW_MEMORIES_FOR_GLOBAL: usize = 1_024;
pub const DEFAULT_MEMORIES_MAX_UNUSED_DAYS: i64 = 30;
#[derive(Serialize, Deserialize, Debug, Clone, Copy, PartialEq, Eq, JsonSchema)]
#[serde(rename_all = "kebab-case")]
@@ -98,6 +99,10 @@ pub struct McpServerConfig {
/// Optional OAuth scopes to request during MCP login.
#[serde(default, skip_serializing_if = "Option::is_none")]
pub scopes: Option<Vec<String>>,
/// Optional OAuth resource parameter to include during MCP login (RFC 8707).
#[serde(default, skip_serializing_if = "Option::is_none")]
pub oauth_resource: Option<String>,
}
// Raw MCP config shape used for deserialization and JSON Schema generation.
@@ -142,6 +147,8 @@ pub(crate) struct RawMcpServerConfig {
pub disabled_tools: Option<Vec<String>>,
#[serde(default)]
pub scopes: Option<Vec<String>>,
#[serde(default)]
pub oauth_resource: Option<String>,
}
impl<'de> Deserialize<'de> for McpServerConfig {
@@ -165,6 +172,7 @@ impl<'de> Deserialize<'de> for McpServerConfig {
let enabled_tools = raw.enabled_tools.clone();
let disabled_tools = raw.disabled_tools.clone();
let scopes = raw.scopes.clone();
let oauth_resource = raw.oauth_resource.clone();
fn throw_if_set<E, T>(transport: &str, field: &str, value: Option<&T>) -> Result<(), E>
where
@@ -188,6 +196,7 @@ impl<'de> Deserialize<'de> for McpServerConfig {
throw_if_set("stdio", "bearer_token", raw.bearer_token.as_ref())?;
throw_if_set("stdio", "http_headers", raw.http_headers.as_ref())?;
throw_if_set("stdio", "env_http_headers", raw.env_http_headers.as_ref())?;
throw_if_set("stdio", "oauth_resource", raw.oauth_resource.as_ref())?;
McpServerTransportConfig::Stdio {
command,
args: raw.args.clone().unwrap_or_default(),
@@ -221,6 +230,7 @@ impl<'de> Deserialize<'de> for McpServerConfig {
enabled_tools,
disabled_tools,
scopes,
oauth_resource,
})
}
}
@@ -361,8 +371,14 @@ pub struct FeedbackConfigToml {
#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Default, JsonSchema)]
#[schemars(deny_unknown_fields)]
pub struct MemoriesToml {
/// When `false`, newly created threads are stored with `memory_mode = "disabled"` in the state DB.
pub generate_memories: Option<bool>,
/// When `false`, skip injecting memory usage instructions into developer prompts.
pub use_memories: Option<bool>,
/// Maximum number of recent raw memories retained for global consolidation.
pub max_raw_memories_for_global: Option<usize>,
/// Maximum number of days since a memory was last used before it becomes ineligible for phase 2 selection.
pub max_unused_days: Option<i64>,
/// Maximum age of the threads used for memories.
pub max_rollout_age_days: Option<i64>,
/// Maximum number of rollout candidates processed per pass.
@@ -378,7 +394,10 @@ pub struct MemoriesToml {
/// Effective memories settings after defaults are applied.
#[derive(Debug, Clone, PartialEq, Eq)]
pub struct MemoriesConfig {
pub generate_memories: bool,
pub use_memories: bool,
pub max_raw_memories_for_global: usize,
pub max_unused_days: i64,
pub max_rollout_age_days: i64,
pub max_rollouts_per_startup: usize,
pub min_rollout_idle_hours: i64,
@@ -389,7 +408,10 @@ pub struct MemoriesConfig {
impl Default for MemoriesConfig {
fn default() -> Self {
Self {
generate_memories: true,
use_memories: true,
max_raw_memories_for_global: DEFAULT_MEMORIES_MAX_RAW_MEMORIES_FOR_GLOBAL,
max_unused_days: DEFAULT_MEMORIES_MAX_UNUSED_DAYS,
max_rollout_age_days: DEFAULT_MEMORIES_MAX_ROLLOUT_AGE_DAYS,
max_rollouts_per_startup: DEFAULT_MEMORIES_MAX_ROLLOUTS_PER_STARTUP,
min_rollout_idle_hours: DEFAULT_MEMORIES_MIN_ROLLOUT_IDLE_HOURS,
@@ -403,10 +425,16 @@ impl From<MemoriesToml> for MemoriesConfig {
fn from(toml: MemoriesToml) -> Self {
let defaults = Self::default();
Self {
generate_memories: toml.generate_memories.unwrap_or(defaults.generate_memories),
use_memories: toml.use_memories.unwrap_or(defaults.use_memories),
max_raw_memories_for_global: toml
.max_raw_memories_for_global
.unwrap_or(defaults.max_raw_memories_for_global)
.min(4096),
max_unused_days: toml
.max_unused_days
.unwrap_or(defaults.max_unused_days)
.clamp(0, 365),
max_rollout_age_days: toml
.max_rollout_age_days
.unwrap_or(defaults.max_rollout_age_days)
@@ -1084,6 +1112,22 @@ mod tests {
);
}
#[test]
fn deserialize_streamable_http_server_config_with_oauth_resource() {
let cfg: McpServerConfig = toml::from_str(
r#"
url = "https://example.com/mcp"
oauth_resource = "https://api.example.com"
"#,
)
.expect("should deserialize http config with oauth_resource");
assert_eq!(
cfg.oauth_resource,
Some("https://api.example.com".to_string())
);
}
#[test]
fn deserialize_server_config_with_tool_filters() {
let cfg: McpServerConfig = toml::from_str(
@@ -1138,6 +1182,20 @@ mod tests {
"#,
)
.expect_err("should reject env_http_headers for stdio transport");
let err = toml::from_str::<McpServerConfig>(
r#"
command = "echo"
oauth_resource = "https://api.example.com"
"#,
)
.expect_err("should reject oauth_resource for stdio transport");
assert!(
err.to_string()
.contains("oauth_resource is not supported for stdio"),
"unexpected error: {err}"
);
}
#[test]
codex-rs/core/src/context_manager/history.rs
+31 -51
View File
@@ -1,14 +1,11 @@
use crate::codex::TurnContext;
use crate::context_manager::normalize;
use crate::instructions::SkillInstructions;
use crate::instructions::UserInstructions;
use crate::session_prefix::is_session_prefix;
use crate::event_mapping::is_contextual_user_message_content;
use crate::truncate::TruncationPolicy;
use crate::truncate::approx_token_count;
use crate::truncate::approx_tokens_from_byte_count_i64;
use crate::truncate::truncate_function_output_items_with_policy;
use crate::truncate::truncate_text;
use crate::user_shell_command::is_user_shell_command_text;
use codex_protocol::models::BaseInstructions;
use codex_protocol::models::ContentItem;
use codex_protocol::models::FunctionCallOutputBody;
@@ -347,32 +344,21 @@ impl ContextManager {
let policy_with_serialization_budget = policy * 1.2;
match item {
ResponseItem::FunctionCallOutput { call_id, output } => {
let body = match &output.body {
FunctionCallOutputBody::Text(content) => FunctionCallOutputBody::Text(
truncate_text(content, policy_with_serialization_budget),
),
FunctionCallOutputBody::ContentItems(items) => {
FunctionCallOutputBody::ContentItems(
truncate_function_output_items_with_policy(
items,
policy_with_serialization_budget,
),
)
}
};
ResponseItem::FunctionCallOutput {
call_id: call_id.clone(),
output: FunctionCallOutputPayload {
body,
success: output.success,
},
output: truncate_function_output_payload(
output,
policy_with_serialization_budget,
),
}
}
ResponseItem::CustomToolCallOutput { call_id, output } => {
let truncated = truncate_text(output, policy_with_serialization_budget);
ResponseItem::CustomToolCallOutput {
call_id: call_id.clone(),
output: truncated,
output: truncate_function_output_payload(
output,
policy_with_serialization_budget,
),
}
}
ResponseItem::Message { .. }
@@ -388,6 +374,25 @@ impl ContextManager {
}
}
fn truncate_function_output_payload(
output: &FunctionCallOutputPayload,
policy: TruncationPolicy,
) -> FunctionCallOutputPayload {
let body = match &output.body {
FunctionCallOutputBody::Text(content) => {
FunctionCallOutputBody::Text(truncate_text(content, policy))
}
FunctionCallOutputBody::ContentItems(items) => FunctionCallOutputBody::ContentItems(
truncate_function_output_items_with_policy(items, policy),
),
};
FunctionCallOutputPayload {
body,
success: output.success,
}
}
/// API messages include every non-system item (user/assistant messages, reasoning,
/// tool calls, tool outputs, shell calls, and web-search calls).
fn is_api_message(message: &ResponseItem) -> bool {
@@ -511,7 +516,8 @@ fn image_data_url_estimate_adjustment(item: &ResponseItem) -> (i64, i64) {
}
}
}
ResponseItem::FunctionCallOutput { output, .. } => {
ResponseItem::FunctionCallOutput { output, .. }
| ResponseItem::CustomToolCallOutput { output, .. } => {
if let FunctionCallOutputBody::ContentItems(items) = &output.body {
for content_item in items {
if let FunctionCallOutputContentItem::InputImage { image_url } = content_item {
@@ -554,33 +560,7 @@ pub(crate) fn is_user_turn_boundary(item: &ResponseItem) -> bool {
return false;
};
if role != "user" {
return false;
}
if UserInstructions::is_user_instructions(content)
|| SkillInstructions::is_skill_instructions(content)
{
return false;
}
for content_item in content {
match content_item {
ContentItem::InputText { text } => {
if is_session_prefix(text) || is_user_shell_command_text(text) {
return false;
}
}
ContentItem::OutputText { text } => {
if is_session_prefix(text) {
return false;
}
}
ContentItem::InputImage { .. } => {}
}
}
true
role == "user" && !is_contextual_user_message_content(content)
}
fn user_message_positions(items: &[ResponseItem]) -> Vec<usize> {
codex-rs/core/src/context_manager/history_tests.rs
+68 -13
View File
@@ -67,7 +67,7 @@ fn user_input_text_msg(text: &str) -> ResponseItem {
fn custom_tool_call_output(call_id: &str, output: &str) -> ResponseItem {
ResponseItem::CustomToolCallOutput {
call_id: call_id.to_string(),
output: output.to_string(),
output: FunctionCallOutputPayload::from_text(output.to_string()),
}
}
@@ -279,6 +279,24 @@ fn for_prompt_strips_images_when_model_does_not_support_images() {
},
]),
},
ResponseItem::CustomToolCall {
id: None,
status: None,
call_id: "tool-1".to_string(),
name: "js_repl".to_string(),
input: "view_image".to_string(),
},
ResponseItem::CustomToolCallOutput {
call_id: "tool-1".to_string(),
output: FunctionCallOutputPayload::from_content_items(vec![
FunctionCallOutputContentItem::InputText {
text: "js repl result".to_string(),
},
FunctionCallOutputContentItem::InputImage {
image_url: "https://example.com/js-repl-result.png".to_string(),
},
]),
},
];
let history = create_history_with_items(items);
let text_only_modalities = vec![InputModality::Text];
@@ -321,6 +339,25 @@ fn for_prompt_strips_images_when_model_does_not_support_images() {
},
]),
},
ResponseItem::CustomToolCall {
id: None,
status: None,
call_id: "tool-1".to_string(),
name: "js_repl".to_string(),
input: "view_image".to_string(),
},
ResponseItem::CustomToolCallOutput {
call_id: "tool-1".to_string(),
output: FunctionCallOutputPayload::from_content_items(vec![
FunctionCallOutputContentItem::InputText {
text: "js repl result".to_string(),
},
FunctionCallOutputContentItem::InputText {
text: "image content omitted because you do not support image input"
.to_string(),
},
]),
},
];
assert_eq!(stripped, expected);
@@ -563,7 +600,6 @@ fn drop_last_n_user_turns_preserves_prefix() {
fn drop_last_n_user_turns_ignores_session_prefix_user_messages() {
let items = vec![
user_input_text_msg("<environment_context>ctx</environment_context>"),
user_input_text_msg("<user_instructions>do the thing</user_instructions>"),
user_input_text_msg(
"# AGENTS.md instructions for test_directory\n\n<INSTRUCTIONS>\ntest_text\n</INSTRUCTIONS>",
),
@@ -586,7 +622,6 @@ fn drop_last_n_user_turns_ignores_session_prefix_user_messages() {
let expected_prefix_and_first_turn = vec![
user_input_text_msg("<environment_context>ctx</environment_context>"),
user_input_text_msg("<user_instructions>do the thing</user_instructions>"),
user_input_text_msg(
"# AGENTS.md instructions for test_directory\n\n<INSTRUCTIONS>\ntest_text\n</INSTRUCTIONS>",
),
@@ -608,7 +643,6 @@ fn drop_last_n_user_turns_ignores_session_prefix_user_messages() {
let expected_prefix_only = vec![
user_input_text_msg("<environment_context>ctx</environment_context>"),
user_input_text_msg("<user_instructions>do the thing</user_instructions>"),
user_input_text_msg(
"# AGENTS.md instructions for test_directory\n\n<INSTRUCTIONS>\ntest_text\n</INSTRUCTIONS>",
),
@@ -623,7 +657,6 @@ fn drop_last_n_user_turns_ignores_session_prefix_user_messages() {
let mut history = create_history_with_items(vec![
user_input_text_msg("<environment_context>ctx</environment_context>"),
user_input_text_msg("<user_instructions>do the thing</user_instructions>"),
user_input_text_msg(
"# AGENTS.md instructions for test_directory\n\n<INSTRUCTIONS>\ntest_text\n</INSTRUCTIONS>",
),
@@ -644,7 +677,6 @@ fn drop_last_n_user_turns_ignores_session_prefix_user_messages() {
let mut history = create_history_with_items(vec![
user_input_text_msg("<environment_context>ctx</environment_context>"),
user_input_text_msg("<user_instructions>do the thing</user_instructions>"),
user_input_text_msg(
"# AGENTS.md instructions for test_directory\n\n<INSTRUCTIONS>\ntest_text\n</INSTRUCTIONS>",
),
@@ -676,7 +708,7 @@ fn remove_first_item_handles_custom_tool_pair() {
},
ResponseItem::CustomToolCallOutput {
call_id: "tool-1".to_string(),
output: "ok".to_string(),
output: FunctionCallOutputPayload::from_text("ok".to_string()),
},
];
let mut h = create_history_with_items(items);
@@ -755,7 +787,7 @@ fn record_items_truncates_custom_tool_call_output_content() {
let long_output = line.repeat(2_500);
let item = ResponseItem::CustomToolCallOutput {
call_id: "tool-200".to_string(),
output: long_output.clone(),
output: FunctionCallOutputPayload::from_text(long_output.clone()),
};
history.record_items([&item], policy);
@@ -763,7 +795,8 @@ fn record_items_truncates_custom_tool_call_output_content() {
assert_eq!(history.items.len(), 1);
match &history.items[0] {
ResponseItem::CustomToolCallOutput { output, .. } => {
assert_ne!(output, &long_output);
let output = output.text_content().unwrap_or_default();
assert_ne!(output, long_output);
assert!(
output.contains("tokens truncated"),
"expected token-based truncation marker, got {output}"
@@ -954,7 +987,7 @@ fn normalize_adds_missing_output_for_custom_tool_call() {
},
ResponseItem::CustomToolCallOutput {
call_id: "tool-x".to_string(),
output: "aborted".to_string(),
output: FunctionCallOutputPayload::from_text("aborted".to_string()),
},
]
);
@@ -1021,7 +1054,7 @@ fn normalize_removes_orphan_function_call_output() {
fn normalize_removes_orphan_custom_tool_call_output() {
let items = vec![ResponseItem::CustomToolCallOutput {
call_id: "orphan-2".to_string(),
output: "ok".to_string(),
output: FunctionCallOutputPayload::from_text("ok".to_string()),
}];
let mut h = create_history_with_items(items);
@@ -1094,7 +1127,7 @@ fn normalize_mixed_inserts_and_removals() {
},
ResponseItem::CustomToolCallOutput {
call_id: "t1".to_string(),
output: "aborted".to_string(),
output: FunctionCallOutputPayload::from_text("aborted".to_string()),
},
ResponseItem::LocalShellCall {
id: None,
@@ -1196,7 +1229,7 @@ fn normalize_removes_orphan_function_call_output_panics_in_debug() {
fn normalize_removes_orphan_custom_tool_call_output_panics_in_debug() {
let items = vec![ResponseItem::CustomToolCallOutput {
call_id: "orphan-2".to_string(),
output: "ok".to_string(),
output: FunctionCallOutputPayload::from_text("ok".to_string()),
}];
let mut h = create_history_with_items(items);
h.normalize_history(&default_input_modalities());
@@ -1299,6 +1332,28 @@ fn image_data_url_payload_does_not_dominate_function_call_output_estimate() {
assert!(estimated < raw_len);
}
#[test]
fn image_data_url_payload_does_not_dominate_custom_tool_call_output_estimate() {
let payload = "C".repeat(50_000);
let image_url = format!("data:image/png;base64,{payload}");
let item = ResponseItem::CustomToolCallOutput {
call_id: "call-js-repl".to_string(),
output: FunctionCallOutputPayload::from_content_items(vec![
FunctionCallOutputContentItem::InputText {
text: "Screenshot captured".to_string(),
},
FunctionCallOutputContentItem::InputImage { image_url },
]),
};
let raw_len = serde_json::to_string(&item).unwrap().len() as i64;
let estimated = estimate_response_item_model_visible_bytes(&item);
let expected = raw_len - payload.len() as i64 + IMAGE_BYTES_ESTIMATE;
assert_eq!(estimated, expected);
assert!(estimated < raw_len);
}
#[test]
fn non_base64_image_urls_are_unchanged() {
let message_item = ResponseItem::Message {
codex-rs/core/src/context_manager/normalize.rs
+5 -11
View File
@@ -1,7 +1,6 @@
use std::collections::HashSet;
use codex_protocol::models::ContentItem;
use codex_protocol::models::FunctionCallOutputBody;
use codex_protocol::models::FunctionCallOutputContentItem;
use codex_protocol::models::FunctionCallOutputPayload;
use codex_protocol::models::ResponseItem;
@@ -35,10 +34,7 @@ pub(crate) fn ensure_call_outputs_present(items: &mut Vec<ResponseItem>) {
idx,
ResponseItem::FunctionCallOutput {
call_id: call_id.clone(),
output: FunctionCallOutputPayload {
body: FunctionCallOutputBody::Text("aborted".to_string()),
..Default::default()
},
output: FunctionCallOutputPayload::from_text("aborted".to_string()),
},
));
}
@@ -59,7 +55,7 @@ pub(crate) fn ensure_call_outputs_present(items: &mut Vec<ResponseItem>) {
idx,
ResponseItem::CustomToolCallOutput {
call_id: call_id.clone(),
output: "aborted".to_string(),
output: FunctionCallOutputPayload::from_text("aborted".to_string()),
},
));
}
@@ -82,10 +78,7 @@ pub(crate) fn ensure_call_outputs_present(items: &mut Vec<ResponseItem>) {
idx,
ResponseItem::FunctionCallOutput {
call_id: call_id.clone(),
output: FunctionCallOutputPayload {
body: FunctionCallOutputBody::Text("aborted".to_string()),
..Default::default()
},
output: FunctionCallOutputPayload::from_text("aborted".to_string()),
},
));
}
@@ -245,7 +238,8 @@ pub(crate) fn strip_images_when_unsupported(
}
*content = normalized_content;
}
ResponseItem::FunctionCallOutput { output, .. } => {
ResponseItem::FunctionCallOutput { output, .. }
| ResponseItem::CustomToolCallOutput { output, .. } => {
if let Some(content_items) = output.content_items_mut() {
let mut normalized_content_items = Vec::with_capacity(content_items.len());
for content_item in content_items.iter() {
codex-rs/core/src/context_manager/updates.rs
+65 -40
View File
@@ -4,6 +4,7 @@ use crate::features::Feature;
use crate::shell::Shell;
use codex_execpolicy::Policy;
use codex_protocol::config_types::Personality;
use codex_protocol::models::ContentItem;
use codex_protocol::models::DeveloperInstructions;
use codex_protocol::models::ResponseItem;
use codex_protocol::openai_models::ModelInfo;
@@ -30,7 +31,7 @@ fn build_permissions_update_item(
previous: Option<&TurnContextItem>,
next: &TurnContext,
exec_policy: &Policy,
) -> Option<ResponseItem> {
) -> Option<DeveloperInstructions> {
let prev = previous?;
if prev.sandbox_policy == *next.sandbox_policy.get()
&& prev.approval_policy == next.approval_policy.value()
@@ -38,27 +39,26 @@ fn build_permissions_update_item(
return None;
}
Some(
DeveloperInstructions::from_policy(
next.sandbox_policy.get(),
next.approval_policy.value(),
exec_policy,
&next.cwd,
next.features.enabled(Feature::RequestPermissions),
)
.into(),
)
Some(DeveloperInstructions::from_policy(
next.sandbox_policy.get(),
next.approval_policy.value(),
exec_policy,
&next.cwd,
next.features.enabled(Feature::RequestPermissions),
))
}
fn build_collaboration_mode_update_item(
previous: Option<&TurnContextItem>,
next: &TurnContext,
) -> Option<ResponseItem> {
) -> Option<DeveloperInstructions> {
let prev = previous?;
if prev.collaboration_mode.as_ref() != Some(&next.collaboration_mode) {
// If the next mode has empty developer instructions, this returns None and we emit no
// update, so prior collaboration instructions remain in the prompt history.
Some(DeveloperInstructions::from_collaboration_mode(&next.collaboration_mode)?.into())
Some(DeveloperInstructions::from_collaboration_mode(
&next.collaboration_mode,
)?)
} else {
None
}
@@ -68,7 +68,7 @@ fn build_personality_update_item(
previous: Option<&TurnContextItem>,
next: &TurnContext,
personality_feature_enabled: bool,
) -> Option<ResponseItem> {
) -> Option<DeveloperInstructions> {
if !personality_feature_enabled {
return None;
}
@@ -82,8 +82,7 @@ fn build_personality_update_item(
{
let model_info = &next.model_info;
let personality_message = personality_message_for(model_info, personality);
personality_message
.map(|message| DeveloperInstructions::personality_spec_message(message).into())
personality_message.map(DeveloperInstructions::personality_spec_message)
} else {
None
}
@@ -103,7 +102,7 @@ pub(crate) fn personality_message_for(
pub(crate) fn build_model_instructions_update_item(
previous_user_turn_model: Option<&str>,
next: &TurnContext,
) -> Option<ResponseItem> {
) -> Option<DeveloperInstructions> {
let previous_model = previous_user_turn_model?;
if previous_model == next.model_info.slug {
return None;
@@ -114,7 +113,36 @@ pub(crate) fn build_model_instructions_update_item(
return None;
}
Some(DeveloperInstructions::model_switch_message(model_instructions).into())
Some(DeveloperInstructions::model_switch_message(
model_instructions,
))
}
pub(crate) fn build_developer_update_item(text_sections: Vec<String>) -> Option<ResponseItem> {
build_text_message("developer", text_sections)
}
pub(crate) fn build_contextual_user_message(text_sections: Vec<String>) -> Option<ResponseItem> {
build_text_message("user", text_sections)
}
fn build_text_message(role: &str, text_sections: Vec<String>) -> Option<ResponseItem> {
if text_sections.is_empty() {
return None;
}
let content = text_sections
.into_iter()
.map(|text| ContentItem::InputText { text })
.collect();
Some(ResponseItem::Message {
id: None,
role: role.to_string(),
content,
end_turn: None,
phase: None,
})
}
pub(crate) fn build_settings_update_items(
@@ -125,29 +153,26 @@ pub(crate) fn build_settings_update_items(
exec_policy: &Policy,
personality_feature_enabled: bool,
) -> Vec<ResponseItem> {
let mut update_items = Vec::new();
let contextual_user_message = build_environment_update_item(previous, next, shell);
let developer_update_sections = [
// Keep model-switch instructions first so model-specific guidance is read before
// any other context diffs on this turn.
build_model_instructions_update_item(previous_user_turn_model, next),
build_permissions_update_item(previous, next, exec_policy),
build_collaboration_mode_update_item(previous, next),
build_personality_update_item(previous, next, personality_feature_enabled),
]
.into_iter()
.flatten()
.map(DeveloperInstructions::into_text)
.collect();
// Keep model-switch instructions first so model-specific guidance is read before
// any other context diffs on this turn.
if let Some(model_instructions_item) =
build_model_instructions_update_item(previous_user_turn_model, next)
{
update_items.push(model_instructions_item);
let mut items = Vec::with_capacity(2);
if let Some(developer_message) = build_developer_update_item(developer_update_sections) {
items.push(developer_message);
}
if let Some(env_item) = build_environment_update_item(previous, next, shell) {
update_items.push(env_item);
if let Some(contextual_user_message) = contextual_user_message {
items.push(contextual_user_message);
}
if let Some(permissions_item) = build_permissions_update_item(previous, next, exec_policy) {
update_items.push(permissions_item);
}
if let Some(collaboration_mode_item) = build_collaboration_mode_update_item(previous, next) {
update_items.push(collaboration_mode_item);
}
if let Some(personality_item) =
build_personality_update_item(previous, next, personality_feature_enabled)
{
update_items.push(personality_item);
}
update_items
items
}
codex-rs/core/src/contextual_user_message.rs
+139
View File
@@ -0,0 +1,139 @@
use codex_protocol::models::ContentItem;
use codex_protocol::models::ResponseItem;
use codex_protocol::protocol::ENVIRONMENT_CONTEXT_CLOSE_TAG;
use codex_protocol::protocol::ENVIRONMENT_CONTEXT_OPEN_TAG;
pub(crate) const AGENTS_MD_START_MARKER: &str = "# AGENTS.md instructions for ";
pub(crate) const AGENTS_MD_END_MARKER: &str = "</INSTRUCTIONS>";
pub(crate) const SKILL_OPEN_TAG: &str = "<skill>";
pub(crate) const SKILL_CLOSE_TAG: &str = "</skill>";
pub(crate) const USER_SHELL_COMMAND_OPEN_TAG: &str = "<user_shell_command>";
pub(crate) const USER_SHELL_COMMAND_CLOSE_TAG: &str = "</user_shell_command>";
pub(crate) const TURN_ABORTED_OPEN_TAG: &str = "<turn_aborted>";
pub(crate) const TURN_ABORTED_CLOSE_TAG: &str = "</turn_aborted>";
pub(crate) const SUBAGENT_NOTIFICATION_OPEN_TAG: &str = "<subagent_notification>";
pub(crate) const SUBAGENT_NOTIFICATION_CLOSE_TAG: &str = "</subagent_notification>";
#[derive(Clone, Copy)]
pub(crate) struct ContextualUserFragmentDefinition {
start_marker: &'static str,
end_marker: &'static str,
}
impl ContextualUserFragmentDefinition {
pub(crate) const fn new(start_marker: &'static str, end_marker: &'static str) -> Self {
Self {
start_marker,
end_marker,
}
}
pub(crate) fn matches_text(&self, text: &str) -> bool {
let trimmed = text.trim_start();
let starts_with_marker = trimmed
.get(..self.start_marker.len())
.is_some_and(|candidate| candidate.eq_ignore_ascii_case(self.start_marker));
let trimmed = trimmed.trim_end();
let ends_with_marker = trimmed
.get(trimmed.len().saturating_sub(self.end_marker.len())..)
.is_some_and(|candidate| candidate.eq_ignore_ascii_case(self.end_marker));
starts_with_marker && ends_with_marker
}
pub(crate) const fn start_marker(&self) -> &'static str {
self.start_marker
}
pub(crate) const fn end_marker(&self) -> &'static str {
self.end_marker
}
pub(crate) fn wrap(&self, body: String) -> String {
format!("{}\n{}\n{}", self.start_marker, body, self.end_marker)
}
pub(crate) fn into_message(self, text: String) -> ResponseItem {
ResponseItem::Message {
id: None,
role: "user".to_string(),
content: vec![ContentItem::InputText { text }],
end_turn: None,
phase: None,
}
}
}
pub(crate) const AGENTS_MD_FRAGMENT: ContextualUserFragmentDefinition =
ContextualUserFragmentDefinition::new(AGENTS_MD_START_MARKER, AGENTS_MD_END_MARKER);
pub(crate) const ENVIRONMENT_CONTEXT_FRAGMENT: ContextualUserFragmentDefinition =
ContextualUserFragmentDefinition::new(
ENVIRONMENT_CONTEXT_OPEN_TAG,
ENVIRONMENT_CONTEXT_CLOSE_TAG,
);
pub(crate) const SKILL_FRAGMENT: ContextualUserFragmentDefinition =
ContextualUserFragmentDefinition::new(SKILL_OPEN_TAG, SKILL_CLOSE_TAG);
pub(crate) const USER_SHELL_COMMAND_FRAGMENT: ContextualUserFragmentDefinition =
ContextualUserFragmentDefinition::new(
USER_SHELL_COMMAND_OPEN_TAG,
USER_SHELL_COMMAND_CLOSE_TAG,
);
pub(crate) const TURN_ABORTED_FRAGMENT: ContextualUserFragmentDefinition =
ContextualUserFragmentDefinition::new(TURN_ABORTED_OPEN_TAG, TURN_ABORTED_CLOSE_TAG);
pub(crate) const SUBAGENT_NOTIFICATION_FRAGMENT: ContextualUserFragmentDefinition =
ContextualUserFragmentDefinition::new(
SUBAGENT_NOTIFICATION_OPEN_TAG,
SUBAGENT_NOTIFICATION_CLOSE_TAG,
);
const CONTEXTUAL_USER_FRAGMENTS: &[ContextualUserFragmentDefinition] = &[
AGENTS_MD_FRAGMENT,
ENVIRONMENT_CONTEXT_FRAGMENT,
SKILL_FRAGMENT,
USER_SHELL_COMMAND_FRAGMENT,
TURN_ABORTED_FRAGMENT,
SUBAGENT_NOTIFICATION_FRAGMENT,
];
pub(crate) fn is_contextual_user_fragment(content_item: &ContentItem) -> bool {
let ContentItem::InputText { text } = content_item else {
return false;
};
CONTEXTUAL_USER_FRAGMENTS
.iter()
.any(|definition| definition.matches_text(text))
}
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn detects_environment_context_fragment() {
assert!(is_contextual_user_fragment(&ContentItem::InputText {
text: "<environment_context>\n<cwd>/tmp</cwd>\n</environment_context>".to_string(),
}));
}
#[test]
fn detects_agents_instructions_fragment() {
assert!(is_contextual_user_fragment(&ContentItem::InputText {
text: "# AGENTS.md instructions for /tmp\n\n<INSTRUCTIONS>\nbody\n</INSTRUCTIONS>"
.to_string(),
}));
}
#[test]
fn detects_subagent_notification_fragment_case_insensitively() {
assert!(
SUBAGENT_NOTIFICATION_FRAGMENT
.matches_text("<SUBAGENT_NOTIFICATION>{}</subagent_notification>")
);
}
#[test]
fn ignores_regular_user_text() {
assert!(!is_contextual_user_fragment(&ContentItem::InputText {
text: "hello".to_string(),
}));
}
}
codex-rs/core/src/environment_context.rs
+195 -32
View File
@@ -1,9 +1,7 @@
use crate::codex::TurnContext;
use crate::contextual_user_message::ENVIRONMENT_CONTEXT_FRAGMENT;
use crate::shell::Shell;
use codex_protocol::models::ContentItem;
use codex_protocol::models::ResponseItem;
use codex_protocol::protocol::ENVIRONMENT_CONTEXT_CLOSE_TAG;
use codex_protocol::protocol::ENVIRONMENT_CONTEXT_OPEN_TAG;
use codex_protocol::protocol::TurnContextItem;
use codex_protocol::protocol::TurnContextNetworkItem;
use serde::Deserialize;
@@ -15,7 +13,10 @@ use std::path::PathBuf;
pub(crate) struct EnvironmentContext {
pub cwd: Option<PathBuf>,
pub shell: Shell,
pub current_date: Option<String>,
pub timezone: Option<String>,
pub network: Option<NetworkContext>,
pub subagents: Option<String>,
}
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq, Default)]
@@ -25,11 +26,21 @@ pub(crate) struct NetworkContext {
}
impl EnvironmentContext {
pub fn new(cwd: Option<PathBuf>, shell: Shell, network: Option<NetworkContext>) -> Self {
pub fn new(
cwd: Option<PathBuf>,
shell: Shell,
current_date: Option<String>,
timezone: Option<String>,
network: Option<NetworkContext>,
subagents: Option<String>,
) -> Self {
Self {
cwd,
shell,
current_date,
timezone,
network,
subagents,
}
}
@@ -39,11 +50,17 @@ impl EnvironmentContext {
pub fn equals_except_shell(&self, other: &EnvironmentContext) -> bool {
let EnvironmentContext {
cwd,
current_date,
timezone,
network,
// should compare all fields except shell
subagents,
shell: _,
} = other;
self.cwd == *cwd && self.network == *network
self.cwd == *cwd
&& self.current_date == *current_date
&& self.timezone == *timezone
&& self.network == *network
&& self.subagents == *subagents
}
pub fn diff_from_turn_context_item(
@@ -58,19 +75,24 @@ impl EnvironmentContext {
} else {
None
};
let current_date = after.current_date.clone();
let timezone = after.timezone.clone();
let network = if before_network != after_network {
after_network
} else {
before_network
};
EnvironmentContext::new(cwd, shell.clone(), network)
EnvironmentContext::new(cwd, shell.clone(), current_date, timezone, network, None)
}
pub fn from_turn_context(turn_context: &TurnContext, shell: &Shell) -> Self {
Self::new(
Some(turn_context.cwd.clone()),
shell.clone(),
turn_context.current_date.clone(),
turn_context.timezone.clone(),
Self::network_from_turn_context(turn_context),
None,
)
}
@@ -78,10 +100,20 @@ impl EnvironmentContext {
Self::new(
Some(turn_context_item.cwd.clone()),
shell.clone(),
turn_context_item.current_date.clone(),
turn_context_item.timezone.clone(),
Self::network_from_turn_context_item(turn_context_item),
None,
)
}
pub fn with_subagents(mut self, subagents: String) -> Self {
if !subagents.is_empty() {
self.subagents = Some(subagents);
}
self
}
fn network_from_turn_context(turn_context: &TurnContext) -> Option<NetworkContext> {
let network = turn_context
.config
@@ -122,13 +154,19 @@ impl EnvironmentContext {
/// </environment_context>
/// ```
pub fn serialize_to_xml(self) -> String {
let mut lines = vec![ENVIRONMENT_CONTEXT_OPEN_TAG.to_string()];
let mut lines = Vec::new();
if let Some(cwd) = self.cwd {
lines.push(format!(" <cwd>{}</cwd>", cwd.to_string_lossy()));
}
let shell_name = self.shell.name();
lines.push(format!(" <shell>{shell_name}</shell>"));
if let Some(current_date) = self.current_date {
lines.push(format!(" <current_date>{current_date}</current_date>"));
}
if let Some(timezone) = self.timezone {
lines.push(format!(" <timezone>{timezone}</timezone>"));
}
match self.network {
Some(ref network) => {
lines.push(" <network enabled=\"true\">".to_string());
@@ -145,22 +183,18 @@ impl EnvironmentContext {
// lines.push(" <network enabled=\"false\" />".to_string());
}
}
lines.push(ENVIRONMENT_CONTEXT_CLOSE_TAG.to_string());
lines.join("\n")
if let Some(subagents) = self.subagents {
lines.push(" <subagents>".to_string());
lines.extend(subagents.lines().map(|line| format!(" {line}")));
lines.push(" </subagents>".to_string());
}
ENVIRONMENT_CONTEXT_FRAGMENT.wrap(lines.join("\n"))
}
}
impl From<EnvironmentContext> for ResponseItem {
fn from(ec: EnvironmentContext) -> Self {
ResponseItem::Message {
id: None,
role: "user".to_string(),
content: vec![ContentItem::InputText {
text: ec.serialize_to_xml(),
}],
end_turn: None,
phase: None,
}
ENVIRONMENT_CONTEXT_FRAGMENT.into_message(ec.serialize_to_xml())
}
}
@@ -183,12 +217,21 @@ mod tests {
#[test]
fn serialize_workspace_write_environment_context() {
let cwd = test_path_buf("/repo");
let context = EnvironmentContext::new(Some(cwd.clone()), fake_shell(), None);
let context = EnvironmentContext::new(
Some(cwd.clone()),
fake_shell(),
Some("2026-02-26".to_string()),
Some("America/Los_Angeles".to_string()),
None,
None,
);
let expected = format!(
r#"<environment_context>
<cwd>{cwd}</cwd>
<shell>bash</shell>
<current_date>2026-02-26</current_date>
<timezone>America/Los_Angeles</timezone>
</environment_context>"#,
cwd = cwd.display(),
);
@@ -202,13 +245,21 @@ mod tests {
allowed_domains: vec!["api.example.com".to_string(), "*.openai.com".to_string()],
denied_domains: vec!["blocked.example.com".to_string()],
};
let context =
EnvironmentContext::new(Some(test_path_buf("/repo")), fake_shell(), Some(network));
let context = EnvironmentContext::new(
Some(test_path_buf("/repo")),
fake_shell(),
Some("2026-02-26".to_string()),
Some("America/Los_Angeles".to_string()),
Some(network),
None,
);
let expected = format!(
r#"<environment_context>
<cwd>{}</cwd>
<shell>bash</shell>
<current_date>2026-02-26</current_date>
<timezone>America/Los_Angeles</timezone>
<network enabled="true">
<allowed>api.example.com</allowed>
<allowed>*.openai.com</allowed>
@@ -223,10 +274,19 @@ mod tests {
#[test]
fn serialize_read_only_environment_context() {
let context = EnvironmentContext::new(None, fake_shell(), None);
let context = EnvironmentContext::new(
None,
fake_shell(),
Some("2026-02-26".to_string()),
Some("America/Los_Angeles".to_string()),
None,
None,
);
let expected = r#"<environment_context>
<shell>bash</shell>
<current_date>2026-02-26</current_date>
<timezone>America/Los_Angeles</timezone>
</environment_context>"#;
assert_eq!(context.serialize_to_xml(), expected);
@@ -234,10 +294,19 @@ mod tests {
#[test]
fn serialize_external_sandbox_environment_context() {
let context = EnvironmentContext::new(None, fake_shell(), None);
let context = EnvironmentContext::new(
None,
fake_shell(),
Some("2026-02-26".to_string()),
Some("America/Los_Angeles".to_string()),
None,
None,
);
let expected = r#"<environment_context>
<shell>bash</shell>
<current_date>2026-02-26</current_date>
<timezone>America/Los_Angeles</timezone>
</environment_context>"#;
assert_eq!(context.serialize_to_xml(), expected);
@@ -245,10 +314,19 @@ mod tests {
#[test]
fn serialize_external_sandbox_with_restricted_network_environment_context() {
let context = EnvironmentContext::new(None, fake_shell(), None);
let context = EnvironmentContext::new(
None,
fake_shell(),
Some("2026-02-26".to_string()),
Some("America/Los_Angeles".to_string()),
None,
None,
);
let expected = r#"<environment_context>
<shell>bash</shell>
<current_date>2026-02-26</current_date>
<timezone>America/Los_Angeles</timezone>
</environment_context>"#;
assert_eq!(context.serialize_to_xml(), expected);
@@ -256,10 +334,19 @@ mod tests {
#[test]
fn serialize_full_access_environment_context() {
let context = EnvironmentContext::new(None, fake_shell(), None);
let context = EnvironmentContext::new(
None,
fake_shell(),
Some("2026-02-26".to_string()),
Some("America/Los_Angeles".to_string()),
None,
None,
);
let expected = r#"<environment_context>
<shell>bash</shell>
<current_date>2026-02-26</current_date>
<timezone>America/Los_Angeles</timezone>
</environment_context>"#;
assert_eq!(context.serialize_to_xml(), expected);
@@ -267,23 +354,65 @@ mod tests {
#[test]
fn equals_except_shell_compares_cwd() {
let context1 = EnvironmentContext::new(Some(PathBuf::from("/repo")), fake_shell(), None);
let context2 = EnvironmentContext::new(Some(PathBuf::from("/repo")), fake_shell(), None);
let context1 = EnvironmentContext::new(
Some(PathBuf::from("/repo")),
fake_shell(),
None,
None,
None,
None,
);
let context2 = EnvironmentContext::new(
Some(PathBuf::from("/repo")),
fake_shell(),
None,
None,
None,
None,
);
assert!(context1.equals_except_shell(&context2));
}
#[test]
fn equals_except_shell_ignores_sandbox_policy() {
let context1 = EnvironmentContext::new(Some(PathBuf::from("/repo")), fake_shell(), None);
let context2 = EnvironmentContext::new(Some(PathBuf::from("/repo")), fake_shell(), None);
let context1 = EnvironmentContext::new(
Some(PathBuf::from("/repo")),
fake_shell(),
None,
None,
None,
None,
);
let context2 = EnvironmentContext::new(
Some(PathBuf::from("/repo")),
fake_shell(),
None,
None,
None,
None,
);
assert!(context1.equals_except_shell(&context2));
}
#[test]
fn equals_except_shell_compares_cwd_differences() {
let context1 = EnvironmentContext::new(Some(PathBuf::from("/repo1")), fake_shell(), None);
let context2 = EnvironmentContext::new(Some(PathBuf::from("/repo2")), fake_shell(), None);
let context1 = EnvironmentContext::new(
Some(PathBuf::from("/repo1")),
fake_shell(),
None,
None,
None,
None,
);
let context2 = EnvironmentContext::new(
Some(PathBuf::from("/repo2")),
fake_shell(),
None,
None,
None,
None,
);
assert!(!context1.equals_except_shell(&context2));
}
@@ -298,6 +427,9 @@ mod tests {
shell_snapshot: crate::shell::empty_shell_snapshot_receiver(),
},
None,
None,
None,
None,
);
let context2 = EnvironmentContext::new(
Some(PathBuf::from("/repo")),
@@ -307,8 +439,39 @@ mod tests {
shell_snapshot: crate::shell::empty_shell_snapshot_receiver(),
},
None,
None,
None,
None,
);
assert!(context1.equals_except_shell(&context2));
}
#[test]
fn serialize_environment_context_with_subagents() {
let context = EnvironmentContext::new(
Some(test_path_buf("/repo")),
fake_shell(),
Some("2026-02-26".to_string()),
Some("America/Los_Angeles".to_string()),
None,
Some("- agent-1: atlas\n- agent-2".to_string()),
);
let expected = format!(
r#"<environment_context>
<cwd>{}</cwd>
<shell>bash</shell>
<current_date>2026-02-26</current_date>
<timezone>America/Los_Angeles</timezone>
<subagents>
- agent-1: atlas
- agent-2
</subagents>
</environment_context>"#,
test_path_buf("/repo").display()
);
assert_eq!(context.serialize_to_xml(), expected);
}
}
codex-rs/core/src/event_mapping.rs
+23 -14
View File
@@ -18,16 +18,15 @@ use codex_protocol::user_input::UserInput;
use tracing::warn;
use uuid::Uuid;
use crate::instructions::SkillInstructions;
use crate::instructions::UserInstructions;
use crate::session_prefix::is_session_prefix;
use crate::user_shell_command::is_user_shell_command_text;
use crate::contextual_user_message::is_contextual_user_fragment;
use crate::web_search::web_search_action_detail;
pub(crate) fn is_contextual_user_message_content(message: &[ContentItem]) -> bool {
message.iter().any(is_contextual_user_fragment)
}
fn parse_user_message(message: &[ContentItem]) -> Option<UserMessageItem> {
if UserInstructions::is_user_instructions(message)
|| SkillInstructions::is_skill_instructions(message)
{
if is_contextual_user_message_content(message) {
return None;
}
@@ -44,9 +43,6 @@ fn parse_user_message(message: &[ContentItem]) -> Option<UserMessageItem> {
{
continue;
}
if is_session_prefix(text) || is_user_shell_command_text(text) {
return None;
}
content.push(UserInput::Text {
text: text.clone(),
// Model input content does not carry UI element ranges.
@@ -59,9 +55,6 @@ fn parse_user_message(message: &[ContentItem]) -> Option<UserMessageItem> {
});
}
ContentItem::OutputText { text } => {
if is_session_prefix(text) {
return None;
}
warn!("Output text in user message: {}", text);
}
}
@@ -299,7 +292,7 @@ mod tests {
id: None,
role: "user".to_string(),
content: vec![ContentItem::InputText {
text: "<user_instructions>test_text</user_instructions>".to_string(),
text: "# AGENTS.md instructions for test_directory\n\n<INSTRUCTIONS>\ntest_text\n</INSTRUCTIONS>".to_string(),
}],
end_turn: None,
phase: None,
@@ -341,6 +334,22 @@ mod tests {
end_turn: None,
phase: None,
},
ResponseItem::Message {
id: None,
role: "user".to_string(),
content: vec![
ContentItem::InputText {
text: "<environment_context>ctx</environment_context>".to_string(),
},
ContentItem::InputText {
text:
"# AGENTS.md instructions for dir\n\n<INSTRUCTIONS>\nbody\n</INSTRUCTIONS>"
.to_string(),
},
],
end_turn: None,
phase: None,
},
];
for item in items {
codex-rs/core/src/exec.rs
+2
View File
@@ -219,6 +219,8 @@ pub async fn process_exec_tool_call(
enforce_managed_network,
network: network.as_ref(),
sandbox_policy_cwd: sandbox_cwd,
#[cfg(target_os = "macos")]
macos_seatbelt_profile_extensions: None,
codex_linux_sandbox_exe: codex_linux_sandbox_exe.as_ref(),
use_linux_sandbox_bwrap,
windows_sandbox_level,
codex-rs/core/src/features.rs
+1 -9
View File
@@ -133,8 +133,6 @@ pub enum Feature {
SkillMcpDependencyInstall,
/// Prompt for missing skill env var dependencies.
SkillEnvVarDependencyPrompt,
/// Emit skill approval test prompts/events.
SkillApproval,
/// Steer feature flag - when enabled, Enter submits immediately instead of queuing.
/// Kept for config backward compatibility; behavior is always steer-enabled.
Steer,
@@ -465,7 +463,7 @@ pub const FEATURES: &[FeatureSpec] = &[
key: "js_repl",
stage: Stage::Experimental {
name: "JavaScript REPL",
menu_description: "Enable a persistent Node-backed JavaScript REPL for interactive website debugging and other inline JavaScript execution capabilities. Requires Node >= v24.13.1 installed.",
menu_description: "Enable a persistent Node-backed JavaScript REPL for interactive website debugging and other inline JavaScript execution capabilities. Requires Node >= v22.22.0 installed.",
announcement: "NEW: JavaScript REPL is now available in /experimental. Enable it, then start a new chat or restart Codex to use it.",
},
default_enabled: false,
@@ -630,12 +628,6 @@ pub const FEATURES: &[FeatureSpec] = &[
stage: Stage::UnderDevelopment,
default_enabled: false,
},
FeatureSpec {
id: Feature::SkillApproval,
key: "skill_approval",
stage: Stage::UnderDevelopment,
default_enabled: false,
},
FeatureSpec {
id: Feature::Steer,
key: "steer",
codex-rs/core/src/instructions/mod.rs
-1
View File
@@ -1,6 +1,5 @@
mod user_instructions;
pub(crate) use user_instructions::SkillInstructions;
pub use user_instructions::USER_INSTRUCTIONS_OPEN_TAG_LEGACY;
pub use user_instructions::USER_INSTRUCTIONS_PREFIX;
pub(crate) use user_instructions::UserInstructions;
codex-rs/core/src/instructions/user_instructions.rs
+25 -69
View File
@@ -1,12 +1,12 @@
use serde::Deserialize;
use serde::Serialize;
use codex_protocol::models::ContentItem;
use codex_protocol::models::ResponseItem;
pub const USER_INSTRUCTIONS_OPEN_TAG_LEGACY: &str = "<user_instructions>";
use crate::contextual_user_message::AGENTS_MD_FRAGMENT;
use crate::contextual_user_message::SKILL_FRAGMENT;
pub const USER_INSTRUCTIONS_PREFIX: &str = "# AGENTS.md instructions for ";
pub const SKILL_INSTRUCTIONS_PREFIX: &str = "<skill";
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
#[serde(rename = "user_instructions", rename_all = "snake_case")]
@@ -16,31 +16,20 @@ pub(crate) struct UserInstructions {
}
impl UserInstructions {
pub fn is_user_instructions(message: &[ContentItem]) -> bool {
if let [ContentItem::InputText { text }] = message {
text.starts_with(USER_INSTRUCTIONS_PREFIX)
|| text.starts_with(USER_INSTRUCTIONS_OPEN_TAG_LEGACY)
} else {
false
}
pub(crate) fn serialize_to_text(&self) -> String {
format!(
"{prefix}{directory}\n\n<INSTRUCTIONS>\n{contents}\n{suffix}",
prefix = AGENTS_MD_FRAGMENT.start_marker(),
directory = self.directory,
contents = self.text,
suffix = AGENTS_MD_FRAGMENT.end_marker(),
)
}
}
impl From<UserInstructions> for ResponseItem {
fn from(ui: UserInstructions) -> Self {
ResponseItem::Message {
id: None,
role: "user".to_string(),
content: vec![ContentItem::InputText {
text: format!(
"{USER_INSTRUCTIONS_PREFIX}{directory}\n\n<INSTRUCTIONS>\n{contents}\n</INSTRUCTIONS>",
directory = ui.directory,
contents = ui.text
),
}],
end_turn: None,
phase: None,
}
AGENTS_MD_FRAGMENT.into_message(ui.serialize_to_text())
}
}
@@ -52,36 +41,21 @@ pub(crate) struct SkillInstructions {
pub contents: String,
}
impl SkillInstructions {
pub fn is_skill_instructions(message: &[ContentItem]) -> bool {
if let [ContentItem::InputText { text }] = message {
text.starts_with(SKILL_INSTRUCTIONS_PREFIX)
} else {
false
}
}
}
impl SkillInstructions {}
impl From<SkillInstructions> for ResponseItem {
fn from(si: SkillInstructions) -> Self {
ResponseItem::Message {
id: None,
role: "user".to_string(),
content: vec![ContentItem::InputText {
text: format!(
"<skill>\n<name>{}</name>\n<path>{}</path>\n{}\n</skill>",
si.name, si.path, si.contents
),
}],
end_turn: None,
phase: None,
}
SKILL_FRAGMENT.into_message(SKILL_FRAGMENT.wrap(format!(
"<name>{}</name>\n<path>{}</path>\n{}",
si.name, si.path, si.contents
)))
}
}
#[cfg(test)]
mod tests {
use super::*;
use codex_protocol::models::ContentItem;
use pretty_assertions::assert_eq;
#[test]
@@ -110,21 +84,10 @@ mod tests {
#[test]
fn test_is_user_instructions() {
assert!(UserInstructions::is_user_instructions(
&[ContentItem::InputText {
text: "# AGENTS.md instructions for test_directory\n\n<INSTRUCTIONS>\ntest_text\n</INSTRUCTIONS>".to_string(),
}]
assert!(AGENTS_MD_FRAGMENT.matches_text(
"# AGENTS.md instructions for test_directory\n\n<INSTRUCTIONS>\ntest_text\n</INSTRUCTIONS>"
));
assert!(UserInstructions::is_user_instructions(&[
ContentItem::InputText {
text: "<user_instructions>test_text</user_instructions>".to_string(),
}
]));
assert!(!UserInstructions::is_user_instructions(&[
ContentItem::InputText {
text: "test_text".to_string(),
}
]));
assert!(!AGENTS_MD_FRAGMENT.matches_text("test_text"));
}
#[test]
@@ -154,16 +117,9 @@ mod tests {
#[test]
fn test_is_skill_instructions() {
assert!(SkillInstructions::is_skill_instructions(&[
ContentItem::InputText {
text: "<skill>\n<name>demo-skill</name>\n<path>skills/demo/SKILL.md</path>\nbody\n</skill>"
.to_string(),
}
]));
assert!(!SkillInstructions::is_skill_instructions(&[
ContentItem::InputText {
text: "regular text".to_string(),
}
]));
assert!(SKILL_FRAGMENT.matches_text(
"<skill>\n<name>demo-skill</name>\n<path>skills/demo/SKILL.md</path>\nbody\n</skill>"
));
assert!(!SKILL_FRAGMENT.matches_text("regular text"));
}
}
codex-rs/core/src/lib.rs
+1
View File
@@ -27,6 +27,7 @@ pub mod config;
pub mod config_loader;
pub mod connectors;
mod context_manager;
mod contextual_user_message;
pub mod custom_prompts;
pub mod env;
mod environment_context;
codex-rs/core/src/mcp/mod.rs
+1
View File
@@ -139,6 +139,7 @@ fn codex_apps_mcp_server_config(config: &Config, auth: Option<&CodexAuth>) -> Mc
enabled_tools: None,
disabled_tools: None,
scopes: None,
oauth_resource: None,
}
}
codex-rs/core/src/mcp/skill_dependencies.rs
+5
View File
@@ -241,6 +241,7 @@ pub(crate) async fn maybe_install_mcp_dependencies(
oauth_config.http_headers,
oauth_config.env_http_headers,
&[],
server_config.oauth_resource.as_deref(),
config.mcp_oauth_callback_port,
config.mcp_oauth_callback_url.as_deref(),
)
@@ -387,6 +388,7 @@ fn mcp_dependency_to_server_config(
enabled_tools: None,
disabled_tools: None,
scopes: None,
oauth_resource: None,
});
}
@@ -411,6 +413,7 @@ fn mcp_dependency_to_server_config(
enabled_tools: None,
disabled_tools: None,
scopes: None,
oauth_resource: None,
});
}
@@ -468,6 +471,7 @@ mod tests {
enabled_tools: None,
disabled_tools: None,
scopes: None,
oauth_resource: None,
},
)]);
@@ -516,6 +520,7 @@ mod tests {
enabled_tools: None,
disabled_tools: None,
scopes: None,
oauth_resource: None,
},
)]);
codex-rs/core/src/mcp_connection_manager.rs
+2
View File
@@ -2115,6 +2115,7 @@ mod tests {
enabled_tools: None,
disabled_tools: None,
scopes: None,
oauth_resource: None,
},
auth_status: McpAuthStatus::Unsupported,
};
@@ -2162,6 +2163,7 @@ mod tests {
enabled_tools: None,
disabled_tools: None,
scopes: None,
oauth_resource: None,
},
auth_status: McpAuthStatus::Unsupported,
};
codex-rs/core/src/memories/README.md
+28 -1
View File
@@ -59,7 +59,14 @@ Phase 2 consolidates the latest stage-1 outputs into the filesystem memory artif
What it does:
- claims a single global phase-2 job (so only one consolidation runs at a time)
- loads a bounded set of the most recent stage-1 outputs from the state DB (the per-rollout memories produced by Phase 1, used as the consolidation input set)
- loads a bounded set of stage-1 outputs from the state DB using phase-2
selection rules:
- ignores memories whose `last_usage` falls outside the configured
`max_unused_days` window
- for memories with no `last_usage`, falls back to `generated_at` so fresh
never-used memories can still be selected
- ranks eligible memories by `usage_count` first, then by the most recent
`last_usage` / `generated_at`
- computes a completion watermark from the claimed watermark + newest input timestamps
- syncs local memory artifacts under the memories root:
- `raw_memories.md` (merged raw memories, latest first)
@@ -70,11 +77,31 @@ What it does:
If there is input, it then:
- spawns an internal consolidation sub-agent
- builds the Phase 2 prompt with a diff of the current Phase 1 input
selection versus the last successful Phase 2 selection (`added`,
`retained`, `removed`)
- runs it with no approvals, no network, and local write access only
- disables collab for that agent (to prevent recursive delegation)
- watches the agent status and heartbeats the global job lease while it runs
- marks the phase-2 job success/failure in the state DB when the agent finishes
Selection diff behavior:
- successful Phase 2 runs mark the exact stage-1 snapshots they consumed with
`selected_for_phase2 = 1` and persist the matching
`selected_for_phase2_source_updated_at`
- Phase 1 upserts preserve the previous `selected_for_phase2` baseline until
the next successful Phase 2 run rewrites it
- the next Phase 2 run compares the current top-N stage-1 inputs against that
prior snapshot selection to label inputs as `added` or `retained`; a
refreshed thread stays `added` until Phase 2 successfully selects its newer
snapshot
- rows that were previously selected but still exist outside the current top-N
selection are surfaced as `removed`
- before the agent starts, local `rollout_summaries/` and `raw_memories.md`
keep the union of the current selection and the previous successful
selection, so removed-thread evidence stays available during forgetting
Watermark behavior:
- The global phase-2 job claim includes an input watermark representing the latest input timestamp known when the job was claimed.
codex-rs/core/src/memories/phase2.rs
+59 -10
View File
@@ -8,6 +8,7 @@ use crate::memories::metrics;
use crate::memories::phase_two;
use crate::memories::prompts::build_consolidation_prompt;
use crate::memories::storage::rebuild_raw_memories_file_from_memories;
use crate::memories::storage::rollout_summary_file_stem;
use crate::memories::storage::sync_rollout_summaries_from_memories;
use codex_config::Constrained;
use codex_protocol::ThreadId;
@@ -17,8 +18,10 @@ use codex_protocol::protocol::SessionSource;
use codex_protocol::protocol::SubAgentSource;
use codex_protocol::protocol::TokenUsage;
use codex_protocol::user_input::UserInput;
use codex_state::Stage1Output;
use codex_state::StateRuntime;
use codex_utils_absolute_path::AbsolutePathBuf;
use std::collections::HashSet;
use std::sync::Arc;
use std::time::Duration;
use tokio::sync::watch;
@@ -50,6 +53,7 @@ pub(super) async fn run(session: &Arc<Session>, config: Arc<Config>) {
};
let root = memory_root(&config.codex_home);
let max_raw_memories = config.memories.max_raw_memories_for_global;
let max_unused_days = config.memories.max_unused_days;
// 1. Claim the job.
let claim = match job::claim(session, db).await {
@@ -73,21 +77,27 @@ pub(super) async fn run(session: &Arc<Session>, config: Arc<Config>) {
};
// 3. Query the memories
let raw_memories = match db.list_stage1_outputs_for_global(max_raw_memories).await {
Ok(memories) => memories,
let selection = match db
.get_phase2_input_selection(max_raw_memories, max_unused_days)
.await
{
Ok(selection) => selection,
Err(err) => {
tracing::error!("failed to list stage1 outputs from global: {}", err);
job::failed(session, db, &claim, "failed_load_stage1_outputs").await;
return;
}
};
let raw_memories = selection.selected.to_vec();
let artifact_memories = artifact_memories_for_phase2(&selection);
let new_watermark = get_watermark(claim.watermark, &raw_memories);
// 4. Update the file system by syncing the raw memories with the one extracted from DB at
// step 3
// [`rollout_summaries/`]
if let Err(err) =
sync_rollout_summaries_from_memories(&root, &raw_memories, max_raw_memories).await
sync_rollout_summaries_from_memories(&root, &artifact_memories, artifact_memories.len())
.await
{
tracing::error!("failed syncing local memory artifacts for global consolidation: {err}");
job::failed(session, db, &claim, "failed_sync_artifacts").await;
@@ -95,7 +105,8 @@ pub(super) async fn run(session: &Arc<Session>, config: Arc<Config>) {
}
// [`raw_memories.md`]
if let Err(err) =
rebuild_raw_memories_file_from_memories(&root, &raw_memories, max_raw_memories).await
rebuild_raw_memories_file_from_memories(&root, &artifact_memories, artifact_memories.len())
.await
{
tracing::error!("failed syncing local memory artifacts for global consolidation: {err}");
job::failed(session, db, &claim, "failed_rebuild_raw_memories").await;
@@ -103,12 +114,20 @@ pub(super) async fn run(session: &Arc<Session>, config: Arc<Config>) {
}
if raw_memories.is_empty() {
// We check only after sync of the file system.
job::succeed(session, db, &claim, new_watermark, "succeeded_no_input").await;
job::succeed(
session,
db,
&claim,
new_watermark,
&[],
"succeeded_no_input",
)
.await;
return;
}
// 5. Spawn the agent
let prompt = agent::get_prompt(config);
let prompt = agent::get_prompt(config, &selection);
let source = SessionSource::SubAgent(SubAgentSource::MemoryConsolidation);
let thread_id = match session
.services
@@ -129,6 +148,7 @@ pub(super) async fn run(session: &Arc<Session>, config: Arc<Config>) {
session,
claim,
new_watermark,
raw_memories.clone(),
thread_id,
phase_two_e2e_timer,
);
@@ -140,6 +160,22 @@ pub(super) async fn run(session: &Arc<Session>, config: Arc<Config>) {
emit_metrics(session, counters);
}
fn artifact_memories_for_phase2(
selection: &codex_state::Phase2InputSelection,
) -> Vec<Stage1Output> {
let mut seen = HashSet::new();
let mut memories = selection.selected.clone();
for memory in &selection.selected {
seen.insert(rollout_summary_file_stem(memory));
}
for memory in &selection.previous_selected {
if seen.insert(rollout_summary_file_stem(memory)) {
memories.push(memory.clone());
}
}
memories
}
mod job {
use super::*;
@@ -205,6 +241,7 @@ mod job {
db: &StateRuntime,
claim: &Claim,
completion_watermark: i64,
selected_outputs: &[codex_state::Stage1Output],
reason: &'static str,
) {
session.services.otel_manager.counter(
@@ -213,7 +250,7 @@ mod job {
&[("status", reason)],
);
let _ = db
.mark_global_phase2_job_succeeded(&claim.token, completion_watermark)
.mark_global_phase2_job_succeeded(&claim.token, completion_watermark, selected_outputs)
.await;
}
}
@@ -266,9 +303,12 @@ mod agent {
Some(agent_config)
}
pub(super) fn get_prompt(config: Arc<Config>) -> Vec<UserInput> {
pub(super) fn get_prompt(
config: Arc<Config>,
selection: &codex_state::Phase2InputSelection,
) -> Vec<UserInput> {
let root = memory_root(&config.codex_home);
let prompt = build_consolidation_prompt(&root);
let prompt = build_consolidation_prompt(&root, selection);
vec![UserInput::Text {
text: prompt,
text_elements: vec![],
@@ -280,6 +320,7 @@ mod agent {
session: &Arc<Session>,
claim: Claim,
new_watermark: i64,
selected_outputs: Vec<codex_state::Stage1Output>,
thread_id: ThreadId,
phase_two_e2e_timer: Option<codex_otel::Timer>,
) {
@@ -316,7 +357,15 @@ mod agent {
if let Some(token_usage) = agent_control.get_total_token_usage(thread_id).await {
emit_token_usage_metrics(&session, &token_usage);
}
job::succeed(&session, &db, &claim, new_watermark, "succeeded").await;
job::succeed(
&session,
&db,
&claim,
new_watermark,
&selected_outputs,
"succeeded",
)
.await;
} else {
job::failed(&session, &db, &claim, "failed_agent").await;
}
codex-rs/core/src/memories/prompts.rs
+81 -2
View File
@@ -1,9 +1,13 @@
use crate::memories::memory_root;
use crate::memories::phase_one;
use crate::memories::storage::rollout_summary_file_stem_from_parts;
use crate::truncate::TruncationPolicy;
use crate::truncate::truncate_text;
use askama::Template;
use codex_protocol::openai_models::ModelInfo;
use codex_state::Phase2InputSelection;
use codex_state::Stage1Output;
use codex_state::Stage1OutputRef;
use std::path::Path;
use tokio::fs;
use tracing::warn;
@@ -12,6 +16,7 @@ use tracing::warn;
#[template(path = "memories/consolidation.md", escape = "none")]
struct ConsolidationPromptTemplate<'a> {
memory_root: &'a str,
phase2_input_selection: &'a str,
}
#[derive(Template)]
@@ -30,17 +35,91 @@ struct MemoryToolDeveloperInstructionsTemplate<'a> {
}
/// Builds the consolidation subagent prompt for a specific memory root.
pub(super) fn build_consolidation_prompt(memory_root: &Path) -> String {
pub(super) fn build_consolidation_prompt(
memory_root: &Path,
selection: &Phase2InputSelection,
) -> String {
let memory_root = memory_root.display().to_string();
let phase2_input_selection = render_phase2_input_selection(selection);
let template = ConsolidationPromptTemplate {
memory_root: &memory_root,
phase2_input_selection: &phase2_input_selection,
};
template.render().unwrap_or_else(|err| {
warn!("failed to render memories consolidation prompt template: {err}");
format!("## Memory Phase 2 (Consolidation)\nConsolidate Codex memories in: {memory_root}")
format!(
"## Memory Phase 2 (Consolidation)\nConsolidate Codex memories in: {memory_root}\n\n{phase2_input_selection}"
)
})
}
fn render_phase2_input_selection(selection: &Phase2InputSelection) -> String {
let retained = selection.retained_thread_ids.len();
let added = selection.selected.len().saturating_sub(retained);
let selected = if selection.selected.is_empty() {
"- none".to_string()
} else {
selection
.selected
.iter()
.map(|item| {
render_selected_input_line(
item,
selection.retained_thread_ids.contains(&item.thread_id),
)
})
.collect::<Vec<_>>()
.join("\n")
};
let removed = if selection.removed.is_empty() {
"- none".to_string()
} else {
selection
.removed
.iter()
.map(render_removed_input_line)
.collect::<Vec<_>>()
.join("\n")
};
format!(
"- selected inputs this run: {}\n- newly added since the last successful Phase 2 run: {added}\n- retained from the last successful Phase 2 run: {retained}\n- removed from the last successful Phase 2 run: {}\n\nCurrent selected Phase 1 inputs:\n{selected}\n\nRemoved from the last successful Phase 2 selection:\n{removed}\n",
selection.selected.len(),
selection.removed.len(),
)
}
fn render_selected_input_line(item: &Stage1Output, retained: bool) -> String {
let status = if retained { "retained" } else { "added" };
let rollout_summary_file = format!(
"rollout_summaries/{}.md",
rollout_summary_file_stem_from_parts(
item.thread_id,
item.source_updated_at,
item.rollout_slug.as_deref(),
)
);
format!(
"- [{status}] thread_id={}, rollout_summary_file={rollout_summary_file}",
item.thread_id
)
}
fn render_removed_input_line(item: &Stage1OutputRef) -> String {
let rollout_summary_file = format!(
"rollout_summaries/{}.md",
rollout_summary_file_stem_from_parts(
item.thread_id,
item.source_updated_at,
item.rollout_slug.as_deref(),
)
);
format!(
"- thread_id={}, rollout_summary_file={rollout_summary_file}",
item.thread_id
)
}
/// Builds the stage-1 user message containing rollout metadata and content.
///
/// Large rollout payloads are truncated to 70% of the active model's effective
codex-rs/core/src/memories/storage.rs
+9
View File
@@ -82,6 +82,8 @@ async fn rebuild_raw_memories_file(
)
.map_err(raw_memories_format_error)?;
writeln!(body, "cwd: {}", memory.cwd.display()).map_err(raw_memories_format_error)?;
writeln!(body, "rollout_path: {}", memory.rollout_path.display())
.map_err(raw_memories_format_error)?;
let rollout_summary_file = format!("{}.md", rollout_summary_file_stem(memory));
writeln!(body, "rollout_summary_file: {rollout_summary_file}")
.map_err(raw_memories_format_error)?;
@@ -138,7 +140,12 @@ async fn write_rollout_summary_for_thread(
memory.source_updated_at.to_rfc3339()
)
.map_err(rollout_summary_format_error)?;
writeln!(body, "rollout_path: {}", memory.rollout_path.display())
.map_err(rollout_summary_format_error)?;
writeln!(body, "cwd: {}", memory.cwd.display()).map_err(rollout_summary_format_error)?;
if let Some(git_branch) = memory.git_branch.as_deref() {
writeln!(body, "git_branch: {git_branch}").map_err(rollout_summary_format_error)?;
}
writeln!(body).map_err(rollout_summary_format_error)?;
body.push_str(&memory.rollout_summary);
body.push('\n');
@@ -267,7 +274,9 @@ mod tests {
raw_memory: "raw memory".to_string(),
rollout_summary: "summary".to_string(),
rollout_slug: rollout_slug.map(ToString::to_string),
rollout_path: PathBuf::from("/tmp/rollout.jsonl"),
cwd: PathBuf::from("/tmp/workspace"),
git_branch: None,
generated_at: Utc.timestamp_opt(124, 0).single().expect("timestamp"),
}
}
codex-rs/core/src/memories/tests.rs
+33 -5
View File
@@ -86,7 +86,9 @@ async fn sync_rollout_summaries_and_raw_memories_file_keeps_latest_memories_only
raw_memory: "raw memory".to_string(),
rollout_summary: "short summary".to_string(),
rollout_slug: None,
rollout_path: PathBuf::from("/tmp/rollout-100.jsonl"),
cwd: PathBuf::from("/tmp/workspace"),
git_branch: None,
generated_at: Utc.timestamp_opt(101, 0).single().expect("timestamp"),
}];
@@ -135,6 +137,7 @@ async fn sync_rollout_summaries_and_raw_memories_file_keeps_latest_memories_only
assert!(raw_memories.contains("raw memory"));
assert!(raw_memories.contains(&keep_id));
assert!(raw_memories.contains("cwd: /tmp/workspace"));
assert!(raw_memories.contains("rollout_path: /tmp/rollout-100.jsonl"));
assert!(raw_memories.contains(&format!(
"rollout_summary_file: {canonical_rollout_summary_file}"
)));
@@ -150,6 +153,10 @@ async fn sync_rollout_summaries_and_raw_memories_file_keeps_latest_memories_only
.find("cwd: /tmp/workspace")
.map(|offset| thread_pos + offset)
.expect("cwd should exist after thread header");
let rollout_path_pos = raw_memories[thread_pos..]
.find("rollout_path: /tmp/rollout-100.jsonl")
.map(|offset| thread_pos + offset)
.expect("rollout_path should exist after thread header");
let file_pos = raw_memories[thread_pos..]
.find(&format!(
"rollout_summary_file: {canonical_rollout_summary_file}"
@@ -158,7 +165,8 @@ async fn sync_rollout_summaries_and_raw_memories_file_keeps_latest_memories_only
.expect("rollout_summary_file should exist after thread header");
assert!(thread_pos < updated_pos);
assert!(updated_pos < cwd_pos);
assert!(cwd_pos < file_pos);
assert!(cwd_pos < rollout_path_pos);
assert!(rollout_path_pos < file_pos);
}
#[tokio::test]
@@ -184,7 +192,9 @@ async fn sync_rollout_summaries_uses_timestamp_hash_and_sanitized_slug_filename(
raw_memory: "raw memory".to_string(),
rollout_summary: "short summary".to_string(),
rollout_slug: Some("Unsafe Slug/With Spaces & Symbols + EXTRA_LONG_12345".to_string()),
rollout_path: PathBuf::from("/tmp/rollout-200.jsonl"),
cwd: PathBuf::from("/tmp/workspace"),
git_branch: Some("feature/memory-branch".to_string()),
generated_at: Utc.timestamp_opt(201, 0).single().expect("timestamp"),
}];
@@ -239,6 +249,8 @@ async fn sync_rollout_summaries_uses_timestamp_hash_and_sanitized_slug_filename(
.await
.expect("read rollout summary");
assert!(summary.contains(&format!("thread_id: {thread_id}")));
assert!(summary.contains("rollout_path: /tmp/rollout-200.jsonl"));
assert!(summary.contains("git_branch: feature/memory-branch"));
assert!(
!tokio::fs::try_exists(&stale_unslugged_path)
.await
@@ -283,7 +295,9 @@ task_outcome: success
.to_string(),
rollout_summary: "short summary".to_string(),
rollout_slug: Some("Unsafe Slug/With Spaces & Symbols + EXTRA_LONG_12345".to_string()),
rollout_path: PathBuf::from("/tmp/rollout-200.jsonl"),
cwd: PathBuf::from("/tmp/workspace"),
git_branch: None,
generated_at: Utc.timestamp_opt(201, 0).single().expect("timestamp"),
}];
@@ -316,6 +330,12 @@ task_outcome: success
let raw_memories = tokio::fs::read_to_string(raw_memories_file(&root))
.await
.expect("read raw memories");
let summary = tokio::fs::read_to_string(
rollout_summaries_dir(&root).join(canonical_rollout_summary_file),
)
.await
.expect("read rollout summary");
assert!(summary.contains("rollout_path: /tmp/rollout-200.jsonl"));
assert!(raw_memories.contains(&format!(
"rollout_summary_file: {canonical_rollout_summary_file}"
)));
@@ -360,7 +380,9 @@ mod phase2 {
raw_memory: "raw memory".to_string(),
rollout_summary: "rollout summary".to_string(),
rollout_slug: None,
rollout_path: PathBuf::from("/tmp/rollout-summary.jsonl"),
cwd: PathBuf::from("/tmp/workspace"),
git_branch: None,
generated_at: chrono::DateTime::<Utc>::from_timestamp(source_updated_at + 1, 0)
.expect("valid generated_at timestamp"),
}
@@ -542,7 +564,7 @@ mod phase2 {
#[tokio::test]
async fn dispatch_reclaims_stale_global_lock_and_starts_consolidation() {
let harness = DispatchHarness::new().await;
harness.seed_stage1_output(100).await;
harness.seed_stage1_output(Utc::now().timestamp()).await;
let stale_claim = harness
.state_db
@@ -556,12 +578,18 @@ mod phase2 {
phase2::run(&harness.session, Arc::clone(&harness.config)).await;
let running_claim = harness
let post_dispatch_claim = harness
.state_db
.try_claim_global_phase2_job(ThreadId::new(), 3_600)
.await
.expect("claim while running");
pretty_assertions::assert_eq!(running_claim, Phase2JobClaimOutcome::SkippedRunning);
.expect("claim after stale lock dispatch");
assert!(
matches!(
post_dispatch_claim,
Phase2JobClaimOutcome::SkippedRunning | Phase2JobClaimOutcome::SkippedNotDirty
),
"stale-lock dispatch should either keep the reclaimed job running or finish it before re-claim"
);
let user_input_ops = harness.user_input_ops_count();
pretty_assertions::assert_eq!(user_input_ops, 1);
codex-rs/core/src/models_manager/model_info.rs
+3
View File
@@ -1,3 +1,4 @@
use codex_protocol::config_types::ReasoningSummary;
use codex_protocol::openai_models::ConfigShellToolType;
use codex_protocol::openai_models::ModelInfo;
use codex_protocol::openai_models::ModelInstructionsVariables;
@@ -68,10 +69,12 @@ pub(crate) fn model_info_from_slug(slug: &str) -> ModelInfo {
visibility: ModelVisibility::None,
supported_in_api: true,
priority: 99,
availability_nux: None,
upgrade: None,
base_instructions: BASE_INSTRUCTIONS.to_string(),
model_messages: local_personality_messages_for_slug(slug),
supports_reasoning_summaries: false,
default_reasoning_summary: ReasoningSummary::Auto,
support_verbosity: false,
default_verbosity: None,
apply_patch_tool_type: None,
codex-rs/core/src/realtime_conversation.rs
+9
View File
@@ -31,6 +31,7 @@ use tokio::sync::Mutex;
use tokio::task::JoinHandle;
use tracing::debug;
use tracing::error;
use tracing::info;
use tracing::warn;
const AUDIO_IN_QUEUE_CAPACITY: usize = 256;
@@ -184,6 +185,7 @@ pub(crate) async fn handle_start(
let requested_session_id = params
.session_id
.or_else(|| Some(sess.conversation_id.to_string()));
info!("starting realtime conversation");
let events_rx = match sess
.conversation
.start(api_provider, None, prompt, requested_session_id.clone())
@@ -191,11 +193,14 @@ pub(crate) async fn handle_start(
{
Ok(events_rx) => events_rx,
Err(err) => {
error!("failed to start realtime conversation: {err}");
send_conversation_error(sess, sub_id, err.to_string(), CodexErrorInfo::Other).await;
return Ok(());
}
};
info!("realtime conversation started");
sess.send_event_raw(Event {
id: sub_id.clone(),
msg: EventMsg::RealtimeConversationStarted(RealtimeConversationStartedEvent {
@@ -211,6 +216,7 @@ pub(crate) async fn handle_start(
msg,
};
while let Ok(event) = events_rx.recv().await {
debug!(conversation_id = %sess_clone.conversation_id, "received realtime conversation event");
let maybe_routed_text = match &event {
RealtimeEvent::ConversationItemAdded(item) => {
realtime_text_from_conversation_item(item)
@@ -231,6 +237,7 @@ pub(crate) async fn handle_start(
.await;
}
if let Some(()) = sess_clone.conversation.running_state().await {
info!("realtime conversation transport closed");
sess_clone
.send_event_raw(ev(EventMsg::RealtimeConversationClosed(
RealtimeConversationClosedEvent {
@@ -250,6 +257,7 @@ pub(crate) async fn handle_audio(
params: ConversationAudioParams,
) {
if let Err(err) = sess.conversation.audio_in(params.frame).await {
error!("failed to append realtime audio: {err}");
send_conversation_error(sess, sub_id, err.to_string(), CodexErrorInfo::BadRequest).await;
}
}
@@ -284,6 +292,7 @@ pub(crate) async fn handle_text(
debug!(text = %params.text, "[realtime-text] appending realtime conversation text input");
if let Err(err) = sess.conversation.text_in(params.text).await {
error!("failed to append realtime text: {err}");
send_conversation_error(sess, sub_id, err.to_string(), CodexErrorInfo::BadRequest).await;
}
}
codex-rs/core/src/rollout/policy.rs
-1
View File
@@ -141,7 +141,6 @@ fn event_msg_persistence_mode(ev: &EventMsg) -> Option<EventPersistenceMode> {
| EventMsg::TerminalInteraction(_)
| EventMsg::ExecCommandOutputDelta(_)
| EventMsg::ExecApprovalRequest(_)
| EventMsg::SkillRequestApproval(_)
| EventMsg::RequestUserInput(_)
| EventMsg::ElicitationRequest(_)
| EventMsg::ApplyPatchApprovalRequest(_)
codex-rs/core/src/rollout/recorder.rs
+123 -20
View File
@@ -326,7 +326,9 @@ impl RolloutRecorder {
else {
break;
};
if let Some(path) = select_resume_path_from_db_page(&db_page, filter_cwd) {
if let Some(path) =
select_resume_path_from_db_page(&db_page, filter_cwd, default_provider).await
{
return Ok(Some(path));
}
db_cursor = db_page.next_anchor.map(Into::into);
@@ -348,7 +350,7 @@ impl RolloutRecorder {
default_provider,
)
.await?;
if let Some(path) = select_resume_path(&page, filter_cwd) {
if let Some(path) = select_resume_path(&page, filter_cwd, default_provider).await {
return Ok(Some(path));
}
cursor = page.next_cursor;
@@ -458,6 +460,7 @@ impl RolloutRecorder {
state_db_ctx.clone(),
state_builder,
config.model_provider_id.clone(),
config.memories.generate_memories,
));
Ok(Self {
@@ -709,6 +712,7 @@ async fn rollout_writer(
state_db_ctx: Option<StateDbHandle>,
mut state_builder: Option<ThreadMetadataBuilder>,
default_provider: String,
generate_memories: bool,
) -> std::io::Result<()> {
let mut writer = file.map(|file| JsonlWriter { file });
let mut buffered_items = Vec::<RolloutItem>::new();
@@ -729,6 +733,7 @@ async fn rollout_writer(
state_db_ctx.as_deref(),
&mut state_builder,
default_provider.as_str(),
generate_memories,
)
.await?;
}
@@ -782,6 +787,7 @@ async fn rollout_writer(
state_db_ctx.as_deref(),
&mut state_builder,
default_provider.as_str(),
generate_memories,
)
.await?;
}
@@ -829,6 +835,7 @@ async fn rollout_writer(
Ok(())
}
#[allow(clippy::too_many_arguments)]
async fn write_session_meta(
mut writer: Option<&mut JsonlWriter>,
session_meta: SessionMeta,
@@ -837,6 +844,7 @@ async fn write_session_meta(
state_db_ctx: Option<&StateRuntime>,
state_builder: &mut Option<ThreadMetadataBuilder>,
default_provider: &str,
generate_memories: bool,
) -> std::io::Result<()> {
let git_info = collect_git_info(cwd).await;
let session_meta_line = SessionMetaLine {
@@ -858,6 +866,7 @@ async fn write_session_meta(
state_builder.as_ref(),
std::slice::from_ref(&rollout_item),
None,
(!generate_memories).then_some("disabled"),
)
.await;
Ok(())
@@ -886,6 +895,7 @@ async fn write_and_reconcile_items(
state_builder.as_ref(),
items,
"rollout_writer",
None,
)
.await;
Ok(())
@@ -961,35 +971,79 @@ impl From<codex_state::ThreadsPage> for ThreadsPage {
}
}
fn select_resume_path(page: &ThreadsPage, filter_cwd: Option<&Path>) -> Option<PathBuf> {
async fn select_resume_path(
page: &ThreadsPage,
filter_cwd: Option<&Path>,
default_provider: &str,
) -> Option<PathBuf> {
match filter_cwd {
Some(cwd) => page.items.iter().find_map(|item| {
if item
.cwd
.as_ref()
.is_some_and(|session_cwd| cwd_matches(session_cwd, cwd))
{
Some(item.path.clone())
} else {
None
Some(cwd) => {
for item in &page.items {
if resume_candidate_matches_cwd(
item.path.as_path(),
item.cwd.as_deref(),
cwd,
default_provider,
)
.await
{
return Some(item.path.clone());
}
}
}),
None
}
None => page.items.first().map(|item| item.path.clone()),
}
}
fn select_resume_path_from_db_page(
async fn resume_candidate_matches_cwd(
rollout_path: &Path,
cached_cwd: Option<&Path>,
cwd: &Path,
default_provider: &str,
) -> bool {
if cached_cwd.is_some_and(|session_cwd| cwd_matches(session_cwd, cwd)) {
return true;
}
if let Ok((items, _, _)) = RolloutRecorder::load_rollout_items(rollout_path).await
&& let Some(latest_turn_context_cwd) = items.iter().rev().find_map(|item| match item {
RolloutItem::TurnContext(turn_context) => Some(turn_context.cwd.as_path()),
RolloutItem::SessionMeta(_)
| RolloutItem::ResponseItem(_)
| RolloutItem::Compacted(_)
| RolloutItem::EventMsg(_) => None,
})
{
return cwd_matches(latest_turn_context_cwd, cwd);
}
metadata::extract_metadata_from_rollout(rollout_path, default_provider, None)
.await
.is_ok_and(|outcome| cwd_matches(outcome.metadata.cwd.as_path(), cwd))
}
async fn select_resume_path_from_db_page(
page: &codex_state::ThreadsPage,
filter_cwd: Option<&Path>,
default_provider: &str,
) -> Option<PathBuf> {
match filter_cwd {
Some(cwd) => page.items.iter().find_map(|item| {
if cwd_matches(item.cwd.as_path(), cwd) {
Some(item.rollout_path.clone())
} else {
None
Some(cwd) => {
for item in &page.items {
if resume_candidate_matches_cwd(
item.rollout_path.as_path(),
Some(item.cwd.as_path()),
cwd,
default_provider,
)
.await
{
return Some(item.rollout_path.clone());
}
}
}),
None
}
None => page.items.first().map(|item| item.rollout_path.clone()),
}
}
@@ -1010,8 +1064,12 @@ mod tests {
use crate::config::ConfigBuilder;
use crate::features::Feature;
use chrono::TimeZone;
use codex_protocol::config_types::ReasoningSummary as ReasoningSummaryConfig;
use codex_protocol::protocol::AgentMessageEvent;
use codex_protocol::protocol::AskForApproval;
use codex_protocol::protocol::EventMsg;
use codex_protocol::protocol::SandboxPolicy;
use codex_protocol::protocol::TurnContextItem;
use codex_protocol::protocol::UserMessageEvent;
use pretty_assertions::assert_eq;
use std::fs::File;
@@ -1320,4 +1378,49 @@ mod tests {
assert_eq!(repaired_path, Some(real_path));
Ok(())
}
#[tokio::test]
async fn resume_candidate_matches_cwd_reads_latest_turn_context() -> std::io::Result<()> {
let home = TempDir::new().expect("temp dir");
let stale_cwd = home.path().join("stale");
let latest_cwd = home.path().join("latest");
fs::create_dir_all(&stale_cwd)?;
fs::create_dir_all(&latest_cwd)?;
let path = write_session_file(home.path(), "2025-01-03T13-00-00", Uuid::from_u128(9012))?;
let mut file = std::fs::OpenOptions::new().append(true).open(&path)?;
let turn_context = RolloutLine {
timestamp: "2025-01-03T13:00:01Z".to_string(),
item: RolloutItem::TurnContext(TurnContextItem {
turn_id: Some("turn-1".to_string()),
cwd: latest_cwd.clone(),
current_date: None,
timezone: None,
approval_policy: AskForApproval::Never,
sandbox_policy: SandboxPolicy::new_read_only_policy(),
network: None,
model: "test-model".to_string(),
personality: None,
collaboration_mode: None,
effort: None,
summary: ReasoningSummaryConfig::Auto,
user_instructions: None,
developer_instructions: None,
final_output_json_schema: None,
truncation_policy: None,
}),
};
writeln!(file, "{}", serde_json::to_string(&turn_context)?)?;
assert!(
resume_candidate_matches_cwd(
path.as_path(),
Some(stale_cwd.as_path()),
latest_cwd.as_path(),
"test-provider",
)
.await
);
Ok(())
}
}
codex-rs/core/src/rollout/truncation.rs
+1 -1
View File
@@ -193,7 +193,7 @@ mod tests {
#[tokio::test]
async fn ignores_session_prefix_messages_when_truncating_rollout_from_start() {
let (session, turn_context) = make_session_and_context().await;
let mut items = session.build_initial_context(&turn_context).await;
let mut items = session.build_initial_context(&turn_context, None).await;
items.push(user_msg("feature request"));
items.push(assistant_msg("ack"));
items.push(user_msg("second question"));
codex-rs/core/src/sandboxing/mod.rs
+9 -2
View File
@@ -17,7 +17,7 @@ use crate::protocol::SandboxPolicy;
#[cfg(target_os = "macos")]
use crate::seatbelt::MACOS_PATH_TO_SEATBELT_EXECUTABLE;
#[cfg(target_os = "macos")]
use crate::seatbelt::create_seatbelt_command_args;
use crate::seatbelt::create_seatbelt_command_args_with_extensions;
#[cfg(target_os = "macos")]
use crate::spawn::CODEX_SANDBOX_ENV_VAR;
use crate::spawn::CODEX_SANDBOX_NETWORK_DISABLED_ENV_VAR;
@@ -25,6 +25,8 @@ use crate::tools::sandboxing::SandboxablePreference;
use codex_network_proxy::NetworkProxy;
use codex_protocol::config_types::WindowsSandboxLevel;
use codex_protocol::models::FileSystemPermissions;
#[cfg(target_os = "macos")]
use codex_protocol::models::MacOsSeatbeltProfileExtensions;
use codex_protocol::models::PermissionProfile;
pub use codex_protocol::models::SandboxPermissions;
use codex_protocol::protocol::ReadOnlyAccess;
@@ -73,6 +75,8 @@ pub(crate) struct SandboxTransformRequest<'a> {
// to make shared ownership explicit across runtime/sandbox plumbing.
pub network: Option<&'a NetworkProxy>,
pub sandbox_policy_cwd: &'a Path,
#[cfg(target_os = "macos")]
pub macos_seatbelt_profile_extensions: Option<&'a MacOsSeatbeltProfileExtensions>,
pub codex_linux_sandbox_exe: Option<&'a PathBuf>,
pub use_linux_sandbox_bwrap: bool,
pub windows_sandbox_level: WindowsSandboxLevel,
@@ -342,6 +346,8 @@ impl SandboxManager {
enforce_managed_network,
network,
sandbox_policy_cwd,
#[cfg(target_os = "macos")]
macos_seatbelt_profile_extensions,
codex_linux_sandbox_exe,
use_linux_sandbox_bwrap,
windows_sandbox_level,
@@ -370,12 +376,13 @@ impl SandboxManager {
SandboxType::MacosSeatbelt => {
let mut seatbelt_env = HashMap::new();
seatbelt_env.insert(CODEX_SANDBOX_ENV_VAR.to_string(), "seatbelt".to_string());
let mut args = create_seatbelt_command_args(
let mut args = create_seatbelt_command_args_with_extensions(
command.clone(),
&effective_policy,
sandbox_policy_cwd,
enforce_managed_network,
network,
macos_seatbelt_profile_extensions,
);
let mut full_command = Vec::with_capacity(1 + args.len());
full_command.push(MACOS_PATH_TO_SEATBELT_EXECUTABLE.to_string());
codex-rs/core/src/seatbelt_permissions.rs
+17 -41
View File
@@ -3,34 +3,9 @@
use std::collections::BTreeSet;
use std::path::PathBuf;
#[allow(dead_code)]
#[derive(Debug, Clone, PartialEq, Eq, Default)]
pub enum MacOsPreferencesPermission {
// IMPORTANT: ReadOnly needs to be the default because it's the security-sensitive default.
// it's important for allowing cf prefs to work.
#[default]
ReadOnly,
ReadWrite,
None,
}
#[allow(dead_code)]
#[derive(Debug, Clone, PartialEq, Eq, Default)]
pub enum MacOsAutomationPermission {
#[default]
None,
All,
BundleIds(Vec<String>),
}
#[allow(dead_code)]
#[derive(Debug, Clone, PartialEq, Eq, Default)]
pub struct MacOsSeatbeltProfileExtensions {
pub macos_preferences: MacOsPreferencesPermission,
pub macos_automation: MacOsAutomationPermission,
pub macos_accessibility: bool,
pub macos_calendar: bool,
}
pub use codex_protocol::models::MacOsAutomationPermission;
pub use codex_protocol::models::MacOsPreferencesPermission;
pub use codex_protocol::models::MacOsSeatbeltProfileExtensions;
#[derive(Debug, Clone, PartialEq, Eq, Default)]
pub(crate) struct SeatbeltExtensionPolicy {
@@ -38,25 +13,26 @@ pub(crate) struct SeatbeltExtensionPolicy {
pub(crate) dir_params: Vec<(String, PathBuf)>,
}
impl MacOsSeatbeltProfileExtensions {
pub fn normalized(&self) -> Self {
let mut normalized = self.clone();
if let MacOsAutomationPermission::BundleIds(bundle_ids) = &self.macos_automation {
let bundle_ids = normalize_bundle_ids(bundle_ids);
normalized.macos_automation = if bundle_ids.is_empty() {
MacOsAutomationPermission::None
} else {
MacOsAutomationPermission::BundleIds(bundle_ids)
};
}
normalized
fn normalized_extensions(
extensions: &MacOsSeatbeltProfileExtensions,
) -> MacOsSeatbeltProfileExtensions {
let mut normalized = extensions.clone();
if let MacOsAutomationPermission::BundleIds(bundle_ids) = &extensions.macos_automation {
let bundle_ids = normalize_bundle_ids(bundle_ids);
normalized.macos_automation = if bundle_ids.is_empty() {
MacOsAutomationPermission::None
} else {
MacOsAutomationPermission::BundleIds(bundle_ids)
};
}
normalized
}
pub(crate) fn build_seatbelt_extensions(
extensions: &MacOsSeatbeltProfileExtensions,
) -> SeatbeltExtensionPolicy {
let extensions = extensions.normalized();
let extensions = normalized_extensions(extensions);
let mut clauses = Vec::new();
match extensions.macos_preferences {
codex-rs/core/src/session_prefix.rs
+8 -35
View File
@@ -1,28 +1,8 @@
use codex_protocol::protocol::AgentStatus;
/// Helpers for identifying model-visible "session prefix" messages.
///
/// A session prefix is a user-role message that carries configuration or state needed by
/// follow-up turns (e.g. `<environment_context>`, `<turn_aborted>`). These items are persisted in
/// history so the model can see them, but they are not user intent and must not create user-turn
/// boundaries.
pub(crate) const ENVIRONMENT_CONTEXT_OPEN_TAG: &str = "<environment_context>";
pub(crate) const TURN_ABORTED_OPEN_TAG: &str = "<turn_aborted>";
pub(crate) const SUBAGENT_NOTIFICATION_OPEN_TAG: &str = "<subagent_notification>";
pub(crate) const SUBAGENT_NOTIFICATION_CLOSE_TAG: &str = "</subagent_notification>";
fn starts_with_ascii_case_insensitive(text: &str, prefix: &str) -> bool {
text.get(..prefix.len())
.is_some_and(|candidate| candidate.eq_ignore_ascii_case(prefix))
}
/// Returns true if `text` starts with a session prefix marker (case-insensitive).
pub(crate) fn is_session_prefix(text: &str) -> bool {
let trimmed = text.trim_start();
starts_with_ascii_case_insensitive(trimmed, ENVIRONMENT_CONTEXT_OPEN_TAG)
|| starts_with_ascii_case_insensitive(trimmed, TURN_ABORTED_OPEN_TAG)
|| starts_with_ascii_case_insensitive(trimmed, SUBAGENT_NOTIFICATION_OPEN_TAG)
}
/// Helpers for model-visible session state markers that are stored in user-role
/// messages but are not user intent.
use crate::contextual_user_message::SUBAGENT_NOTIFICATION_FRAGMENT;
pub(crate) fn format_subagent_notification_message(agent_id: &str, status: &AgentStatus) -> String {
let payload_json = serde_json::json!({
@@ -30,19 +10,12 @@ pub(crate) fn format_subagent_notification_message(agent_id: &str, status: &Agen
"status": status,
})
.to_string();
format!("{SUBAGENT_NOTIFICATION_OPEN_TAG}\n{payload_json}\n{SUBAGENT_NOTIFICATION_CLOSE_TAG}")
SUBAGENT_NOTIFICATION_FRAGMENT.wrap(payload_json)
}
#[cfg(test)]
mod tests {
use super::*;
use pretty_assertions::assert_eq;
#[test]
fn is_session_prefix_is_case_insensitive() {
assert_eq!(
is_session_prefix("<SUBAGENT_NOTIFICATION>{}</subagent_notification>"),
true
);
pub(crate) fn format_subagent_context_line(agent_id: &str, agent_nickname: Option<&str>) -> String {
match agent_nickname.filter(|nickname| !nickname.is_empty()) {
Some(agent_nickname) => format!("- {agent_id}: {agent_nickname}"),
None => format!("- {agent_id}"),
}
}

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