Merge branch 'main' into fix-readme-agents-md

.github/actions/codex/bun.lock

@@ -11,7 +11,7 @@
         "@types/bun": "^1.2.19",
         "@types/node": "^24.1.0",
         "prettier": "^3.6.2",
-        "typescript": "^5.8.3",
+        "typescript": "^5.9.2",
       },
     },
   },
@@ -68,7 +68,7 @@
 
     "tunnel": ["tunnel@0.0.6", "", {}, "sha512-1h/Lnq9yajKY2PEbBadPXj3VxsDDu844OnaAo52UVmIzIvwwtBPIuNvkjuzBlTWpfJyUbG3ez0KSBibQkj4ojg=="],
 
-    "typescript": ["typescript@5.8.3", "", { "bin": { "tsc": "bin/tsc", "tsserver": "bin/tsserver" } }, "sha512-p1diW6TqL9L07nNxvRMM7hMMw4c5XOo/1ibL4aAIGmSAt9slTE1Xgw5KWuof2uTOvCg9BY7ZRi+GaF+7sfgPeQ=="],
+    "typescript": ["typescript@5.9.2", "", { "bin": { "tsc": "bin/tsc", "tsserver": "bin/tsserver" } }, "sha512-CWBzXQrc/qOkhidw1OzBTQuYRbfyxDXJMVJ1XNwUHGROVmuaeiEm3OslpZ1RV96d7SKKjZKrSJu3+t/xlw3R9A=="],
 
     "undici": ["undici@5.29.0", "", { "dependencies": { "@fastify/busboy": "^2.0.0" } }, "sha512-raqeBD6NQK4SkWhQzeYKd1KmIG6dllBOTt55Rmkt4HtI9mwdWtJljnrXjAFUBLTSN67HWrOIZ3EPF4kjUw80Bg=="],
 

.github/actions/codex/package.json

@@ -16,6 +16,6 @@
         "@types/bun": "^1.2.19",
         "@types/node": "^24.1.0",
         "prettier": "^3.6.2",
-        "typescript": "^5.8.3"
+        "typescript": "^5.9.2"
     }
 }

.github/actions/codex/src/process-label.ts

@@ -91,7 +91,38 @@ async function processLabel(
   labelConfig: LabelConfig,
 ): Promise<void> {
   const template = labelConfig.getPromptTemplate();
-  const populatedTemplate = await renderPromptTemplate(template, ctx);
+
+  // If this is a review label, prepend explicit PR-diff scoping guidance to
+  // reduce out-of-scope feedback. Do this before rendering so placeholders in
+  // the guidance (e.g., {CODEX_ACTION_GITHUB_EVENT_PATH}) are substituted.
+  const isReview = label.toLowerCase().includes("review");
+  const reviewScopeGuidance = `
+PR Diff Scope
+- Only review changes between the PR's merge-base and head; do not comment on commits or files outside this range.
+- Derive the base/head SHAs from the event JSON at {CODEX_ACTION_GITHUB_EVENT_PATH}, then compute and use the PR diff for all analysis and comments.
+
+Commands to determine scope
+- Resolve SHAs:
+  - BASE_SHA=$(jq -r '.pull_request.base.sha // .pull_request.base.ref' "{CODEX_ACTION_GITHUB_EVENT_PATH}")
+  - HEAD_SHA=$(jq -r '.pull_request.head.sha // .pull_request.head.ref' "{CODEX_ACTION_GITHUB_EVENT_PATH}")
+  - BASE_SHA=$(git rev-parse "$BASE_SHA")
+  - HEAD_SHA=$(git rev-parse "$HEAD_SHA")
+- Prefer triple-dot (merge-base) semantics for PR diffs:
+  - Changed commits: git log --oneline "$BASE_SHA...$HEAD_SHA"
+  - Changed files: git diff --name-status "$BASE_SHA...$HEAD_SHA"
+  - Review hunks: git diff -U0 "$BASE_SHA...$HEAD_SHA"
+
+Review rules
+- Anchor every comment to a file and hunk present in git diff "$BASE_SHA...$HEAD_SHA".
+- If you mention context outside the diff, label it as "Follow-up (outside this PR scope)" and keep it brief (<=2 bullets).
+- Do not critique commits or files not reachable in the PR range (merge-base(base, head) → head).
+`.trim();
+
+  const effectiveTemplate = isReview
+    ? `${reviewScopeGuidance}\n\n${template}`
+    : template;
+
+  const populatedTemplate = await renderPromptTemplate(effectiveTemplate, ctx);
 
   // Always run Codex and post the resulting message as a comment.
   let commentBody = await runCodex(populatedTemplate, ctx);

.github/codex/labels/codex-rust-review.md

@@ -6,18 +6,134 @@ Then provide the **review** (1-2 sentences plus bullet points, friendly tone).
 
 Things to look out for when doing the review:
 
+## General Principles
+
 - **Make sure the pull request body explains the motivation behind the change.** If the author has failed to do this, call it out, and if you think you can deduce the motivation behind the change, propose copy.
 - Ideally, the PR body also contains a small summary of the change. For small changes, the PR title may be sufficient.
 - Each PR should ideally do one conceptual thing. For example, if a PR does a refactoring as well as introducing a new feature, push back and suggest the refactoring be done in a separate PR. This makes things easier for the reviewer, as refactoring changes can often be far-reaching, yet quick to review.
-- If the nature of the change seems to have a visual component (which is often the case for changes to `codex-rs/tui`), recommend including a screenshot or video to demonstrate the change, if appropriate.
-- Rust files should generally be organized such that the public parts of the API appear near the top of the file and helper functions go below. This is analagous to the "inverted pyramid" structure that is favored in journalism.
-- Encourage the use of small enums or the newtype pattern in Rust if it helps readability without adding significant cognitive load or lines of code.
-- Be wary of large files and offer suggestions for how to break things into more reasonably-sized files.
-- When modifying a `Cargo.toml` file, make sure that dependency lists stay alphabetically sorted. Also consider whether a new dependency is added to the appropriate place (e.g., `[dependencies]` versus `[dev-dependencies]`)
-- If you see opportunities for the changes in a diff to use more idiomatic Rust, please make specific recommendations. For example, favor the use of expressions over `return`.
 - When introducing new code, be on the lookout for code that duplicates existing code. When found, propose a way to refactor the existing code such that it should be reused.
+
+## Code Organization
+
 - Each create in the Cargo workspace in `codex-rs` has a specific purpose: make a note if you believe new code is not introduced in the correct crate.
 - When possible, try to keep the `core` crate as small as possible. Non-core but shared logic is often a good candidate for `codex-rs/common`.
+- Be wary of large files and offer suggestions for how to break things into more reasonably-sized files.
+- Rust files should generally be organized such that the public parts of the API appear near the top of the file and helper functions go below. This is analagous to the "inverted pyramid" structure that is favored in journalism.
+
+## Assertions in Tests
+
+Assert the equality of the entire objects instead of doing "piecemeal comparisons," performing `assert_eq!()` on individual fields.
+
+Note that unit tests also function as "executable documentation." As shown in the following example, "piecemeal comparisons" are often more verbose, provide less coverage, and are not as useful as executable documentation.
+
+For example, suppose you have the following enum:
+
+```rust
+#[derive(Debug, PartialEq)]
+enum Message {
+    Request {
+        id: String,
+        method: String,
+        params: Option<serde_json::Value>,
+    },
+    Notification {
+        method: String,
+        params: Option<serde_json::Value>,
+    },
+}
+```
+
+This is an example of a _piecemeal_ comparison:
+
+```rust
+// BAD: Piecemeal Comparison
+
+#[test]
+fn test_get_latest_messages() {
+    let messages = get_latest_messages();
+    assert_eq!(messages.len(), 2);
+
+    let m0 = &messages[0];
+    match m0 {
+        Message::Request { id, method, params } => {
+            assert_eq!(id, "123");
+            assert_eq!(method, "subscribe");
+            assert_eq!(
+                *params,
+                Some(json!({
+                    "conversation_id": "x42z86"
+                }))
+            )
+        }
+        Message::Notification { .. } => {
+            panic!("expected Request");
+        }
+    }
+
+    let m1 = &messages[1];
+    match m1 {
+        Message::Request { .. } => {
+            panic!("expected Notification");
+        }
+        Message::Notification { method, params } => {
+            assert_eq!(method, "log");
+            assert_eq!(
+                *params,
+                Some(json!({
+                    "level": "info",
+                    "message": "subscribed"
+                }))
+            )
+        }
+    }
+}
+```
+
+This is a _deep_ comparison:
+
+```rust
+// GOOD: Verify the entire structure with a single assert_eq!().
+
+use pretty_assertions::assert_eq;
+
+#[test]
+fn test_get_latest_messages() {
+    let messages = get_latest_messages();
+
+    assert_eq!(
+        vec![
+            Message::Request {
+                id: "123".to_string(),
+                method: "subscribe".to_string(),
+                params: Some(json!({
+                    "conversation_id": "x42z86"
+                })),
+            },
+            Message::Notification {
+                method: "log".to_string(),
+                params: Some(json!({
+                    "level": "info",
+                    "message": "subscribed"
+                })),
+            },
+        ],
+        messages,
+    );
+}
+```
+
+## More Tactical Rust Things To Look Out For
+
+- Do not use `unsafe` (unless you have a really, really good reason like using an operating system API directly and no safe wrapper exists). For example, there are cases where it is tempting to use `unsafe` in order to use `std::env::set_var()`, but this indeed `unsafe` and has led to race conditions on multiple occasions. (When this happens, find a mechanism other than environment variables to use for configuration.)
+- Encourage the use of small enums or the newtype pattern in Rust if it helps readability without adding significant cognitive load or lines of code.
+- If you see opportunities for the changes in a diff to use more idiomatic Rust, please make specific recommendations. For example, favor the use of expressions over `return`.
+- When modifying a `Cargo.toml` file, make sure that dependency lists stay alphabetically sorted. Also consider whether a new dependency is added to the appropriate place (e.g., `[dependencies]` versus `[dev-dependencies]`)
+
+## Pull Request Body
+
+- If the nature of the change seems to have a visual component (which is often the case for changes to `codex-rs/tui`), recommend including a screenshot or video to demonstrate the change, if appropriate.
 - References to existing GitHub issues and PRs are encouraged, where appropriate, though you likely do not have network access, so may not be able to help here.
 
+# PR Information
+
 {CODEX_ACTION_GITHUB_EVENT_PATH} contains the JSON that triggered this GitHub workflow. It contains the `base` and `head` refs that define this PR. Both refs are available locally.

.github/workflows/rust-release.yml

@@ -181,9 +181,9 @@ jobs:
           name: ${{ steps.release_name.outputs.name }}
           tag_name: ${{ github.ref_name }}
           files: dist/**
-          # For now, tag releases as "prerelease" because we are not claiming
-          # the Rust CLI is stable yet.
-          prerelease: true
+          # Mark as prerelease only when the version has a suffix after x.y.z
+          # (e.g. -alpha, -beta). Otherwise publish a normal release.
+          prerelease: ${{ contains(steps.release_name.outputs.name, '-') }}
 
       - uses: facebook/dotslash-publish-release@v2
         env:

.vscode/settings.json

@@ -11,6 +11,8 @@
         "editor.defaultFormatter": "tamasfe.even-better-toml",
         "editor.formatOnSave": true,
     },
-    "evenBetterToml.formatter.reorderArrays": true,
+    // Array order for options in ~/.codex/config.toml such as `notify` and the
+    // `args` for an MCP server is significant, so we disable reordering.
+    "evenBetterToml.formatter.reorderArrays": false,
     "evenBetterToml.formatter.reorderKeys": true,
 }

INIT.md

@@ -0,0 +1,40 @@
+Generate a file named AGENTS.md that serves as a contributor guide for this repository.
+Your goal is to produce a clear, concise, and well-structured document with descriptive headings and actionable explanations for each section.
+Follow the outline below, but adapt as needed — add sections if relevant, and omit those that do not apply to this project.
+
+Document Requirements
+
+- Title the document "Repository Guidelines".
+- Use Markdown headings (#, ##, etc.) for structure.
+- Keep the document concise. 200-400 words is optimal.
+- Keep explanations short, direct, and specific to this repository.
+- Provide examples where helpful (commands, directory paths, naming patterns).
+- Maintain a professional, instructional tone.
+
+Recommended Sections
+
+Project Structure & Module Organization
+
+- Outline the project structure, including where the source code, tests, and assets are located.
+
+Build, Test, and Development Commands
+
+- List key commands for building, testing, and running locally (e.g., npm test, make build).
+- Briefly explain what each command does.
+
+Coding Style & Naming Conventions
+
+- Specify indentation rules, language-specific style preferences, and naming patterns.
+- Include any formatting or linting tools used.
+
+Testing Guidelines
+
+- Identify testing frameworks and coverage requirements.
+- State test naming conventions and how to run tests.
+
+Commit & Pull Request Guidelines
+
+- Summarize commit message conventions found in the project’s Git history.
+- Outline pull request requirements (descriptions, linked issues, screenshots, etc.).
+
+(Optional) Add other sections if relevant, such as Security & Configuration Tips, Architecture Overview, or Agent-Specific Instructions.

README.md

@@ -18,6 +18,7 @@ This is the home of the **Codex CLI**, which is a coding agent from OpenAI that
 - [Quickstart](#quickstart)
   - [OpenAI API Users](#openai-api-users)
   - [OpenAI Plus/Pro Users](#openai-pluspro-users)
+- [Using Open Source Models](#using-open-source-models)
 - [Why Codex?](#why-codex)
 - [Security model & permissions](#security-model--permissions)
   - [Platform sandboxing details](#platform-sandboxing-details)
@@ -186,6 +187,41 @@ they'll be committed to your working directory.
 
 ---
 
+## Using Open Source Models
+
+Codex can run fully locally against an OpenAI-compatible OSS host (like Ollama) using the `--oss` flag:
+
+- Interactive UI:
+  - codex --oss
+- Non-interactive (programmatic) mode:
+  - echo "Refactor utils" | codex exec --oss
+
+Model selection when using `--oss`:
+
+- If you omit `-m/--model`, Codex defaults to -m gpt-oss:20b and will verify it exists locally (downloading if needed).
+- To pick a different size, pass one of:
+  - -m "gpt-oss:20b"
+  - -m "gpt-oss:120b"
+
+Point Codex at your own OSS host:
+
+- By default, `--oss` talks to http://localhost:11434/v1.
+- To use a different host, set one of these environment variables before running Codex:
+  - CODEX_OSS_BASE_URL, for example:
+    - CODEX_OSS_BASE_URL="http://my-ollama.example.com:11434/v1" codex --oss -m gpt-oss:20b
+  - or CODEX_OSS_PORT (when the host is localhost):
+    - CODEX_OSS_PORT=11434 codex --oss
+
+Advanced: you can persist this in your config instead of environment variables by overriding the built-in `oss` provider in `~/.codex/config.toml`:
+
+```toml
+[model_providers.oss]
+name = "Open Source"
+base_url = "http://my-ollama.example.com:11434/v1"
+```
+
+---
+
 ## Why Codex?
 
 Codex CLI is built for developers who already **live in the terminal** and want

codex-rs/Cargo.lock

@@ -661,7 +661,7 @@ dependencies = [
  "clap",
  "codex-core",
  "serde",
- "toml 0.9.2",
+ "toml 0.9.4",
 ]
 
 [[package]]
@@ -707,7 +707,7 @@ dependencies = [
  "tokio",
  "tokio-test",
  "tokio-util",
- "toml 0.9.2",
+ "toml 0.9.4",
  "tracing",
  "tree-sitter",
  "tree-sitter-bash",
@@ -729,6 +729,7 @@ dependencies = [
  "codex-arg0",
  "codex-common",
  "codex-core",
+ "codex-ollama",
  "owo-colors",
  "predicates",
  "serde_json",
@@ -831,13 +832,30 @@ dependencies = [
  "tempfile",
  "tokio",
  "tokio-test",
- "toml 0.9.2",
+ "toml 0.9.4",
  "tracing",
  "tracing-subscriber",
  "uuid",
  "wiremock",
 ]
 
+[[package]]
+name = "codex-ollama"
+version = "0.0.0"
+dependencies = [
+ "async-stream",
+ "bytes",
+ "codex-core",
+ "futures",
+ "reqwest",
+ "serde_json",
+ "tempfile",
+ "tokio",
+ "toml 0.9.4",
+ "tracing",
+ "wiremock",
+]
+
 [[package]]
 name = "codex-tui"
 version = "0.0.0"
@@ -852,6 +870,7 @@ dependencies = [
  "codex-core",
  "codex-file-search",
  "codex-login",
+ "codex-ollama",
  "color-eyre",
  "crossterm",
  "image",
@@ -881,6 +900,7 @@ dependencies = [
  "unicode-segmentation",
  "unicode-width 0.1.14",
  "uuid",
+ "vt100",
 ]
 
 [[package]]
@@ -1473,7 +1493,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "778e2ac28f6c47af28e4907f13ffd1e1ddbd400980a9abd7c8df189bf578a5ad"
 dependencies = [
  "libc",
- "windows-sys 0.60.2",
+ "windows-sys 0.52.0",
 ]
 
 [[package]]
@@ -1553,7 +1573,7 @@ checksum = "0ce92ff622d6dadf7349484f42c93271a0d49b7cc4d466a936405bacbe10aa78"
 dependencies = [
  "cfg-if",
  "rustix 1.0.8",
- "windows-sys 0.59.0",
+ "windows-sys 0.52.0",
 ]
 
 [[package]]
@@ -1756,7 +1776,7 @@ version = "0.2.23"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "cba6ae63eb948698e300f645f87c70f76630d505f23b8907cf1e193ee85048c1"
 dependencies = [
- "unicode-width 0.2.0",
+ "unicode-width 0.2.1",
 ]
 
 [[package]]
@@ -2017,7 +2037,7 @@ dependencies = [
  "libc",
  "percent-encoding",
  "pin-project-lite",
- "socket2 0.6.0",
+ "socket2",
  "system-configuration",
  "tokio",
  "tower-service",
@@ -2336,7 +2356,7 @@ checksum = "e04d7f318608d35d4b61ddd75cbdaee86b023ebe2bd5a66ee0915f0bf93095a9"
 dependencies = [
  "hermit-abi",
  "libc",
- "windows-sys 0.59.0",
+ "windows-sys 0.52.0",
 ]
 
 [[package]]
@@ -3392,7 +3412,7 @@ dependencies = [
 [[package]]
 name = "ratatui"
 version = "0.29.0"
-source = "git+https://github.com/nornagon/ratatui?branch=nornagon-v0.29.0-patch#bca287ddc5d38fe088c79e2eda22422b96226f2e"
+source = "git+https://github.com/nornagon/ratatui?branch=nornagon-v0.29.0-patch#9b2ad1298408c45918ee9f8241a6f95498cdbed2"
 dependencies = [
  "bitflags 2.9.1",
  "cassowary",
@@ -3406,7 +3426,7 @@ dependencies = [
  "strum 0.26.3",
  "unicode-segmentation",
  "unicode-truncate",
- "unicode-width 0.2.0",
+ "unicode-width 0.2.1",
 ]
 
 [[package]]
@@ -3720,7 +3740,7 @@ dependencies = [
  "errno",
  "libc",
  "linux-raw-sys 0.4.15",
- "windows-sys 0.59.0",
+ "windows-sys 0.52.0",
 ]
 
 [[package]]
@@ -3733,7 +3753,7 @@ dependencies = [
  "errno",
  "libc",
  "linux-raw-sys 0.9.4",
- "windows-sys 0.60.2",
+ "windows-sys 0.52.0",
 ]
 
 [[package]]
@@ -3997,9 +4017,9 @@ dependencies = [
 
 [[package]]
 name = "serde_json"
-version = "1.0.141"
+version = "1.0.142"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "30b9eff21ebe718216c6ec64e1d9ac57087aad11efc64e32002bce4a0d4c03d3"
+checksum = "030fedb782600dcbd6f02d479bf0d817ac3bb40d644745b769d6a96bc3afc5a7"
 dependencies = [
  "indexmap 2.10.0",
  "itoa",
@@ -4188,16 +4208,6 @@ version = "0.3.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b7c388c1b5e93756d0c740965c41e8822f866621d41acbdf6336a6a168f8840c"
 
-[[package]]
-name = "socket2"
-version = "0.5.10"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e22376abed350d73dd1cd119b57ffccad95b4e585a7cda43e286245ce23c0678"
-dependencies = [
- "libc",
- "windows-sys 0.52.0",
-]
-
 [[package]]
 name = "socket2"
 version = "0.6.0"
@@ -4509,7 +4519,7 @@ dependencies = [
  "getrandom 0.3.3",
  "once_cell",
  "rustix 1.0.8",
- "windows-sys 0.59.0",
+ "windows-sys 0.52.0",
 ]
 
 [[package]]
@@ -4556,7 +4566,7 @@ checksum = "c13547615a44dc9c452a8a534638acdf07120d4b6847c8178705da06306a3057"
 dependencies = [
  "smawk",
  "unicode-linebreak",
- "unicode-width 0.2.0",
+ "unicode-width 0.2.1",
 ]
 
 [[package]]
@@ -4673,9 +4683,9 @@ dependencies = [
 
 [[package]]
 name = "tokio"
-version = "1.46.1"
+version = "1.47.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0cc3a2344dafbe23a245241fe8b09735b521110d30fcefbbd5feb1797ca35d17"
+checksum = "89e49afdadebb872d3145a5638b59eb0691ea23e46ca484037cfab3b76b95038"
 dependencies = [
  "backtrace",
  "bytes",
@@ -4686,9 +4696,9 @@ dependencies = [
  "pin-project-lite",
  "signal-hook-registry",
  "slab",
- "socket2 0.5.10",
+ "socket2",
  "tokio-macros",
- "windows-sys 0.52.0",
+ "windows-sys 0.59.0",
 ]
 
 [[package]]
@@ -4773,9 +4783,9 @@ dependencies = [
 
 [[package]]
 name = "toml"
-version = "0.9.2"
+version = "0.9.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ed0aee96c12fa71097902e0bb061a5e1ebd766a6636bb605ba401c45c1650eac"
+checksum = "41ae868b5a0f67631c14589f7e250c1ea2c574ee5ba21c6c8dd4b1485705a5a1"
 dependencies = [
  "indexmap 2.10.0",
  "serde",
@@ -5004,7 +5014,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "911e93158bf80bbc94bad533b2b16e3d711e1132d69a6a6980c3920a63422c19"
 dependencies = [
  "ratatui",
- "unicode-width 0.2.0",
+ "unicode-width 0.2.1",
 ]
 
 [[package]]
@@ -5072,9 +5082,9 @@ checksum = "7dd6e30e90baa6f72411720665d41d89b9a3d039dc45b8faea1ddd07f617f6af"
 
 [[package]]
 name = "unicode-width"
-version = "0.2.0"
+version = "0.2.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1fc81956842c57dac11422a97c3b8195a1ff727f06e85c84ed2e8aa277c9a0fd"
+checksum = "4a1a07cc7db3810833284e8d372ccdc6da29741639ecc70c9ec107df0fa6154c"
 
 [[package]]
 name = "unicode-xid"
@@ -5159,6 +5169,27 @@ version = "0.9.5"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "0b928f33d975fc6ad9f86c8f283853ad26bdd5b10b7f1542aa2fa15e2289105a"
 
+[[package]]
+name = "vt100"
+version = "0.16.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "054ff75fb8fa83e609e685106df4faeffdf3a735d3c74ebce97ec557d5d36fd9"
+dependencies = [
+ "itoa",
+ "unicode-width 0.2.1",
+ "vte",
+]
+
+[[package]]
+name = "vte"
+version = "0.15.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "a5924018406ce0063cd67f8e008104968b74b563ee1b85dde3ed1f7cb87d3dbd"
+dependencies = [
+ "arrayvec",
+ "memchr",
+]
+
 [[package]]
 name = "wait-timeout"
 version = "0.2.1"
@@ -5347,7 +5378,7 @@ version = "0.1.9"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "cf221c93e13a30d793f7645a0e7762c55d169dbb0a49671918a2319d289b10bb"
 dependencies = [
- "windows-sys 0.59.0",
+ "windows-sys 0.52.0",
 ]
 
 [[package]]

codex-rs/Cargo.toml

@@ -14,6 +14,7 @@ members = [
     "mcp-client",
     "mcp-server",
     "mcp-types",
+    "ollama",
     "tui",
 ]
 resolver = "2"

codex-rs/common/src/approval_mode_cli_arg.rs

@@ -18,6 +18,9 @@ pub enum ApprovalModeCliArg {
     /// will escalate to the user to ask for un-sandboxed execution.
     OnFailure,
 
+    /// The model decides when to ask the user for approval.
+    OnRequest,
+
     /// Never ask for user approval
     /// Execution failures are immediately returned to the model.
     Never,
@@ -28,6 +31,7 @@ impl From<ApprovalModeCliArg> for AskForApproval {
         match value {
             ApprovalModeCliArg::Untrusted => AskForApproval::UnlessTrusted,
             ApprovalModeCliArg::OnFailure => AskForApproval::OnFailure,
+            ApprovalModeCliArg::OnRequest => AskForApproval::OnRequest,
             ApprovalModeCliArg::Never => AskForApproval::Never,
         }
     }

codex-rs/common/src/config_summary.rs

@@ -0,0 +1,29 @@
+use codex_core::WireApi;
+use codex_core::config::Config;
+
+use crate::sandbox_summary::summarize_sandbox_policy;
+
+/// Build a list of key/value pairs summarizing the effective configuration.
+pub fn create_config_summary_entries(config: &Config) -> Vec<(&'static str, String)> {
+    let mut entries = vec![
+        ("workdir", config.cwd.display().to_string()),
+        ("model", config.model.clone()),
+        ("provider", config.model_provider_id.clone()),
+        ("approval", config.approval_policy.to_string()),
+        ("sandbox", summarize_sandbox_policy(&config.sandbox_policy)),
+    ];
+    if config.model_provider.wire_api == WireApi::Responses
+        && config.model_family.supports_reasoning_summaries
+    {
+        entries.push((
+            "reasoning effort",
+            config.model_reasoning_effort.to_string(),
+        ));
+        entries.push((
+            "reasoning summaries",
+            config.model_reasoning_summary.to_string(),
+        ));
+    }
+
+    entries
+}

codex-rs/common/src/lib.rs

@@ -23,3 +23,7 @@ mod sandbox_summary;
 
 #[cfg(feature = "sandbox_summary")]
 pub use sandbox_summary::summarize_sandbox_policy;
+
+mod config_summary;
+
+pub use config_summary::create_config_summary_entries;

codex-rs/config.md

@@ -148,12 +148,20 @@ Determines when the user should be prompted to approve whether Codex can execute
 approval_policy = "untrusted"
 ```
 
+If you want to be notified whenever a command fails, use "on-failure":
 ```toml
 # If the command fails when run in the sandbox, Codex asks for permission to
 # retry the command outside the sandbox.
 approval_policy = "on-failure"
 ```
 
+If you want the model to run until it decides that it needs to ask you for escalated permissions, use "on-request":
+```toml
+# The model decides when to escalate
+approval_policy = "on-request"
+```
+
+Alternatively, you can have the model run until it is done, and never ask to run a command with escalated permissions:
 ```toml
 # User is never prompted: if the command fails, Codex will automatically try
 # something out. Note the `exec` subcommand always uses this mode.
@@ -483,6 +491,19 @@ Setting `hide_agent_reasoning` to `true` suppresses these events in **both** the
 hide_agent_reasoning = true   # defaults to false
 ```
 
+## show_raw_agent_reasoning
+
+Surfaces the model’s raw chain-of-thought ("raw reasoning content") when available.
+
+Notes:
+- Only takes effect if the selected model/provider actually emits raw reasoning content. Many models do not. When unsupported, this option has no visible effect.
+- Raw reasoning may include intermediate thoughts or sensitive context. Enable only if acceptable for your workflow.
+
+Example:
+```toml
+show_raw_agent_reasoning = true  # defaults to false
+```
+
 ## model_context_window
 
 The size of the context window for the model, in tokens.

codex-rs/core/Cargo.toml

@@ -46,7 +46,7 @@ tokio = { version = "1", features = [
     "signal",
 ] }
 tokio-util = "0.7.14"
-toml = "0.9.2"
+toml = "0.9.4"
 tracing = { version = "0.1.41", features = ["log"] }
 tree-sitter = "0.25.8"
 tree-sitter-bash = "0.25.0"

codex-rs/core/prompt.md

@@ -1,16 +1,31 @@
-Please resolve the user's task by editing and testing the code files in your current code execution session.
-You are a deployed coding agent.
-Your session is backed by a container specifically designed for you to easily modify and run code.
-The repo(s) are already cloned in your working directory, and you must fully solve the problem for your answer to be considered correct.
+You are operating as and within the Codex CLI, an open-source, terminal-based agentic coding assistant built by OpenAI. It wraps OpenAI models to enable natural language interaction with a local codebase. You are expected to be precise, safe, and helpful.
 
+Your capabilities:
+- Receive user prompts, project context, and files.
+- Stream responses and emit function calls (e.g., shell commands, code edits).
+- Run commands, like apply_patch, and manage user approvals based on policy.
+- Work inside a workspace with sandboxing instructions specified by the policy described in (## Sandbox environment and approval instructions)
+
+Within this context, Codex refers to the open-source agentic coding interface (not the old Codex language model built by OpenAI).
+
+## General guidelines
+As a deployed coding agent, please continue working on the user's task until their query is resolved, before ending your turn and yielding back to the user. Only terminate your turn when you are sure that the task is solved. If you are not sure about file content or codebase structure pertaining to the user's request, use your tools to read files and gather the relevant information. Do NOT guess or make up an answer.
+
+After a user sends their first message, you should immediately provide a brief message acknowledging their request to set the tone and expectation of future work to be done (no more than 8-10 words). This should be done before performing work like exploring the codebase, writing or reading files, or other tool calls needed to complete the task. Use a natural, collaborative tone similar to how a teammate would receive a task during a pair programming session.
+
+Please resolve the user's task by editing the code files in your current code execution session. Your session allows for you to modify and run code. The repo(s) are already cloned in your working directory, and you must fully solve the problem for your answer to be considered correct.
+
+### Task execution
 You MUST adhere to the following criteria when executing the task:
 
 - Working on the repo(s) in the current environment is allowed, even if they are proprietary.
 - Analyzing code for vulnerabilities is allowed.
 - Showing user code and tool call details is allowed.
 - User instructions may overwrite the _CODING GUIDELINES_ section in this developer message.
+- `user_instructions` are not part of the user's request, but guidance for how to complete the task.
+- Do not cite `user_instructions` back to the user unless a specific piece is relevant.
 - Do not use \`ls -R\`, \`find\`, or \`grep\` - these are slow in large repos. Use \`rg\` and \`rg --files\`.
-- Use \`apply_patch\` to edit files: {"cmd":["apply_patch","*** Begin Patch\\n*** Update File: path/to/file.py\\n@@ def example():\\n- pass\\n+ return 123\\n*** End Patch"]}
+- Use the \`apply_patch\` shell command to edit files: {"command":["apply_patch","*** Begin Patch\\n*** Update File: path/to/file.py\\n@@ def example():\\n- pass\\n+ return 123\\n*** End Patch"]}
 - If completing the user's task requires writing or modifying files:
   - Your code and final answer should follow these _CODING GUIDELINES_:
     - Fix the problem at the root cause rather than applying surface-level patches, when possible.
@@ -33,23 +48,22 @@ You MUST adhere to the following criteria when executing the task:
 - If completing the user's task DOES NOT require writing or modifying files (e.g., the user asks a question about the code base):
   - Respond in a friendly tune as a remote teammate, who is knowledgeable, capable and eager to help with coding.
 - When your task involves writing or modifying files:
-  - Do NOT tell the user to "save the file" or "copy the code into a file" if you already created or modified the file using \`apply_patch\`. Instead, reference the file as already saved.
+  - Do NOT tell the user to "save the file" or "copy the code into a file" if you already created or modified the file using the `apply_patch` shell command. Instead, reference the file as already saved.
   - Do NOT show the full contents of large files you have already written, unless the user explicitly asks for them.
 
-§ `apply-patch` Specification
+## Using the shell command `apply_patch` to edit files
+`apply_patch` is a shell command for editing files. Your patch language is a stripped‑down, file‑oriented diff format designed to be easy to parse and safe to apply. You can think of it as a high‑level envelope:
 
-Your patch language is a stripped‑down, file‑oriented diff format designed to be easy to parse and safe to apply. You can think of it as a high‑level envelope:
-
-**_ Begin Patch
+*** Begin Patch
 [ one or more file sections ]
-_** End Patch
+*** End Patch
 
 Within that envelope, you get a sequence of file operations.
 You MUST include a header to specify the action you are taking.
 Each operation starts with one of three headers:
 
-**_ Add File: <path> - create a new file. Every following line is a + line (the initial contents).
-_** Delete File: <path> - remove an existing file. Nothing follows.
+*** Add File: <path> - create a new file. Every following line is a + line (the initial contents).
+*** Delete File: <path> - remove an existing file. Nothing follows.
 \*\*\* Update File: <path> - patch an existing file in place (optionally with a rename).
 
 May be immediately followed by \*\*\* Move to: <new path> if you want to rename the file.
@@ -63,45 +77,60 @@ Within a hunk each line starts with:
   At the end of a truncated hunk you can emit \*\*\* End of File.
 
 Patch := Begin { FileOp } End
-Begin := "**_ Begin Patch" NEWLINE
-End := "_** End Patch" NEWLINE
+Begin := "*** Begin Patch" NEWLINE
+End := "*** End Patch" NEWLINE
 FileOp := AddFile | DeleteFile | UpdateFile
-AddFile := "**_ Add File: " path NEWLINE { "+" line NEWLINE }
-DeleteFile := "_** Delete File: " path NEWLINE
-UpdateFile := "**_ Update File: " path NEWLINE [ MoveTo ] { Hunk }
-MoveTo := "_** Move to: " newPath NEWLINE
+AddFile := "*** Add File: " path NEWLINE { "+" line NEWLINE }
+DeleteFile := "*** Delete File: " path NEWLINE
+UpdateFile := "*** Update File: " path NEWLINE [ MoveTo ] { Hunk }
+MoveTo := "*** Move to: " newPath NEWLINE
 Hunk := "@@" [ header ] NEWLINE { HunkLine } [ "*** End of File" NEWLINE ]
 HunkLine := (" " | "-" | "+") text NEWLINE
 
 A full patch can combine several operations:
 
-**_ Begin Patch
-_** Add File: hello.txt
+*** Begin Patch
+*** Add File: hello.txt
 +Hello world
-**_ Update File: src/app.py
-_** Move to: src/main.py
+*** Update File: src/app.py
+*** Move to: src/main.py
 @@ def greet():
 -print("Hi")
 +print("Hello, world!")
-**_ Delete File: obsolete.txt
-_** End Patch
+*** Delete File: obsolete.txt
+*** End Patch
 
 It is important to remember:
 
 - You must include a header with your intended action (Add/Delete/Update)
 - You must prefix new lines with `+` even when creating a new file
+- You must follow this schema exactly when providing a patch
 
-You can invoke apply_patch like:
+You can invoke apply_patch with the following shell command:
 
 ```
 shell {"command":["apply_patch","*** Begin Patch\n*** Add File: hello.txt\n+Hello, world!\n*** End Patch\n"]}
 ```
 
-Plan updates
+## Sandbox environment and approval instructions
+
+You are running in a sandboxed workspace backed by version control. The sandbox might be configured by the user to restrict certain behaviors, like accessing the internet or writing to files outside the current directory.
+
+Commands that are blocked by sandbox settings will be automatically sent to the user for approval. The result of the request will be returned (i.e. the command result, or the request denial).
+The user also has an opportunity to approve the same command for the rest of the session.
+
+Guidance on running within the sandbox:
+- When running commands that will likely require approval, attempt to use simple, precise commands, to reduce frequency of approval requests.
+- When approval is denied or a command fails due to a permission error, do not retry the exact command in a different way. Move on and continue trying to address the user's request.
+
+
+## Tools available
+### Plan updates
 
 A tool named `update_plan` is available. Use it to keep an up‑to‑date, step‑by‑step plan for the task so you can follow your progress. When making your plans, keep in mind that you are a deployed coding agent - `update_plan` calls should not involve doing anything that you aren't capable of doing. For example, `update_plan` calls should NEVER contain tasks to merge your own pull requests. Only stop to ask the user if you genuinely need their feedback on a change.
 
-- At the start of the task, call `update_plan` with an initial plan: a short list of 1‑sentence steps with a `status` for each step (`pending`, `in_progress`, or `completed`). There should always be exactly one `in_progress` step until everything is done.
+- At the start of any nontrivial task, call `update_plan` with an initial plan: a short list of 1‑sentence steps with a `status` for each step (`pending`, `in_progress`, or `completed`). There should always be exactly one `in_progress` step until everything is done.
 - Whenever you finish a step, call `update_plan` again, marking the finished step as `completed` and the next step as `in_progress`.
 - If your plan needs to change, call `update_plan` with the revised steps and include an `explanation` describing the change.
 - When all steps are complete, make a final `update_plan` call with all steps marked `completed`.
+

codex-rs/core/src/chat_completions.rs

@@ -21,7 +21,9 @@ use crate::client_common::ResponseEvent;
 use crate::client_common::ResponseStream;
 use crate::error::CodexErr;
 use crate::error::Result;
+use crate::model_family::ModelFamily;
 use crate::models::ContentItem;
+use crate::models::ReasoningItemContent;
 use crate::models::ResponseItem;
 use crate::openai_tools::create_tools_json_for_chat_completions_api;
 use crate::util::backoff;
@@ -29,22 +31,19 @@ use crate::util::backoff;
 /// Implementation for the classic Chat Completions API.
 pub(crate) async fn stream_chat_completions(
     prompt: &Prompt,
-    model: &str,
-    include_plan_tool: bool,
+    model_family: &ModelFamily,
     client: &reqwest::Client,
     provider: &ModelProviderInfo,
 ) -> Result<ResponseStream> {
     // Build messages array
     let mut messages = Vec::<serde_json::Value>::new();
 
-    let full_instructions = prompt.get_full_instructions(model);
+    let full_instructions = prompt.get_full_instructions(model_family);
     messages.push(json!({"role": "system", "content": full_instructions}));
 
-    if let Some(instr) = &prompt.user_instructions {
-        messages.push(json!({"role": "user", "content": instr}));
-    }
+    let input = prompt.get_formatted_input();
 
-    for item in &prompt.input {
+    for item in &input {
         match item {
             ResponseItem::Message { role, content, .. } => {
                 let mut text = String::new();
@@ -110,9 +109,9 @@ pub(crate) async fn stream_chat_completions(
         }
     }
 
-    let tools_json = create_tools_json_for_chat_completions_api(prompt, model, include_plan_tool)?;
+    let tools_json = create_tools_json_for_chat_completions_api(&prompt.tools)?;
     let payload = json!({
-        "model": model,
+        "model": model_family.slug,
         "messages": messages,
         "stream": true,
         "tools": tools_json,
@@ -120,7 +119,7 @@ pub(crate) async fn stream_chat_completions(
 
     debug!(
         "POST to {}: {}",
-        provider.get_full_url(),
+        provider.get_full_url(&None),
         serde_json::to_string_pretty(&payload).unwrap_or_default()
     );
 
@@ -129,7 +128,7 @@ pub(crate) async fn stream_chat_completions(
     loop {
         attempt += 1;
 
-        let req_builder = provider.create_request_builder(client)?;
+        let req_builder = provider.create_request_builder(client, &None).await?;
 
         let res = req_builder
             .header(reqwest::header::ACCEPT, "text/event-stream")
@@ -207,6 +206,8 @@ async fn process_chat_sse<S>(
     }
 
     let mut fn_call_state = FunctionCallState::default();
+    let mut assistant_text = String::new();
+    let mut reasoning_text = String::new();
 
     loop {
         let sse = match timeout(idle_timeout, stream.next()).await {
@@ -235,6 +236,31 @@ async fn process_chat_sse<S>(
 
         // OpenAI Chat streaming sends a literal string "[DONE]" when finished.
         if sse.data.trim() == "[DONE]" {
+            // Emit any finalized items before closing so downstream consumers receive
+            // terminal events for both assistant content and raw reasoning.
+            if !assistant_text.is_empty() {
+                let item = ResponseItem::Message {
+                    role: "assistant".to_string(),
+                    content: vec![ContentItem::OutputText {
+                        text: std::mem::take(&mut assistant_text),
+                    }],
+                    id: None,
+                };
+                let _ = tx_event.send(Ok(ResponseEvent::OutputItemDone(item))).await;
+            }
+
+            if !reasoning_text.is_empty() {
+                let item = ResponseItem::Reasoning {
+                    id: String::new(),
+                    summary: Vec::new(),
+                    content: Some(vec![ReasoningItemContent::ReasoningText {
+                        text: std::mem::take(&mut reasoning_text),
+                    }]),
+                    encrypted_content: None,
+                };
+                let _ = tx_event.send(Ok(ResponseEvent::OutputItemDone(item))).await;
+            }
+
             let _ = tx_event
                 .send(Ok(ResponseEvent::Completed {
                     response_id: String::new(),
@@ -254,21 +280,47 @@ async fn process_chat_sse<S>(
         let choice_opt = chunk.get("choices").and_then(|c| c.get(0));
 
         if let Some(choice) = choice_opt {
-            // Handle assistant content tokens.
+            // Handle assistant content tokens as streaming deltas.
             if let Some(content) = choice
                 .get("delta")
                 .and_then(|d| d.get("content"))
                 .and_then(|c| c.as_str())
             {
-                let item = ResponseItem::Message {
-                    role: "assistant".to_string(),
-                    content: vec![ContentItem::OutputText {
-                        text: content.to_string(),
-                    }],
-                    id: None,
-                };
+                if !content.is_empty() {
+                    assistant_text.push_str(content);
+                    let _ = tx_event
+                        .send(Ok(ResponseEvent::OutputTextDelta(content.to_string())))
+                        .await;
+                }
+            }
 
-                let _ = tx_event.send(Ok(ResponseEvent::OutputItemDone(item))).await;
+            // Forward any reasoning/thinking deltas if present.
+            // Some providers stream `reasoning` as a plain string while others
+            // nest the text under an object (e.g. `{ "reasoning": { "text": "…" } }`).
+            if let Some(reasoning_val) = choice.get("delta").and_then(|d| d.get("reasoning")) {
+                let mut maybe_text = reasoning_val.as_str().map(|s| s.to_string());
+
+                if maybe_text.is_none() && reasoning_val.is_object() {
+                    if let Some(s) = reasoning_val
+                        .get("text")
+                        .and_then(|t| t.as_str())
+                        .filter(|s| !s.is_empty())
+                    {
+                        maybe_text = Some(s.to_string());
+                    } else if let Some(s) = reasoning_val
+                        .get("content")
+                        .and_then(|t| t.as_str())
+                        .filter(|s| !s.is_empty())
+                    {
+                        maybe_text = Some(s.to_string());
+                    }
+                }
+
+                if let Some(reasoning) = maybe_text {
+                    let _ = tx_event
+                        .send(Ok(ResponseEvent::ReasoningContentDelta(reasoning)))
+                        .await;
+                }
             }
 
             // Handle streaming function / tool calls.
@@ -305,7 +357,21 @@ async fn process_chat_sse<S>(
             if let Some(finish_reason) = choice.get("finish_reason").and_then(|v| v.as_str()) {
                 match finish_reason {
                     "tool_calls" if fn_call_state.active => {
-                        // Build the FunctionCall response item.
+                        // First, flush the terminal raw reasoning so UIs can finalize
+                        // the reasoning stream before any exec/tool events begin.
+                        if !reasoning_text.is_empty() {
+                            let item = ResponseItem::Reasoning {
+                                id: String::new(),
+                                summary: Vec::new(),
+                                content: Some(vec![ReasoningItemContent::ReasoningText {
+                                    text: std::mem::take(&mut reasoning_text),
+                                }]),
+                                encrypted_content: None,
+                            };
+                            let _ = tx_event.send(Ok(ResponseEvent::OutputItemDone(item))).await;
+                        }
+
+                        // Then emit the FunctionCall response item.
                         let item = ResponseItem::FunctionCall {
                             id: None,
                             name: fn_call_state.name.clone().unwrap_or_else(|| "".to_string()),
@@ -313,11 +379,33 @@ async fn process_chat_sse<S>(
                             call_id: fn_call_state.call_id.clone().unwrap_or_else(String::new),
                         };
 
-                        // Emit it downstream.
                         let _ = tx_event.send(Ok(ResponseEvent::OutputItemDone(item))).await;
                     }
                     "stop" => {
-                        // Regular turn without tool-call.
+                        // Regular turn without tool-call. Emit the final assistant message
+                        // as a single OutputItemDone so non-delta consumers see the result.
+                        if !assistant_text.is_empty() {
+                            let item = ResponseItem::Message {
+                                role: "assistant".to_string(),
+                                content: vec![ContentItem::OutputText {
+                                    text: std::mem::take(&mut assistant_text),
+                                }],
+                                id: None,
+                            };
+                            let _ = tx_event.send(Ok(ResponseEvent::OutputItemDone(item))).await;
+                        }
+                        // Also emit a terminal Reasoning item so UIs can finalize raw reasoning.
+                        if !reasoning_text.is_empty() {
+                            let item = ResponseItem::Reasoning {
+                                id: String::new(),
+                                summary: Vec::new(),
+                                content: Some(vec![ReasoningItemContent::ReasoningText {
+                                    text: std::mem::take(&mut reasoning_text),
+                                }]),
+                                encrypted_content: None,
+                            };
+                            let _ = tx_event.send(Ok(ResponseEvent::OutputItemDone(item))).await;
+                        }
                     }
                     _ => {}
                 }
@@ -355,10 +443,17 @@ async fn process_chat_sse<S>(
 /// The adapter is intentionally *lossless*: callers who do **not** opt in via
 /// [`AggregateStreamExt::aggregate()`] keep receiving the original unmodified
 /// events.
+#[derive(Copy, Clone, Eq, PartialEq)]
+enum AggregateMode {
+    AggregatedOnly,
+    Streaming,
+}
 pub(crate) struct AggregatedChatStream<S> {
     inner: S,
     cumulative: String,
-    pending_completed: Option<ResponseEvent>,
+    cumulative_reasoning: String,
+    pending: std::collections::VecDeque<ResponseEvent>,
+    mode: AggregateMode,
 }
 
 impl<S> Stream for AggregatedChatStream<S>
@@ -370,8 +465,8 @@ where
     fn poll_next(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Option<Self::Item>> {
         let this = self.get_mut();
 
-        // First, flush any buffered Completed event from the previous call.
-        if let Some(ev) = this.pending_completed.take() {
+        // First, flush any buffered events from the previous call.
+        if let Some(ev) = this.pending.pop_front() {
             return Poll::Ready(Some(Ok(ev)));
         }
 
@@ -388,16 +483,21 @@ where
                     let is_assistant_delta = matches!(&item, crate::models::ResponseItem::Message { role, .. } if role == "assistant");
 
                     if is_assistant_delta {
-                        if let crate::models::ResponseItem::Message { content, .. } = &item {
-                            if let Some(text) = content.iter().find_map(|c| match c {
-                                crate::models::ContentItem::OutputText { text } => Some(text),
-                                _ => None,
-                            }) {
-                                this.cumulative.push_str(text);
+                        // Only use the final assistant message if we have not
+                        // seen any deltas; otherwise, deltas already built the
+                        // cumulative text and this would duplicate it.
+                        if this.cumulative.is_empty() {
+                            if let crate::models::ResponseItem::Message { content, .. } = &item {
+                                if let Some(text) = content.iter().find_map(|c| match c {
+                                    crate::models::ContentItem::OutputText { text } => Some(text),
+                                    _ => None,
+                                }) {
+                                    this.cumulative.push_str(text);
+                                }
                             }
                         }
 
-                        // Swallow partial assistant chunk; keep polling.
+                        // Swallow assistant message here; emit on Completed.
                         continue;
                     }
 
@@ -408,24 +508,50 @@ where
                     response_id,
                     token_usage,
                 }))) => {
+                    // Build any aggregated items in the correct order: Reasoning first, then Message.
+                    let mut emitted_any = false;
+
+                    if !this.cumulative_reasoning.is_empty()
+                        && matches!(this.mode, AggregateMode::AggregatedOnly)
+                    {
+                        let aggregated_reasoning = crate::models::ResponseItem::Reasoning {
+                            id: String::new(),
+                            summary: Vec::new(),
+                            content: Some(vec![
+                                crate::models::ReasoningItemContent::ReasoningText {
+                                    text: std::mem::take(&mut this.cumulative_reasoning),
+                                },
+                            ]),
+                            encrypted_content: None,
+                        };
+                        this.pending
+                            .push_back(ResponseEvent::OutputItemDone(aggregated_reasoning));
+                        emitted_any = true;
+                    }
+
                     if !this.cumulative.is_empty() {
-                        let aggregated_item = crate::models::ResponseItem::Message {
+                        let aggregated_message = crate::models::ResponseItem::Message {
                             id: None,
                             role: "assistant".to_string(),
                             content: vec![crate::models::ContentItem::OutputText {
                                 text: std::mem::take(&mut this.cumulative),
                             }],
                         };
+                        this.pending
+                            .push_back(ResponseEvent::OutputItemDone(aggregated_message));
+                        emitted_any = true;
+                    }
 
-                        // Buffer Completed so it is returned *after* the aggregated message.
-                        this.pending_completed = Some(ResponseEvent::Completed {
-                            response_id,
-                            token_usage,
+                    // Always emit Completed last when anything was aggregated.
+                    if emitted_any {
+                        this.pending.push_back(ResponseEvent::Completed {
+                            response_id: response_id.clone(),
+                            token_usage: token_usage.clone(),
                         });
-
-                        return Poll::Ready(Some(Ok(ResponseEvent::OutputItemDone(
-                            aggregated_item,
-                        ))));
+                        // Return the first pending event now.
+                        if let Some(ev) = this.pending.pop_front() {
+                            return Poll::Ready(Some(Ok(ev)));
+                        }
                     }
 
                     // Nothing aggregated – forward Completed directly.
@@ -439,10 +565,27 @@ where
                     // will never appear in a Chat Completions stream.
                     continue;
                 }
-                Poll::Ready(Some(Ok(ResponseEvent::OutputTextDelta(_))))
-                | Poll::Ready(Some(Ok(ResponseEvent::ReasoningSummaryDelta(_)))) => {
-                    // Deltas are ignored here since aggregation waits for the
-                    // final OutputItemDone.
+                Poll::Ready(Some(Ok(ResponseEvent::OutputTextDelta(delta)))) => {
+                    // Always accumulate deltas so we can emit a final OutputItemDone at Completed.
+                    this.cumulative.push_str(&delta);
+                    if matches!(this.mode, AggregateMode::Streaming) {
+                        // In streaming mode, also forward the delta immediately.
+                        return Poll::Ready(Some(Ok(ResponseEvent::OutputTextDelta(delta))));
+                    } else {
+                        continue;
+                    }
+                }
+                Poll::Ready(Some(Ok(ResponseEvent::ReasoningContentDelta(delta)))) => {
+                    // Always accumulate reasoning deltas so we can emit a final Reasoning item at Completed.
+                    this.cumulative_reasoning.push_str(&delta);
+                    if matches!(this.mode, AggregateMode::Streaming) {
+                        // In streaming mode, also forward the delta immediately.
+                        return Poll::Ready(Some(Ok(ResponseEvent::ReasoningContentDelta(delta))));
+                    } else {
+                        continue;
+                    }
+                }
+                Poll::Ready(Some(Ok(ResponseEvent::ReasoningSummaryDelta(_)))) => {
                     continue;
                 }
             }
@@ -472,12 +615,24 @@ pub(crate) trait AggregateStreamExt: Stream<Item = Result<ResponseEvent>> + Size
     /// }
     /// ```
     fn aggregate(self) -> AggregatedChatStream<Self> {
-        AggregatedChatStream {
-            inner: self,
-            cumulative: String::new(),
-            pending_completed: None,
-        }
+        AggregatedChatStream::new(self, AggregateMode::AggregatedOnly)
     }
 }
 
 impl<T> AggregateStreamExt for T where T: Stream<Item = Result<ResponseEvent>> + Sized {}
+
+impl<S> AggregatedChatStream<S> {
+    fn new(inner: S, mode: AggregateMode) -> Self {
+        AggregatedChatStream {
+            inner,
+            cumulative: String::new(),
+            cumulative_reasoning: String::new(),
+            pending: std::collections::VecDeque::new(),
+            mode,
+        }
+    }
+
+    pub(crate) fn streaming_mode(inner: S) -> Self {
+        Self::new(inner, AggregateMode::Streaming)
+    }
+}

codex-rs/core/src/client.rs

@@ -30,12 +30,10 @@ use crate::config::Config;
 use crate::config_types::ReasoningEffort as ReasoningEffortConfig;
 use crate::config_types::ReasoningSummary as ReasoningSummaryConfig;
 use crate::error::CodexErr;
-use crate::error::EnvVarError;
 use crate::error::Result;
 use crate::flags::CODEX_RS_SSE_FIXTURE;
 use crate::model_provider_info::ModelProviderInfo;
 use crate::model_provider_info::WireApi;
-use crate::models::ContentItem;
 use crate::models::ResponseItem;
 use crate::openai_tools::create_tools_json_for_responses_api;
 use crate::protocol::TokenUsage;
@@ -83,8 +81,7 @@ impl ModelClient {
                 // Create the raw streaming connection first.
                 let response_stream = stream_chat_completions(
                     prompt,
-                    &self.config.model,
-                    self.config.include_plan_tool,
+                    &self.config.model_family,
                     &self.client,
                     &self.provider,
                 )
@@ -93,7 +90,11 @@ impl ModelClient {
                 // Wrap it with the aggregation adapter so callers see *only*
                 // the final assistant message per turn (matching the
                 // behaviour of the Responses API).
-                let mut aggregated = response_stream.aggregate();
+                let mut aggregated = if self.config.show_raw_agent_reasoning {
+                    crate::chat_completions::AggregatedChatStream::streaming_mode(response_stream)
+                } else {
+                    response_stream.aggregate()
+                };
 
                 // Bridge the aggregated stream back into a standard
                 // `ResponseStream` by forwarding events through a channel.
@@ -122,32 +123,19 @@ impl ModelClient {
             return stream_from_fixture(path, self.provider.clone()).await;
         }
 
-        let auth = self.auth.as_ref().ok_or_else(|| {
-            CodexErr::EnvVar(EnvVarError {
-                var: "OPENAI_API_KEY".to_string(),
-                instructions: Some("Create an API key (https://platform.openai.com) and export it as an environment variable.".to_string()),
-            })
-        })?;
+        let auth = self.auth.clone();
 
-        let store = prompt.store && auth.mode != AuthMode::ChatGPT;
+        let auth_mode = auth.as_ref().map(|a| a.mode);
 
-        let base_url = match self.provider.base_url.clone() {
-            Some(url) => url,
-            None => match auth.mode {
-                AuthMode::ChatGPT => "https://chatgpt.com/backend-api/codex".to_string(),
-                AuthMode::ApiKey => "https://api.openai.com/v1".to_string(),
-            },
-        };
+        let store = prompt.store && auth_mode != Some(AuthMode::ChatGPT);
 
-        let token = auth.get_token().await?;
-
-        let full_instructions = prompt.get_full_instructions(&self.config.model);
-        let tools_json = create_tools_json_for_responses_api(
-            prompt,
-            &self.config.model,
-            self.config.include_plan_tool,
-        )?;
-        let reasoning = create_reasoning_param_for_request(&self.config, self.effort, self.summary);
+        let full_instructions = prompt.get_full_instructions(&self.config.model_family);
+        let tools_json = create_tools_json_for_responses_api(&prompt.tools)?;
+        let reasoning = create_reasoning_param_for_request(
+            &self.config.model_family,
+            self.effort,
+            self.summary,
+        );
 
         // Request encrypted COT if we are not storing responses,
         // otherwise reasoning items will be referenced by ID
@@ -157,15 +145,7 @@ impl ModelClient {
             vec![]
         };
 
-        let mut input_with_instructions = Vec::with_capacity(prompt.input.len() + 1);
-        if let Some(ui) = &prompt.user_instructions {
-            input_with_instructions.push(ResponseItem::Message {
-                id: None,
-                role: "user".to_string(),
-                content: vec![ContentItem::InputText { text: ui.clone() }],
-            });
-        }
-        input_with_instructions.extend(prompt.input.clone());
+        let input_with_instructions = prompt.get_formatted_input();
 
         let payload = ResponsesApiRequest {
             model: &self.config.model,
@@ -180,35 +160,36 @@ impl ModelClient {
             include,
         };
 
-        trace!(
-            "POST to {}: {}",
-            self.provider.get_full_url(),
-            serde_json::to_string(&payload)?
-        );
-
         let mut attempt = 0;
         let max_retries = self.provider.request_max_retries();
 
+        trace!(
+            "POST to {}: {}",
+            self.provider.get_full_url(&auth),
+            serde_json::to_string(&payload)?
+        );
+
         loop {
             attempt += 1;
 
             let mut req_builder = self
-                .client
-                .post(format!("{base_url}/responses"))
+                .provider
+                .create_request_builder(&self.client, &auth)
+                .await?;
+
+            req_builder = req_builder
                 .header("OpenAI-Beta", "responses=experimental")
                 .header("session_id", self.session_id.to_string())
-                .bearer_auth(&token)
                 .header(reqwest::header::ACCEPT, "text/event-stream")
                 .json(&payload);
 
-            if auth.mode == AuthMode::ChatGPT {
-                if let Some(account_id) = auth.get_account_id().await {
-                    req_builder = req_builder.header("chatgpt-account-id", account_id);
-                }
+            if let Some(auth) = auth.as_ref()
+                && auth.mode == AuthMode::ChatGPT
+                && let Some(account_id) = auth.get_account_id().await
+            {
+                req_builder = req_builder.header("chatgpt-account-id", account_id);
             }
 
-            req_builder = self.provider.apply_http_headers(req_builder);
-
             let originator = self
                 .config
                 .internal_originator
@@ -446,6 +427,14 @@ async fn process_sse<S>(
                     }
                 }
             }
+            "response.reasoning_text.delta" => {
+                if let Some(delta) = event.delta {
+                    let event = ResponseEvent::ReasoningContentDelta(delta);
+                    if tx_event.send(Ok(event)).await.is_err() {
+                        return;
+                    }
+                }
+            }
             "response.created" => {
                 if event.response.is_some() {
                     let _ = tx_event.send(Ok(ResponseEvent::Created {})).await;

codex-rs/core/src/client_common.rs

@@ -1,13 +1,19 @@
 use crate::config_types::ReasoningEffort as ReasoningEffortConfig;
 use crate::config_types::ReasoningSummary as ReasoningSummaryConfig;
 use crate::error::Result;
+use crate::model_family::ModelFamily;
+use crate::models::ContentItem;
 use crate::models::ResponseItem;
+use crate::openai_tools::OpenAiTool;
+use crate::protocol::AskForApproval;
+use crate::protocol::SandboxPolicy;
 use crate::protocol::TokenUsage;
 use codex_apply_patch::APPLY_PATCH_TOOL_INSTRUCTIONS;
 use futures::Stream;
 use serde::Serialize;
 use std::borrow::Cow;
-use std::collections::HashMap;
+use std::fmt::Display;
+use std::path::PathBuf;
 use std::pin::Pin;
 use std::task::Context;
 use std::task::Poll;
@@ -17,6 +23,47 @@ use tokio::sync::mpsc;
 /// with this content.
 const BASE_INSTRUCTIONS: &str = include_str!("../prompt.md");
 
+/// wraps environment context message in a tag for the model to parse more easily.
+const ENVIRONMENT_CONTEXT_START: &str = "<environment_context>\n\n";
+const ENVIRONMENT_CONTEXT_END: &str = "\n\n</environment_context>";
+
+/// wraps user instructions message in a tag for the model to parse more easily.
+const USER_INSTRUCTIONS_START: &str = "<user_instructions>\n\n";
+const USER_INSTRUCTIONS_END: &str = "\n\n</user_instructions>";
+
+#[derive(Debug, Clone)]
+pub(crate) struct EnvironmentContext {
+    pub cwd: PathBuf,
+    pub approval_policy: AskForApproval,
+    pub sandbox_policy: SandboxPolicy,
+}
+
+impl Display for EnvironmentContext {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        writeln!(
+            f,
+            "Current working directory: {}",
+            self.cwd.to_string_lossy()
+        )?;
+        writeln!(f, "Approval policy: {}", self.approval_policy)?;
+        writeln!(f, "Sandbox policy: {}", self.sandbox_policy)?;
+
+        let network_access = match self.sandbox_policy.clone() {
+            SandboxPolicy::DangerFullAccess => "enabled",
+            SandboxPolicy::ReadOnly => "restricted",
+            SandboxPolicy::WorkspaceWrite { network_access, .. } => {
+                if network_access {
+                    "enabled"
+                } else {
+                    "restricted"
+                }
+            }
+        };
+        writeln!(f, "Network access: {network_access}")?;
+        Ok(())
+    }
+}
+
 /// API request payload for a single model turn.
 #[derive(Default, Debug, Clone)]
 pub struct Prompt {
@@ -28,27 +75,62 @@ pub struct Prompt {
     /// Whether to store response on server side (disable_response_storage = !store).
     pub store: bool,
 
-    /// Additional tools sourced from external MCP servers. Note each key is
-    /// the "fully qualified" tool name (i.e., prefixed with the server name),
-    /// which should be reported to the model in place of Tool::name.
-    pub extra_tools: HashMap<String, mcp_types::Tool>,
+    /// A list of key-value pairs that will be added as a developer message
+    /// for the model to use
+    pub environment_context: Option<EnvironmentContext>,
+
+    /// Tools available to the model, including additional tools sourced from
+    /// external MCP servers.
+    pub tools: Vec<OpenAiTool>,
 
     /// Optional override for the built-in BASE_INSTRUCTIONS.
     pub base_instructions_override: Option<String>,
 }
 
 impl Prompt {
-    pub(crate) fn get_full_instructions(&self, model: &str) -> Cow<'_, str> {
+    pub(crate) fn get_full_instructions(&self, model: &ModelFamily) -> Cow<'_, str> {
         let base = self
             .base_instructions_override
             .as_deref()
             .unwrap_or(BASE_INSTRUCTIONS);
         let mut sections: Vec<&str> = vec![base];
-        if model.starts_with("gpt-4.1") {
+        if model.needs_special_apply_patch_instructions {
             sections.push(APPLY_PATCH_TOOL_INSTRUCTIONS);
         }
         Cow::Owned(sections.join("\n"))
     }
+
+    fn get_formatted_user_instructions(&self) -> Option<String> {
+        self.user_instructions
+            .as_ref()
+            .map(|ui| format!("{USER_INSTRUCTIONS_START}{ui}{USER_INSTRUCTIONS_END}"))
+    }
+
+    fn get_formatted_environment_context(&self) -> Option<String> {
+        self.environment_context
+            .as_ref()
+            .map(|ec| format!("{ENVIRONMENT_CONTEXT_START}{ec}{ENVIRONMENT_CONTEXT_END}"))
+    }
+
+    pub(crate) fn get_formatted_input(&self) -> Vec<ResponseItem> {
+        let mut input_with_instructions = Vec::with_capacity(self.input.len() + 2);
+        if let Some(ec) = self.get_formatted_environment_context() {
+            input_with_instructions.push(ResponseItem::Message {
+                id: None,
+                role: "user".to_string(),
+                content: vec![ContentItem::InputText { text: ec }],
+            });
+        }
+        if let Some(ui) = self.get_formatted_user_instructions() {
+            input_with_instructions.push(ResponseItem::Message {
+                id: None,
+                role: "user".to_string(),
+                content: vec![ContentItem::InputText { text: ui }],
+            });
+        }
+        input_with_instructions.extend(self.input.clone());
+        input_with_instructions
+    }
 }
 
 #[derive(Debug)]
@@ -61,6 +143,7 @@ pub enum ResponseEvent {
     },
     OutputTextDelta(String),
     ReasoningSummaryDelta(String),
+    ReasoningContentDelta(String),
 }
 
 #[derive(Debug, Serialize)]
@@ -134,14 +217,12 @@ pub(crate) struct ResponsesApiRequest<'a> {
     pub(crate) include: Vec<String>,
 }
 
-use crate::config::Config;
-
 pub(crate) fn create_reasoning_param_for_request(
-    config: &Config,
+    model_family: &ModelFamily,
     effort: ReasoningEffortConfig,
     summary: ReasoningSummaryConfig,
 ) -> Option<Reasoning> {
-    if model_supports_reasoning_summaries(config) {
+    if model_family.supports_reasoning_summaries {
         let effort: Option<OpenAiReasoningEffort> = effort.into();
         let effort = effort?;
         Some(Reasoning {
@@ -153,27 +234,6 @@ pub(crate) fn create_reasoning_param_for_request(
     }
 }
 
-pub fn model_supports_reasoning_summaries(config: &Config) -> bool {
-    // Currently, we hardcode this rule to decide whether to enable reasoning.
-    // We expect reasoning to apply only to OpenAI models, but we do not want
-    // users to have to mess with their config to disable reasoning for models
-    // that do not support it, such as `gpt-4.1`.
-    //
-    // Though if a user is using Codex with non-OpenAI models that, say, happen
-    // to start with "o", then they can set `model_reasoning_effort = "none"` in
-    // config.toml to disable reasoning.
-    //
-    // Converseley, if a user has a non-OpenAI provider that supports reasoning,
-    // they can set the top-level `model_supports_reasoning_summaries = true`
-    // config option to enable reasoning.
-    if config.model_supports_reasoning_summaries {
-        return true;
-    }
-
-    let model = &config.model;
-    model.starts_with("o") || model.starts_with("codex")
-}
-
 pub(crate) struct ResponseStream {
     pub(crate) rx_event: mpsc::Receiver<Result<ResponseEvent>>,
 }
@@ -188,6 +248,9 @@ impl Stream for ResponseStream {
 
 #[cfg(test)]
 mod tests {
+    #![allow(clippy::expect_used)]
+    use crate::model_family::find_family_for_model;
+
     use super::*;
 
     #[test]
@@ -197,7 +260,8 @@ mod tests {
             ..Default::default()
         };
         let expected = format!("{BASE_INSTRUCTIONS}\n{APPLY_PATCH_TOOL_INSTRUCTIONS}");
-        let full = prompt.get_full_instructions("gpt-4.1");
+        let model_family = find_family_for_model("gpt-4.1").expect("known model slug");
+        let full = prompt.get_full_instructions(&model_family);
         assert_eq!(full, expected);
     }
 }

codex-rs/core/src/codex.rs

@@ -37,6 +37,7 @@ use crate::apply_patch::convert_apply_patch_to_protocol;
 use crate::apply_patch::get_writable_roots;
 use crate::apply_patch::{self};
 use crate::client::ModelClient;
+use crate::client_common::EnvironmentContext;
 use crate::client_common::Prompt;
 use crate::client_common::ResponseEvent;
 use crate::config::Config;
@@ -56,16 +57,21 @@ use crate::mcp_tool_call::handle_mcp_tool_call;
 use crate::models::ContentItem;
 use crate::models::FunctionCallOutputPayload;
 use crate::models::LocalShellAction;
+use crate::models::ReasoningItemContent;
 use crate::models::ReasoningItemReasoningSummary;
 use crate::models::ResponseInputItem;
 use crate::models::ResponseItem;
 use crate::models::ShellToolCallParams;
+use crate::openai_tools::ToolsConfig;
+use crate::openai_tools::get_openai_tools;
 use crate::plan_tool::handle_update_plan;
 use crate::project_doc::get_user_instructions;
 use crate::protocol::AgentMessageDeltaEvent;
 use crate::protocol::AgentMessageEvent;
 use crate::protocol::AgentReasoningDeltaEvent;
 use crate::protocol::AgentReasoningEvent;
+use crate::protocol::AgentReasoningRawContentDeltaEvent;
+use crate::protocol::AgentReasoningRawContentEvent;
 use crate::protocol::ApplyPatchApprovalRequestEvent;
 use crate::protocol::AskForApproval;
 use crate::protocol::BackgroundEventEvent;
@@ -123,7 +129,7 @@ impl Codex {
         let resume_path = config.experimental_resume.clone();
         info!("resume_path: {resume_path:?}");
         let (tx_sub, rx_sub) = async_channel::bounded(64);
-        let (tx_event, rx_event) = async_channel::bounded(1600);
+        let (tx_event, rx_event) = async_channel::unbounded();
 
         let user_instructions = get_user_instructions(&config).await;
 
@@ -213,6 +219,7 @@ pub(crate) struct Session {
     shell_environment_policy: ShellEnvironmentPolicy,
     pub(crate) writable_roots: Mutex<Vec<PathBuf>>,
     disable_response_storage: bool,
+    tools_config: ToolsConfig,
 
     /// Manager for external MCP servers/tools.
     mcp_connection_manager: McpConnectionManager,
@@ -227,6 +234,7 @@ pub(crate) struct Session {
     state: Mutex<State>,
     codex_linux_sandbox_exe: Option<PathBuf>,
     user_shell: shell::Shell,
+    show_raw_agent_reasoning: bool,
 }
 
 impl Session {
@@ -701,7 +709,7 @@ async fn submission_loop(
                 cwd,
                 resume_path,
             } => {
-                info!(
+                debug!(
                     "Configuring session: model={model}; provider={provider:?}; resume={resume_path:?}"
                 );
                 if !cwd.is_absolute() {
@@ -806,6 +814,12 @@ async fn submission_loop(
                 let default_shell = shell::default_user_shell().await;
                 sess = Some(Arc::new(Session {
                     client,
+                    tools_config: ToolsConfig::new(
+                        &config.model_family,
+                        approval_policy,
+                        sandbox_policy.clone(),
+                        config.include_plan_tool,
+                    ),
                     tx_event: tx_event.clone(),
                     ctrl_c: Arc::clone(&ctrl_c),
                     user_instructions,
@@ -822,6 +836,7 @@ async fn submission_loop(
                     codex_linux_sandbox_exe: config.codex_linux_sandbox_exe.clone(),
                     disable_response_storage,
                     user_shell: default_shell,
+                    show_raw_agent_reasoning: config.show_raw_agent_reasoning,
                 }));
 
                 // Patch restored state into the newly created session.
@@ -1132,6 +1147,7 @@ async fn run_task(sess: Arc<Session>, sub_id: String, input: Vec<InputItem>) {
                             ResponseItem::Reasoning {
                                 id,
                                 summary,
+                                content,
                                 encrypted_content,
                             },
                             None,
@@ -1139,6 +1155,7 @@ async fn run_task(sess: Arc<Session>, sub_id: String, input: Vec<InputItem>) {
                             items_to_record_in_conversation_history.push(ResponseItem::Reasoning {
                                 id: id.clone(),
                                 summary: summary.clone(),
+                                content: content.clone(),
                                 encrypted_content: encrypted_content.clone(),
                             });
                         }
@@ -1197,13 +1214,22 @@ async fn run_turn(
     sub_id: String,
     input: Vec<ResponseItem>,
 ) -> CodexResult<Vec<ProcessedResponseItem>> {
-    let extra_tools = sess.mcp_connection_manager.list_all_tools();
+    let tools = get_openai_tools(
+        &sess.tools_config,
+        Some(sess.mcp_connection_manager.list_all_tools()),
+    );
+
     let prompt = Prompt {
         input,
         user_instructions: sess.user_instructions.clone(),
         store: !sess.disable_response_storage,
-        extra_tools,
+        tools,
         base_instructions_override: sess.base_instructions.clone(),
+        environment_context: Some(EnvironmentContext {
+            cwd: sess.cwd.clone(),
+            approval_policy: sess.approval_policy,
+            sandbox_policy: sess.sandbox_policy.clone(),
+        }),
     };
 
     let mut retries = 0;
@@ -1374,6 +1400,11 @@ async fn try_run_turn(
                 return Ok(output);
             }
             ResponseEvent::OutputTextDelta(delta) => {
+                {
+                    let mut st = sess.state.lock().unwrap();
+                    st.history.append_assistant_text(&delta);
+                }
+
                 let event = Event {
                     id: sub_id.to_string(),
                     msg: EventMsg::AgentMessageDelta(AgentMessageDeltaEvent { delta }),
@@ -1387,6 +1418,17 @@ async fn try_run_turn(
                 };
                 sess.tx_event.send(event).await.ok();
             }
+            ResponseEvent::ReasoningContentDelta(delta) => {
+                if sess.show_raw_agent_reasoning {
+                    let event = Event {
+                        id: sub_id.to_string(),
+                        msg: EventMsg::AgentReasoningRawContentDelta(
+                            AgentReasoningRawContentDeltaEvent { delta },
+                        ),
+                    };
+                    sess.tx_event.send(event).await.ok();
+                }
+            }
         }
     }
 }
@@ -1413,7 +1455,8 @@ async fn run_compact_task(
         input: turn_input,
         user_instructions: None,
         store: !sess.disable_response_storage,
-        extra_tools: HashMap::new(),
+        environment_context: None,
+        tools: Vec::new(),
         base_instructions_override: Some(compact_instructions.clone()),
     };
 
@@ -1493,7 +1536,12 @@ async fn handle_response_item(
             }
             None
         }
-        ResponseItem::Reasoning { summary, .. } => {
+        ResponseItem::Reasoning {
+            id: _,
+            summary,
+            content,
+            encrypted_content: _,
+        } => {
             for item in summary {
                 let text = match item {
                     ReasoningItemReasoningSummary::SummaryText { text } => text,
@@ -1504,6 +1552,21 @@ async fn handle_response_item(
                 };
                 sess.tx_event.send(event).await.ok();
             }
+            if sess.show_raw_agent_reasoning && content.is_some() {
+                let content = content.unwrap();
+                for item in content {
+                    let text = match item {
+                        ReasoningItemContent::ReasoningText { text } => text,
+                    };
+                    let event = Event {
+                        id: sub_id.to_string(),
+                        msg: EventMsg::AgentReasoningRawContent(AgentReasoningRawContentEvent {
+                            text,
+                        }),
+                    };
+                    sess.tx_event.send(event).await.ok();
+                }
+            }
             None
         }
         ResponseItem::FunctionCall {
@@ -1537,6 +1600,8 @@ async fn handle_response_item(
                 command: action.command,
                 workdir: action.working_directory,
                 timeout_ms: action.timeout_ms,
+                with_escalated_permissions: None,
+                justification: None,
             };
             let effective_call_id = match (call_id, id) {
                 (Some(call_id), _) => call_id,
@@ -1625,6 +1690,8 @@ fn to_exec_params(params: ShellToolCallParams, sess: &Session) -> ExecParams {
         cwd: sess.resolve_path(params.workdir.clone()),
         timeout_ms: params.timeout_ms,
         env: create_env(&sess.shell_environment_policy),
+        with_escalated_permissions: params.with_escalated_permissions,
+        justification: params.justification,
     }
 }
 
@@ -1725,13 +1792,19 @@ async fn handle_container_exec_with_params(
                 cwd: cwd.clone(),
                 timeout_ms: params.timeout_ms,
                 env: HashMap::new(),
+                with_escalated_permissions: params.with_escalated_permissions,
+                justification: params.justification.clone(),
             };
             let safety = if *user_explicitly_approved_this_action {
                 SafetyCheck::AutoApprove {
                     sandbox_type: SandboxType::None,
                 }
             } else {
-                assess_safety_for_untrusted_command(sess.approval_policy, &sess.sandbox_policy)
+                assess_safety_for_untrusted_command(
+                    sess.approval_policy,
+                    &sess.sandbox_policy,
+                    params.with_escalated_permissions.unwrap_or(false),
+                )
             };
             (
                 params,
@@ -1747,6 +1820,7 @@ async fn handle_container_exec_with_params(
                     sess.approval_policy,
                     &sess.sandbox_policy,
                     &state.approved_commands,
+                    params.with_escalated_permissions.unwrap_or(false),
                 )
             };
             let command_for_display = params.command.clone();
@@ -1763,7 +1837,7 @@ async fn handle_container_exec_with_params(
                     call_id.clone(),
                     params.command.clone(),
                     params.cwd.clone(),
-                    None,
+                    params.justification.clone(),
                 )
                 .await;
             match rx_approve.await.unwrap_or_default() {
@@ -1901,13 +1975,31 @@ async fn handle_sandbox_error(
     let cwd = exec_command_context.cwd.clone();
     let is_apply_patch = exec_command_context.apply_patch.is_some();
 
-    // Early out if the user never wants to be asked for approval; just return to the model immediately
-    if sess.approval_policy == AskForApproval::Never {
+    // Early out if either the user never wants to be asked for approval, or
+    // we're letting the model manage escalation requests. Otherwise, continue
+    match sess.approval_policy {
+        AskForApproval::Never | AskForApproval::OnRequest => {
+            return ResponseInputItem::FunctionCallOutput {
+                call_id,
+                output: FunctionCallOutputPayload {
+                    content: format!(
+                        "failed in sandbox {sandbox_type:?} with execution error: {error}"
+                    ),
+                    success: Some(false),
+                },
+            };
+        }
+        AskForApproval::UnlessTrusted | AskForApproval::OnFailure => (),
+    }
+
+    // similarly, if the command timed out, we can simply return this failure to the model
+    if matches!(error, SandboxErr::Timeout) {
         return ResponseInputItem::FunctionCallOutput {
             call_id,
             output: FunctionCallOutputPayload {
                 content: format!(
-                    "failed in sandbox {sandbox_type:?} with execution error: {error}"
+                    "command timed out after {} milliseconds",
+                    params.timeout_duration().as_millis()
                 ),
                 success: Some(false),
             },
@@ -1921,7 +2013,8 @@ async fn handle_sandbox_error(
     // include additional metadata on the command to indicate whether non-zero
     // exit codes merit a retry.
 
-    // For now, we categorically ask the user to retry without sandbox.
+    // For now, we categorically ask the user to retry without sandbox and
+    // emit the raw error as a background event.
     sess.notify_background_event(&sub_id, format!("Execution failed: {error}"))
         .await;
 

codex-rs/core/src/config.rs

@@ -10,6 +10,8 @@ use crate::config_types::ShellEnvironmentPolicyToml;
 use crate::config_types::Tui;
 use crate::config_types::UriBasedFileOpener;
 use crate::flags::OPENAI_DEFAULT_MODEL;
+use crate::model_family::ModelFamily;
+use crate::model_family::find_family_for_model;
 use crate::model_provider_info::ModelProviderInfo;
 use crate::model_provider_info::built_in_model_providers;
 use crate::openai_model_info::get_model_info;
@@ -33,6 +35,8 @@ pub struct Config {
     /// Optional override of model selection.
     pub model: String,
 
+    pub model_family: ModelFamily,
+
     /// Size of the context window for the model, in tokens.
     pub model_context_window: Option<u64>,
 
@@ -57,6 +61,10 @@ pub struct Config {
     /// users are only interested in the final agent responses.
     pub hide_agent_reasoning: bool,
 
+    /// When set to `true`, `AgentReasoningRawContentEvent` events will be shown in the UI/output.
+    /// Defaults to `false`.
+    pub show_raw_agent_reasoning: bool,
+
     /// Disable server-side response storage (sends the full conversation
     /// context with every request). Currently necessary for OpenAI customers
     /// who have opted into Zero Data Retention (ZDR).
@@ -134,10 +142,6 @@ pub struct Config {
     /// request using the Responses API.
     pub model_reasoning_summary: ReasoningSummary,
 
-    /// When set to `true`, overrides the default heuristic and forces
-    /// `model_supports_reasoning_summaries()` to return `true`.
-    pub model_supports_reasoning_summaries: bool,
-
     /// Base URL for requests to ChatGPT (as opposed to the OpenAI API).
     pub chatgpt_base_url: String,
 
@@ -325,6 +329,10 @@ pub struct ConfigToml {
     /// UI/output. Defaults to `false`.
     pub hide_agent_reasoning: Option<bool>,
 
+    /// When set to `true`, `AgentReasoningRawContentEvent` events will be shown in the UI/output.
+    /// Defaults to `false`.
+    pub show_raw_agent_reasoning: Option<bool>,
+
     pub model_reasoning_effort: Option<ReasoningEffort>,
     pub model_reasoning_summary: Option<ReasoningSummary>,
 
@@ -377,6 +385,8 @@ pub struct ConfigOverrides {
     pub codex_linux_sandbox_exe: Option<PathBuf>,
     pub base_instructions: Option<String>,
     pub include_plan_tool: Option<bool>,
+    pub disable_response_storage: Option<bool>,
+    pub show_raw_agent_reasoning: Option<bool>,
 }
 
 impl Config {
@@ -400,6 +410,8 @@ impl Config {
             codex_linux_sandbox_exe,
             base_instructions,
             include_plan_tool,
+            disable_response_storage,
+            show_raw_agent_reasoning,
         } = overrides;
 
         let config_profile = match config_profile_key.as_ref().or(cfg.profile.as_ref()) {
@@ -465,7 +477,19 @@ impl Config {
             .or(config_profile.model)
             .or(cfg.model)
             .unwrap_or_else(default_model);
-        let openai_model_info = get_model_info(&model);
+        let model_family = find_family_for_model(&model).unwrap_or_else(|| {
+            let supports_reasoning_summaries =
+                cfg.model_supports_reasoning_summaries.unwrap_or(false);
+            ModelFamily {
+                slug: model.clone(),
+                family: model.clone(),
+                needs_special_apply_patch_instructions: false,
+                supports_reasoning_summaries,
+                uses_local_shell_tool: false,
+            }
+        });
+
+        let openai_model_info = get_model_info(&model_family);
         let model_context_window = cfg
             .model_context_window
             .or_else(|| openai_model_info.as_ref().map(|info| info.context_window));
@@ -490,6 +514,7 @@ impl Config {
 
         let config = Self {
             model,
+            model_family,
             model_context_window,
             model_max_output_tokens,
             model_provider_id,
@@ -504,6 +529,7 @@ impl Config {
             disable_response_storage: config_profile
                 .disable_response_storage
                 .or(cfg.disable_response_storage)
+                .or(disable_response_storage)
                 .unwrap_or(false),
             notify: cfg.notify,
             user_instructions,
@@ -518,6 +544,10 @@ impl Config {
             codex_linux_sandbox_exe,
 
             hide_agent_reasoning: cfg.hide_agent_reasoning.unwrap_or(false),
+            show_raw_agent_reasoning: cfg
+                .show_raw_agent_reasoning
+                .or(show_raw_agent_reasoning)
+                .unwrap_or(false),
             model_reasoning_effort: config_profile
                 .model_reasoning_effort
                 .or(cfg.model_reasoning_effort)
@@ -527,10 +557,6 @@ impl Config {
                 .or(cfg.model_reasoning_summary)
                 .unwrap_or_default(),
 
-            model_supports_reasoning_summaries: cfg
-                .model_supports_reasoning_summaries
-                .unwrap_or(false),
-
             chatgpt_base_url: config_profile
                 .chatgpt_base_url
                 .or(cfg.chatgpt_base_url)
@@ -871,6 +897,7 @@ disable_response_storage = true
         assert_eq!(
             Config {
                 model: "o3".to_string(),
+                model_family: find_family_for_model("o3").expect("known model slug"),
                 model_context_window: Some(200_000),
                 model_max_output_tokens: Some(100_000),
                 model_provider_id: "openai".to_string(),
@@ -891,9 +918,9 @@ disable_response_storage = true
                 tui: Tui::default(),
                 codex_linux_sandbox_exe: None,
                 hide_agent_reasoning: false,
+                show_raw_agent_reasoning: false,
                 model_reasoning_effort: ReasoningEffort::High,
                 model_reasoning_summary: ReasoningSummary::Detailed,
-                model_supports_reasoning_summaries: false,
                 chatgpt_base_url: "https://chatgpt.com/backend-api/".to_string(),
                 experimental_resume: None,
                 base_instructions: None,
@@ -921,6 +948,7 @@ disable_response_storage = true
         )?;
         let expected_gpt3_profile_config = Config {
             model: "gpt-3.5-turbo".to_string(),
+            model_family: find_family_for_model("gpt-3.5-turbo").expect("known model slug"),
             model_context_window: Some(16_385),
             model_max_output_tokens: Some(4_096),
             model_provider_id: "openai-chat-completions".to_string(),
@@ -941,9 +969,9 @@ disable_response_storage = true
             tui: Tui::default(),
             codex_linux_sandbox_exe: None,
             hide_agent_reasoning: false,
+            show_raw_agent_reasoning: false,
             model_reasoning_effort: ReasoningEffort::default(),
             model_reasoning_summary: ReasoningSummary::default(),
-            model_supports_reasoning_summaries: false,
             chatgpt_base_url: "https://chatgpt.com/backend-api/".to_string(),
             experimental_resume: None,
             base_instructions: None,
@@ -986,6 +1014,7 @@ disable_response_storage = true
         )?;
         let expected_zdr_profile_config = Config {
             model: "o3".to_string(),
+            model_family: find_family_for_model("o3").expect("known model slug"),
             model_context_window: Some(200_000),
             model_max_output_tokens: Some(100_000),
             model_provider_id: "openai".to_string(),
@@ -1006,9 +1035,9 @@ disable_response_storage = true
             tui: Tui::default(),
             codex_linux_sandbox_exe: None,
             hide_agent_reasoning: false,
+            show_raw_agent_reasoning: false,
             model_reasoning_effort: ReasoningEffort::default(),
             model_reasoning_summary: ReasoningSummary::default(),
-            model_supports_reasoning_summaries: false,
             chatgpt_base_url: "https://chatgpt.com/backend-api/".to_string(),
             experimental_resume: None,
             base_instructions: None,

codex-rs/core/src/conversation_history.rs

@@ -24,9 +24,52 @@ impl ConversationHistory {
         I::Item: std::ops::Deref<Target = ResponseItem>,
     {
         for item in items {
-            if is_api_message(&item) {
-                // Note agent-loop.ts also does filtering on some of the fields.
-                self.items.push(item.clone());
+            if !is_api_message(&item) {
+                continue;
+            }
+
+            // Merge adjacent assistant messages into a single history entry.
+            // This prevents duplicates when a partial assistant message was
+            // streamed into history earlier in the turn and the final full
+            // message is recorded at turn end.
+            match (&*item, self.items.last_mut()) {
+                (
+                    ResponseItem::Message {
+                        role: new_role,
+                        content: new_content,
+                        ..
+                    },
+                    Some(ResponseItem::Message {
+                        role: last_role,
+                        content: last_content,
+                        ..
+                    }),
+                ) if new_role == "assistant" && last_role == "assistant" => {
+                    append_text_content(last_content, new_content);
+                }
+                _ => {
+                    self.items.push(item.clone());
+                }
+            }
+        }
+    }
+
+    /// Append a text `delta` to the latest assistant message, creating a new
+    /// assistant entry if none exists yet (e.g. first delta for this turn).
+    pub(crate) fn append_assistant_text(&mut self, delta: &str) {
+        match self.items.last_mut() {
+            Some(ResponseItem::Message { role, content, .. }) if role == "assistant" => {
+                append_text_delta(content, delta);
+            }
+            _ => {
+                // Start a new assistant message with the delta.
+                self.items.push(ResponseItem::Message {
+                    id: None,
+                    role: "assistant".to_string(),
+                    content: vec![crate::models::ContentItem::OutputText {
+                        text: delta.to_string(),
+                    }],
+                });
             }
         }
     }
@@ -72,3 +115,140 @@ fn is_api_message(message: &ResponseItem) -> bool {
         ResponseItem::Other => false,
     }
 }
+
+/// Helper to append the textual content from `src` into `dst` in place.
+fn append_text_content(
+    dst: &mut Vec<crate::models::ContentItem>,
+    src: &Vec<crate::models::ContentItem>,
+) {
+    for c in src {
+        if let crate::models::ContentItem::OutputText { text } = c {
+            append_text_delta(dst, text);
+        }
+    }
+}
+
+/// Append a single text delta to the last OutputText item in `content`, or
+/// push a new OutputText item if none exists.
+fn append_text_delta(content: &mut Vec<crate::models::ContentItem>, delta: &str) {
+    if let Some(crate::models::ContentItem::OutputText { text }) = content
+        .iter_mut()
+        .rev()
+        .find(|c| matches!(c, crate::models::ContentItem::OutputText { .. }))
+    {
+        text.push_str(delta);
+    } else {
+        content.push(crate::models::ContentItem::OutputText {
+            text: delta.to_string(),
+        });
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use crate::models::ContentItem;
+
+    fn assistant_msg(text: &str) -> ResponseItem {
+        ResponseItem::Message {
+            id: None,
+            role: "assistant".to_string(),
+            content: vec![ContentItem::OutputText {
+                text: text.to_string(),
+            }],
+        }
+    }
+
+    fn user_msg(text: &str) -> ResponseItem {
+        ResponseItem::Message {
+            id: None,
+            role: "user".to_string(),
+            content: vec![ContentItem::OutputText {
+                text: text.to_string(),
+            }],
+        }
+    }
+
+    #[test]
+    fn merges_adjacent_assistant_messages() {
+        let mut h = ConversationHistory::default();
+        let a1 = assistant_msg("Hello");
+        let a2 = assistant_msg(", world!");
+        h.record_items([&a1, &a2]);
+
+        let items = h.contents();
+        assert_eq!(
+            items,
+            vec![ResponseItem::Message {
+                id: None,
+                role: "assistant".to_string(),
+                content: vec![ContentItem::OutputText {
+                    text: "Hello, world!".to_string()
+                }]
+            }]
+        );
+    }
+
+    #[test]
+    fn append_assistant_text_creates_and_appends() {
+        let mut h = ConversationHistory::default();
+        h.append_assistant_text("Hello");
+        h.append_assistant_text(", world");
+
+        // Now record a final full assistant message and verify it merges.
+        let final_msg = assistant_msg("!");
+        h.record_items([&final_msg]);
+
+        let items = h.contents();
+        assert_eq!(
+            items,
+            vec![ResponseItem::Message {
+                id: None,
+                role: "assistant".to_string(),
+                content: vec![ContentItem::OutputText {
+                    text: "Hello, world!".to_string()
+                }]
+            }]
+        );
+    }
+
+    #[test]
+    fn filters_non_api_messages() {
+        let mut h = ConversationHistory::default();
+        // System message is not an API message; Other is ignored.
+        let system = ResponseItem::Message {
+            id: None,
+            role: "system".to_string(),
+            content: vec![ContentItem::OutputText {
+                text: "ignored".to_string(),
+            }],
+        };
+        h.record_items([&system, &ResponseItem::Other]);
+
+        // User and assistant should be retained.
+        let u = user_msg("hi");
+        let a = assistant_msg("hello");
+        h.record_items([&u, &a]);
+
+        let items = h.contents();
+        assert_eq!(
+            items,
+            vec![
+                ResponseItem::Message {
+                    id: None,
+                    role: "user".to_string(),
+                    content: vec![ContentItem::OutputText {
+                        text: "hi".to_string()
+                    }]
+                },
+                ResponseItem::Message {
+                    id: None,
+                    role: "assistant".to_string(),
+                    content: vec![ContentItem::OutputText {
+                        text: "hello".to_string()
+                    }]
+                }
+            ]
+        );
+    }
+}

codex-rs/core/src/exec.rs

@@ -49,6 +49,14 @@ pub struct ExecParams {
     pub cwd: PathBuf,
     pub timeout_ms: Option<u64>,
     pub env: HashMap<String, String>,
+    pub with_escalated_permissions: Option<bool>,
+    pub justification: Option<String>,
+}
+
+impl ExecParams {
+    pub fn timeout_duration(&self) -> Duration {
+        Duration::from_millis(self.timeout_ms.unwrap_or(DEFAULT_TIMEOUT_MS))
+    }
 }
 
 #[derive(Clone, Copy, Debug, PartialEq)]
@@ -83,11 +91,9 @@ pub async fn process_exec_tool_call(
     {
         SandboxType::None => exec(params, sandbox_policy, ctrl_c, stdout_stream.clone()).await,
         SandboxType::MacosSeatbelt => {
+            let timeout = params.timeout_duration();
             let ExecParams {
-                command,
-                cwd,
-                timeout_ms,
-                env,
+                command, cwd, env, ..
             } = params;
             let child = spawn_command_under_seatbelt(
                 command,
@@ -97,14 +103,12 @@ pub async fn process_exec_tool_call(
                 env,
             )
             .await?;
-            consume_truncated_output(child, ctrl_c, timeout_ms, stdout_stream.clone()).await
+            consume_truncated_output(child, ctrl_c, timeout, stdout_stream.clone()).await
         }
         SandboxType::LinuxSeccomp => {
+            let timeout = params.timeout_duration();
             let ExecParams {
-                command,
-                cwd,
-                timeout_ms,
-                env,
+                command, cwd, env, ..
             } = params;
 
             let codex_linux_sandbox_exe = codex_linux_sandbox_exe
@@ -120,7 +124,7 @@ pub async fn process_exec_tool_call(
             )
             .await?;
 
-            consume_truncated_output(child, ctrl_c, timeout_ms, stdout_stream).await
+            consume_truncated_output(child, ctrl_c, timeout, stdout_stream).await
         }
     };
     let duration = start.elapsed();
@@ -255,16 +259,16 @@ pub struct ExecToolCallOutput {
 }
 
 async fn exec(
-    ExecParams {
-        command,
-        cwd,
-        timeout_ms,
-        env,
-    }: ExecParams,
+    params: ExecParams,
     sandbox_policy: &SandboxPolicy,
     ctrl_c: Arc<Notify>,
     stdout_stream: Option<StdoutStream>,
 ) -> Result<RawExecToolCallOutput> {
+    let timeout = params.timeout_duration();
+    let ExecParams {
+        command, cwd, env, ..
+    } = params;
+
     let (program, args) = command.split_first().ok_or_else(|| {
         CodexErr::Io(io::Error::new(
             io::ErrorKind::InvalidInput,
@@ -282,7 +286,7 @@ async fn exec(
         env,
     )
     .await?;
-    consume_truncated_output(child, ctrl_c, timeout_ms, stdout_stream).await
+    consume_truncated_output(child, ctrl_c, timeout, stdout_stream).await
 }
 
 /// Consumes the output of a child process, truncating it so it is suitable for
@@ -290,7 +294,7 @@ async fn exec(
 pub(crate) async fn consume_truncated_output(
     mut child: Child,
     ctrl_c: Arc<Notify>,
-    timeout_ms: Option<u64>,
+    timeout: Duration,
     stdout_stream: Option<StdoutStream>,
 ) -> Result<RawExecToolCallOutput> {
     // Both stdout and stderr were configured with `Stdio::piped()`
@@ -324,7 +328,6 @@ pub(crate) async fn consume_truncated_output(
     ));
 
     let interrupted = ctrl_c.notified();
-    let timeout = Duration::from_millis(timeout_ms.unwrap_or(DEFAULT_TIMEOUT_MS));
     let exit_status = tokio::select! {
         result = tokio::time::timeout(timeout, child.wait()) => {
             match result {

codex-rs/core/src/git_info.rs

@@ -9,7 +9,7 @@ use tokio::time::timeout;
 /// Timeout for git commands to prevent freezing on large repositories
 const GIT_COMMAND_TIMEOUT: TokioDuration = TokioDuration::from_secs(5);
 
-#[derive(Serialize, Deserialize, Clone)]
+#[derive(Serialize, Deserialize, Clone, Debug)]
 pub struct GitInfo {
     /// Current commit hash (SHA)
     #[serde(skip_serializing_if = "Option::is_none")]

codex-rs/core/src/lib.rs

@@ -28,9 +28,12 @@ mod mcp_connection_manager;
 mod mcp_tool_call;
 mod message_history;
 mod model_provider_info;
+pub use model_provider_info::BUILT_IN_OSS_MODEL_PROVIDER_ID;
 pub use model_provider_info::ModelProviderInfo;
 pub use model_provider_info::WireApi;
 pub use model_provider_info::built_in_model_providers;
+pub use model_provider_info::create_oss_provider_with_base_url;
+pub mod model_family;
 mod models;
 mod openai_model_info;
 mod openai_tools;
@@ -45,7 +48,5 @@ pub mod spawn;
 pub mod turn_diff_tracker;
 mod user_notification;
 pub mod util;
-
 pub use apply_patch::CODEX_APPLY_PATCH_ARG1;
-pub use client_common::model_supports_reasoning_summaries;
 pub use safety::get_platform_sandbox;

codex-rs/core/src/model_family.rs

@@ -0,0 +1,95 @@
+/// A model family is a group of models that share certain characteristics.
+#[derive(Debug, Clone, PartialEq, Eq, Hash)]
+pub struct ModelFamily {
+    /// The full model slug used to derive this model family, e.g.
+    /// "gpt-4.1-2025-04-14".
+    pub slug: String,
+
+    /// The model family name, e.g. "gpt-4.1". Note this should able to be used
+    /// with [`crate::openai_model_info::get_model_info`].
+    pub family: String,
+
+    /// True if the model needs additional instructions on how to use the
+    /// "virtual" `apply_patch` CLI.
+    pub needs_special_apply_patch_instructions: bool,
+
+    // Whether the `reasoning` field can be set when making a request to this
+    // model family. Note it has `effort` and `summary` subfields (though
+    // `summary` is optional).
+    pub supports_reasoning_summaries: bool,
+
+    // This should be set to true when the model expects a tool named
+    // "local_shell" to be provided. Its contract must be understood natively by
+    // the model such that its description can be omitted.
+    // See https://platform.openai.com/docs/guides/tools-local-shell
+    pub uses_local_shell_tool: bool,
+}
+
+macro_rules! model_family {
+    (
+        $slug:expr, $family:expr $(, $key:ident : $value:expr )* $(,)?
+    ) => {{
+        // defaults
+        let mut mf = ModelFamily {
+            slug: $slug.to_string(),
+            family: $family.to_string(),
+            needs_special_apply_patch_instructions: false,
+            supports_reasoning_summaries: false,
+            uses_local_shell_tool: false,
+        };
+        // apply overrides
+        $(
+            mf.$key = $value;
+        )*
+        Some(mf)
+    }};
+}
+
+macro_rules! simple_model_family {
+    (
+        $slug:expr, $family:expr
+    ) => {{
+        Some(ModelFamily {
+            slug: $slug.to_string(),
+            family: $family.to_string(),
+            needs_special_apply_patch_instructions: false,
+            supports_reasoning_summaries: false,
+            uses_local_shell_tool: false,
+        })
+    }};
+}
+
+/// Returns a `ModelFamily` for the given model slug, or `None` if the slug
+/// does not match any known model family.
+pub fn find_family_for_model(slug: &str) -> Option<ModelFamily> {
+    if slug.starts_with("o3") {
+        model_family!(
+            slug, "o3",
+            supports_reasoning_summaries: true,
+        )
+    } else if slug.starts_with("o4-mini") {
+        model_family!(
+            slug, "o4-mini",
+            supports_reasoning_summaries: true,
+        )
+    } else if slug.starts_with("codex-mini-latest") {
+        model_family!(
+            slug, "codex-mini-latest",
+            supports_reasoning_summaries: true,
+            uses_local_shell_tool: true,
+        )
+    } else if slug.starts_with("gpt-4.1") {
+        model_family!(
+            slug, "gpt-4.1",
+            needs_special_apply_patch_instructions: true,
+        )
+    } else if slug.starts_with("gpt-4o") {
+        simple_model_family!(slug, "gpt-4o")
+    } else if slug.starts_with("gpt-oss") {
+        simple_model_family!(slug, "gpt-oss")
+    } else if slug.starts_with("gpt-3.5") {
+        simple_model_family!(slug, "gpt-3.5")
+    } else {
+        None
+    }
+}

codex-rs/core/src/model_provider_info.rs

@@ -5,8 +5,11 @@
 //!   2. User-defined entries inside `~/.codex/config.toml` under the `model_providers`
 //!      key. These override or extend the defaults at runtime.
 
+use codex_login::AuthMode;
+use codex_login::CodexAuth;
 use serde::Deserialize;
 use serde::Serialize;
+use std::borrow::Cow;
 use std::collections::HashMap;
 use std::env::VarError;
 use std::time::Duration;
@@ -88,25 +91,30 @@ impl ModelProviderInfo {
     /// When `require_api_key` is true and the provider declares an `env_key`
     /// but the variable is missing/empty, returns an [`Err`] identical to the
     /// one produced by [`ModelProviderInfo::api_key`].
-    pub fn create_request_builder<'a>(
+    pub async fn create_request_builder<'a>(
         &'a self,
         client: &'a reqwest::Client,
+        auth: &Option<CodexAuth>,
     ) -> crate::error::Result<reqwest::RequestBuilder> {
-        let url = self.get_full_url();
+        let auth: Cow<'_, Option<CodexAuth>> = if auth.is_some() {
+            Cow::Borrowed(auth)
+        } else {
+            Cow::Owned(self.get_fallback_auth()?)
+        };
+
+        let url = self.get_full_url(&auth);
 
         let mut builder = client.post(url);
 
-        let api_key = self.api_key()?;
-        if let Some(key) = api_key {
-            builder = builder.bearer_auth(key);
+        if let Some(auth) = auth.as_ref() {
+            builder = builder.bearer_auth(auth.get_token().await?);
         }
 
         Ok(self.apply_http_headers(builder))
     }
 
-    pub(crate) fn get_full_url(&self) -> String {
-        let query_string = self
-            .query_params
+    fn get_query_string(&self) -> String {
+        self.query_params
             .as_ref()
             .map_or_else(String::new, |params| {
                 let full_params = params
@@ -115,16 +123,29 @@ impl ModelProviderInfo {
                     .collect::<Vec<_>>()
                     .join("&");
                 format!("?{full_params}")
-            });
+            })
+    }
+
+    pub(crate) fn get_full_url(&self, auth: &Option<CodexAuth>) -> String {
+        let default_base_url = if matches!(
+            auth,
+            Some(CodexAuth {
+                mode: AuthMode::ChatGPT,
+                ..
+            })
+        ) {
+            "https://chatgpt.com/backend-api/codex"
+        } else {
+            "https://api.openai.com/v1"
+        };
+        let query_string = self.get_query_string();
         let base_url = self
             .base_url
             .clone()
-            .unwrap_or("https://api.openai.com/v1".to_string());
+            .unwrap_or(default_base_url.to_string());
 
         match self.wire_api {
-            WireApi::Responses => {
-                format!("{base_url}/responses{query_string}")
-            }
+            WireApi::Responses => format!("{base_url}/responses{query_string}"),
             WireApi::Chat => format!("{base_url}/chat/completions{query_string}"),
         }
     }
@@ -132,10 +153,7 @@ impl ModelProviderInfo {
     /// Apply provider-specific HTTP headers (both static and environment-based)
     /// onto an existing `reqwest::RequestBuilder` and return the updated
     /// builder.
-    pub fn apply_http_headers(
-        &self,
-        mut builder: reqwest::RequestBuilder,
-    ) -> reqwest::RequestBuilder {
+    fn apply_http_headers(&self, mut builder: reqwest::RequestBuilder) -> reqwest::RequestBuilder {
         if let Some(extra) = &self.http_headers {
             for (k, v) in extra {
                 builder = builder.header(k, v);
@@ -157,7 +175,7 @@ impl ModelProviderInfo {
     /// If `env_key` is Some, returns the API key for this provider if present
     /// (and non-empty) in the environment. If `env_key` is required but
     /// cannot be found, returns an error.
-    fn api_key(&self) -> crate::error::Result<Option<String>> {
+    pub fn api_key(&self) -> crate::error::Result<Option<String>> {
         match &self.env_key {
             Some(env_key) => {
                 let env_value = std::env::var(env_key);
@@ -198,60 +216,113 @@ impl ModelProviderInfo {
             .map(Duration::from_millis)
             .unwrap_or(Duration::from_millis(DEFAULT_STREAM_IDLE_TIMEOUT_MS))
     }
+
+    fn get_fallback_auth(&self) -> crate::error::Result<Option<CodexAuth>> {
+        let api_key = self.api_key()?;
+        if let Some(api_key) = api_key {
+            return Ok(Some(CodexAuth::from_api_key(api_key)));
+        }
+        Ok(None)
+    }
 }
 
+const DEFAULT_OLLAMA_PORT: u32 = 11434;
+
+pub const BUILT_IN_OSS_MODEL_PROVIDER_ID: &str = "oss";
+
 /// Built-in default provider list.
 pub fn built_in_model_providers() -> HashMap<String, ModelProviderInfo> {
     use ModelProviderInfo as P;
 
     // We do not want to be in the business of adjucating which third-party
-    // providers are bundled with Codex CLI, so we only include the OpenAI
-    // provider by default. Users are encouraged to add to `model_providers`
-    // in config.toml to add their own providers.
-    [(
-        "openai",
-        P {
-            name: "OpenAI".into(),
-            // Allow users to override the default OpenAI endpoint by
-            // exporting `OPENAI_BASE_URL`. This is useful when pointing
-            // Codex at a proxy, mock server, or Azure-style deployment
-            // without requiring a full TOML override for the built-in
-            // OpenAI provider.
-            base_url: std::env::var("OPENAI_BASE_URL")
-                .ok()
-                .filter(|v| !v.trim().is_empty()),
-            env_key: None,
-            env_key_instructions: None,
-            wire_api: WireApi::Responses,
-            query_params: None,
-            http_headers: Some(
-                [("version".to_string(), env!("CARGO_PKG_VERSION").to_string())]
+    // providers are bundled with Codex CLI, so we only include the OpenAI and
+    // open source ("oss") providers by default. Users are encouraged to add to
+    // `model_providers` in config.toml to add their own providers.
+    [
+        (
+            "openai",
+            P {
+                name: "OpenAI".into(),
+                // Allow users to override the default OpenAI endpoint by
+                // exporting `OPENAI_BASE_URL`. This is useful when pointing
+                // Codex at a proxy, mock server, or Azure-style deployment
+                // without requiring a full TOML override for the built-in
+                // OpenAI provider.
+                base_url: std::env::var("OPENAI_BASE_URL")
+                    .ok()
+                    .filter(|v| !v.trim().is_empty()),
+                env_key: None,
+                env_key_instructions: None,
+                wire_api: WireApi::Responses,
+                query_params: None,
+                http_headers: Some(
+                    [("version".to_string(), env!("CARGO_PKG_VERSION").to_string())]
+                        .into_iter()
+                        .collect(),
+                ),
+                env_http_headers: Some(
+                    [
+                        (
+                            "OpenAI-Organization".to_string(),
+                            "OPENAI_ORGANIZATION".to_string(),
+                        ),
+                        ("OpenAI-Project".to_string(), "OPENAI_PROJECT".to_string()),
+                    ]
                     .into_iter()
                     .collect(),
-            ),
-            env_http_headers: Some(
-                [
-                    (
-                        "OpenAI-Organization".to_string(),
-                        "OPENAI_ORGANIZATION".to_string(),
-                    ),
-                    ("OpenAI-Project".to_string(), "OPENAI_PROJECT".to_string()),
-                ]
-                .into_iter()
-                .collect(),
-            ),
-            // Use global defaults for retry/timeout unless overridden in config.toml.
-            request_max_retries: None,
-            stream_max_retries: None,
-            stream_idle_timeout_ms: None,
-            requires_auth: true,
-        },
-    )]
+                ),
+                // Use global defaults for retry/timeout unless overridden in config.toml.
+                request_max_retries: None,
+                stream_max_retries: None,
+                stream_idle_timeout_ms: None,
+                requires_auth: true,
+            },
+        ),
+        (BUILT_IN_OSS_MODEL_PROVIDER_ID, create_oss_provider()),
+    ]
     .into_iter()
     .map(|(k, v)| (k.to_string(), v))
     .collect()
 }
 
+pub fn create_oss_provider() -> ModelProviderInfo {
+    // These CODEX_OSS_ environment variables are experimental: we may
+    // switch to reading values from config.toml instead.
+    let codex_oss_base_url = match std::env::var("CODEX_OSS_BASE_URL")
+        .ok()
+        .filter(|v| !v.trim().is_empty())
+    {
+        Some(url) => url,
+        None => format!(
+            "http://localhost:{port}/v1",
+            port = std::env::var("CODEX_OSS_PORT")
+                .ok()
+                .filter(|v| !v.trim().is_empty())
+                .and_then(|v| v.parse::<u32>().ok())
+                .unwrap_or(DEFAULT_OLLAMA_PORT)
+        ),
+    };
+
+    create_oss_provider_with_base_url(&codex_oss_base_url)
+}
+
+pub fn create_oss_provider_with_base_url(base_url: &str) -> ModelProviderInfo {
+    ModelProviderInfo {
+        name: "gpt-oss".into(),
+        base_url: Some(base_url.into()),
+        env_key: None,
+        env_key_instructions: None,
+        wire_api: WireApi::Chat,
+        query_params: None,
+        http_headers: None,
+        env_http_headers: None,
+        request_max_retries: None,
+        stream_max_retries: None,
+        stream_idle_timeout_ms: None,
+        requires_auth: false,
+    }
+}
+
 #[cfg(test)]
 mod tests {
     #![allow(clippy::unwrap_used)]

codex-rs/core/src/models.rs

@@ -9,7 +9,7 @@ use serde::ser::Serializer;
 
 use crate::protocol::InputItem;
 
-#[derive(Debug, Clone, Serialize, Deserialize)]
+#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
 #[serde(tag = "type", rename_all = "snake_case")]
 pub enum ResponseInputItem {
     Message {
@@ -26,7 +26,7 @@ pub enum ResponseInputItem {
     },
 }
 
-#[derive(Debug, Clone, Serialize, Deserialize)]
+#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
 #[serde(tag = "type", rename_all = "snake_case")]
 pub enum ContentItem {
     InputText { text: String },
@@ -34,7 +34,7 @@ pub enum ContentItem {
     OutputText { text: String },
 }
 
-#[derive(Debug, Clone, Serialize, Deserialize)]
+#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
 #[serde(tag = "type", rename_all = "snake_case")]
 pub enum ResponseItem {
     Message {
@@ -45,6 +45,8 @@ pub enum ResponseItem {
     Reasoning {
         id: String,
         summary: Vec<ReasoningItemReasoningSummary>,
+        #[serde(default, skip_serializing_if = "Option::is_none")]
+        content: Option<Vec<ReasoningItemContent>>,
         encrypted_content: Option<String>,
     },
     LocalShellCall {
@@ -107,7 +109,7 @@ impl From<ResponseInputItem> for ResponseItem {
     }
 }
 
-#[derive(Debug, Clone, Serialize, Deserialize)]
+#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
 #[serde(rename_all = "snake_case")]
 pub enum LocalShellStatus {
     Completed,
@@ -115,13 +117,13 @@ pub enum LocalShellStatus {
     Incomplete,
 }
 
-#[derive(Debug, Clone, Serialize, Deserialize)]
+#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
 #[serde(tag = "type", rename_all = "snake_case")]
 pub enum LocalShellAction {
     Exec(LocalShellExecAction),
 }
 
-#[derive(Debug, Clone, Serialize, Deserialize)]
+#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
 pub struct LocalShellExecAction {
     pub command: Vec<String>,
     pub timeout_ms: Option<u64>,
@@ -130,12 +132,18 @@ pub struct LocalShellExecAction {
     pub user: Option<String>,
 }
 
-#[derive(Debug, Clone, Serialize, Deserialize)]
+#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
 #[serde(tag = "type", rename_all = "snake_case")]
 pub enum ReasoningItemReasoningSummary {
     SummaryText { text: String },
 }
 
+#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
+#[serde(tag = "type", rename_all = "snake_case")]
+pub enum ReasoningItemContent {
+    ReasoningText { text: String },
+}
+
 impl From<Vec<InputItem>> for ResponseInputItem {
     fn from(items: Vec<InputItem>) -> Self {
         Self::Message {
@@ -183,12 +191,15 @@ pub struct ShellToolCallParams {
     // The wire format uses `timeout`, which has ambiguous units, so we use
     // `timeout_ms` as the field name so it is clear in code.
     pub timeout_ms: Option<u64>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub with_escalated_permissions: Option<bool>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub justification: Option<String>,
 }
 
-#[derive(Debug, Clone)]
+#[derive(Debug, Clone, PartialEq)]
 pub struct FunctionCallOutputPayload {
     pub content: String,
-    #[expect(dead_code)]
     pub success: Option<bool>,
 }
 
@@ -295,6 +306,8 @@ mod tests {
                 command: vec!["ls".to_string(), "-l".to_string()],
                 workdir: Some("/tmp".to_string()),
                 timeout_ms: Some(1000),
+                with_escalated_permissions: None,
+                justification: None,
             },
             params
         );

codex-rs/core/src/openai_model_info.rs

@@ -1,3 +1,5 @@
+use crate::model_family::ModelFamily;
+
 /// Metadata about a model, particularly OpenAI models.
 /// We may want to consider including details like the pricing for
 /// input tokens, output tokens, etc., though users will need to be able to
@@ -12,10 +14,19 @@ pub(crate) struct ModelInfo {
     pub(crate) max_output_tokens: u64,
 }
 
-/// Note details such as what a model like gpt-4o is aliased to may be out of
-/// date.
-pub(crate) fn get_model_info(name: &str) -> Option<ModelInfo> {
-    match name {
+pub(crate) fn get_model_info(model_family: &ModelFamily) -> Option<ModelInfo> {
+    match model_family.slug.as_str() {
+        // OSS models have a 128k shared token pool.
+        // Arbitrarily splitting it: 3/4 input context, 1/4 output.
+        // https://openai.com/index/gpt-oss-model-card/
+        "gpt-oss-20b" => Some(ModelInfo {
+            context_window: 96_000,
+            max_output_tokens: 32_000,
+        }),
+        "gpt-oss-120b" => Some(ModelInfo {
+            context_window: 96_000,
+            max_output_tokens: 32_000,
+        }),
         // https://platform.openai.com/docs/models/o3
         "o3" => Some(ModelInfo {
             context_window: 200_000,

codex-rs/core/src/openai_tools.rs

@@ -1,22 +1,28 @@
+use serde::Deserialize;
 use serde::Serialize;
 use serde_json::json;
 use std::collections::BTreeMap;
-use std::sync::LazyLock;
+use std::collections::HashMap;
 
-use crate::client_common::Prompt;
+use crate::model_family::ModelFamily;
 use crate::plan_tool::PLAN_TOOL;
+use crate::protocol::AskForApproval;
+use crate::protocol::SandboxPolicy;
 
-#[derive(Debug, Clone, Serialize)]
-pub(crate) struct ResponsesApiTool {
-    pub(crate) name: &'static str,
-    pub(crate) description: &'static str,
+#[derive(Debug, Clone, Serialize, PartialEq)]
+pub struct ResponsesApiTool {
+    pub(crate) name: String,
+    pub(crate) description: String,
+    /// TODO: Validation. When strict is set to true, the JSON schema,
+    /// `required` and `additional_properties` must be present. All fields in
+    /// `properties` must be present in `required`.
     pub(crate) strict: bool,
     pub(crate) parameters: JsonSchema,
 }
 
 /// When serialized as JSON, this produces a valid "Tool" in the OpenAI
 /// Responses API.
-#[derive(Debug, Clone, Serialize)]
+#[derive(Debug, Clone, Serialize, PartialEq)]
 #[serde(tag = "type")]
 pub(crate) enum OpenAiTool {
     #[serde(rename = "function")]
@@ -25,78 +31,217 @@ pub(crate) enum OpenAiTool {
     LocalShell {},
 }
 
+#[derive(Debug, Clone)]
+pub enum ConfigShellToolType {
+    DefaultShell,
+    ShellWithRequest { sandbox_policy: SandboxPolicy },
+    LocalShell,
+}
+
+#[derive(Debug, Clone)]
+pub struct ToolsConfig {
+    pub shell_type: ConfigShellToolType,
+    pub plan_tool: bool,
+}
+
+impl ToolsConfig {
+    pub fn new(
+        model_family: &ModelFamily,
+        approval_policy: AskForApproval,
+        sandbox_policy: SandboxPolicy,
+        include_plan_tool: bool,
+    ) -> Self {
+        let mut shell_type = if model_family.uses_local_shell_tool {
+            ConfigShellToolType::LocalShell
+        } else {
+            ConfigShellToolType::DefaultShell
+        };
+        if matches!(approval_policy, AskForApproval::OnRequest) {
+            shell_type = ConfigShellToolType::ShellWithRequest {
+                sandbox_policy: sandbox_policy.clone(),
+            }
+        }
+
+        Self {
+            shell_type,
+            plan_tool: include_plan_tool,
+        }
+    }
+}
+
 /// Generic JSON‑Schema subset needed for our tool definitions
-#[derive(Debug, Clone, Serialize)]
+#[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
 #[serde(tag = "type", rename_all = "lowercase")]
 pub(crate) enum JsonSchema {
-    String,
-    Number,
+    Boolean {
+        #[serde(skip_serializing_if = "Option::is_none")]
+        description: Option<String>,
+    },
+    String {
+        #[serde(skip_serializing_if = "Option::is_none")]
+        description: Option<String>,
+    },
+    Number {
+        #[serde(skip_serializing_if = "Option::is_none")]
+        description: Option<String>,
+    },
     Array {
         items: Box<JsonSchema>,
+
+        #[serde(skip_serializing_if = "Option::is_none")]
+        description: Option<String>,
     },
     Object {
         properties: BTreeMap<String, JsonSchema>,
-        required: &'static [&'static str],
-        #[serde(rename = "additionalProperties")]
-        additional_properties: bool,
+        #[serde(skip_serializing_if = "Option::is_none")]
+        required: Option<Vec<String>>,
+        #[serde(
+            rename = "additionalProperties",
+            skip_serializing_if = "Option::is_none"
+        )]
+        additional_properties: Option<bool>,
     },
 }
 
-/// Tool usage specification
-static DEFAULT_TOOLS: LazyLock<Vec<OpenAiTool>> = LazyLock::new(|| {
+fn create_shell_tool() -> OpenAiTool {
     let mut properties = BTreeMap::new();
     properties.insert(
         "command".to_string(),
         JsonSchema::Array {
-            items: Box::new(JsonSchema::String),
+            items: Box::new(JsonSchema::String { description: None }),
+            description: None,
         },
     );
-    properties.insert("workdir".to_string(), JsonSchema::String);
-    properties.insert("timeout".to_string(), JsonSchema::Number);
+    properties.insert(
+        "workdir".to_string(),
+        JsonSchema::String { description: None },
+    );
+    properties.insert(
+        "timeout".to_string(),
+        JsonSchema::Number { description: None },
+    );
 
-    vec![OpenAiTool::Function(ResponsesApiTool {
-        name: "shell",
-        description: "Runs a shell command, and returns its output.",
+    OpenAiTool::Function(ResponsesApiTool {
+        name: "shell".to_string(),
+        description: "Runs a shell command and returns its output".to_string(),
         strict: false,
         parameters: JsonSchema::Object {
             properties,
-            required: &["command"],
-            additional_properties: false,
+            required: Some(vec!["command".to_string()]),
+            additional_properties: Some(false),
         },
-    })]
-});
+    })
+}
 
-static DEFAULT_CODEX_MODEL_TOOLS: LazyLock<Vec<OpenAiTool>> =
-    LazyLock::new(|| vec![OpenAiTool::LocalShell {}]);
+fn create_shell_tool_for_sandbox(sandbox_policy: &SandboxPolicy) -> OpenAiTool {
+    let mut properties = BTreeMap::new();
+    properties.insert(
+        "command".to_string(),
+        JsonSchema::Array {
+            items: Box::new(JsonSchema::String { description: None }),
+            description: Some("The command to execute".to_string()),
+        },
+    );
+    properties.insert(
+        "workdir".to_string(),
+        JsonSchema::String {
+            description: Some("The working directory to execute the command in".to_string()),
+        },
+    );
+    properties.insert(
+        "timeout".to_string(),
+        JsonSchema::Number {
+            description: Some("The timeout for the command in milliseconds".to_string()),
+        },
+    );
+
+    if matches!(sandbox_policy, SandboxPolicy::WorkspaceWrite { .. }) {
+        properties.insert(
+        "with_escalated_permissions".to_string(),
+        JsonSchema::Boolean {
+            description: Some("Whether to request escalated permissions. Set to true if command needs to be run without sandbox restrictions".to_string()),
+        },
+    );
+        properties.insert(
+        "justification".to_string(),
+        JsonSchema::String {
+            description: Some("Only set if ask_for_escalated_permissions is true. 1-sentence explanation of why we want to run this command.".to_string()),
+        },
+    );
+    }
+
+    let description = match sandbox_policy {
+        SandboxPolicy::WorkspaceWrite {
+            network_access,
+            ..
+        } => {
+            format!(
+                r#"
+The shell tool is used to execute shell commands.
+- When invoking the shell tool, your call will be running in a landlock sandbox, and some shell commands will require escalated privileges:
+  - Types of actions that require escalated privileges:
+    - Reading files outside the current directory
+    - Writing files outside the current directory, and protected folders like .git or .env{}
+  - Examples of commands that require escalated privileges:
+    - git commit
+    - npm install or pnpm install
+    - cargo build
+    - cargo test
+- When invoking a command that will require escalated privileges:
+  - Provide the with_escalated_permissions parameter with the boolean value true
+  - Include a short, 1 sentence explanation for why we need to run with_escalated_permissions in the justification parameter."#,
+                if !network_access {
+                    "\n  - Commands that require network access\n"
+                } else {
+                    ""
+                }
+            )
+        }
+        SandboxPolicy::DangerFullAccess => {
+            "Runs a shell command and returns its output.".to_string()
+        }
+        SandboxPolicy::ReadOnly => {
+            r#"
+The shell tool is used to execute shell commands.
+- When invoking the shell tool, your call will be running in a landlock sandbox, and some shell commands (including apply_patch) will require escalated permissions:
+  - Types of actions that require escalated privileges:
+    - Reading files outside the current directory
+    - Writing files
+    - Applying patches
+  - Examples of commands that require escalated privileges:
+    - apply_patch
+    - git commit
+    - npm install or pnpm install
+    - cargo build
+    - cargo test
+- When invoking a command that will require escalated privileges:
+  - Provide the with_escalated_permissions parameter with the boolean value true
+  - Include a short, 1 sentence explanation for why we need to run with_escalated_permissions in the justification parameter"#.to_string()
+        }
+    };
+
+    OpenAiTool::Function(ResponsesApiTool {
+        name: "shell".to_string(),
+        description,
+        strict: false,
+        parameters: JsonSchema::Object {
+            properties,
+            required: Some(vec!["command".to_string()]),
+            additional_properties: Some(false),
+        },
+    })
+}
 
 /// Returns JSON values that are compatible with Function Calling in the
 /// Responses API:
 /// https://platform.openai.com/docs/guides/function-calling?api-mode=responses
 pub(crate) fn create_tools_json_for_responses_api(
-    prompt: &Prompt,
-    model: &str,
-    include_plan_tool: bool,
+    tools: &Vec<OpenAiTool>,
 ) -> crate::error::Result<Vec<serde_json::Value>> {
-    // Assemble tool list: built-in tools + any extra tools from the prompt.
-    let default_tools = if model.starts_with("codex") {
-        &DEFAULT_CODEX_MODEL_TOOLS
-    } else {
-        &DEFAULT_TOOLS
-    };
-    let mut tools_json = Vec::with_capacity(default_tools.len() + prompt.extra_tools.len());
-    for t in default_tools.iter() {
-        tools_json.push(serde_json::to_value(t)?);
-    }
-    tools_json.extend(
-        prompt
-            .extra_tools
-            .clone()
-            .into_iter()
-            .map(|(name, tool)| mcp_tool_to_openai_tool(name, tool)),
-    );
+    let mut tools_json = Vec::new();
 
-    if include_plan_tool {
-        tools_json.push(serde_json::to_value(PLAN_TOOL.clone())?);
+    for tool in tools {
+        tools_json.push(serde_json::to_value(tool)?);
     }
 
     Ok(tools_json)
@@ -106,14 +251,11 @@ pub(crate) fn create_tools_json_for_responses_api(
 /// Chat Completions API:
 /// https://platform.openai.com/docs/guides/function-calling?api-mode=chat
 pub(crate) fn create_tools_json_for_chat_completions_api(
-    prompt: &Prompt,
-    model: &str,
-    include_plan_tool: bool,
+    tools: &Vec<OpenAiTool>,
 ) -> crate::error::Result<Vec<serde_json::Value>> {
     // We start with the JSON for the Responses API and than rewrite it to match
     // the chat completions tool call format.
-    let responses_api_tools_json =
-        create_tools_json_for_responses_api(prompt, model, include_plan_tool)?;
+    let responses_api_tools_json = create_tools_json_for_responses_api(tools)?;
     let tools_json = responses_api_tools_json
         .into_iter()
         .filter_map(|mut tool| {
@@ -136,10 +278,10 @@ pub(crate) fn create_tools_json_for_chat_completions_api(
     Ok(tools_json)
 }
 
-fn mcp_tool_to_openai_tool(
+pub(crate) fn mcp_tool_to_openai_tool(
     fully_qualified_name: String,
     tool: mcp_types::Tool,
-) -> serde_json::Value {
+) -> Result<ResponsesApiTool, serde_json::Error> {
     let mcp_types::Tool {
         description,
         mut input_schema,
@@ -154,12 +296,205 @@ fn mcp_tool_to_openai_tool(
         input_schema.properties = Some(serde_json::Value::Object(serde_json::Map::new()));
     }
 
-    // TODO(mbolin): Change the contract of this function to return
-    // ResponsesApiTool.
-    json!({
-        "name": fully_qualified_name,
-        "description": description,
-        "parameters": input_schema,
-        "type": "function",
+    let serialized_input_schema = serde_json::to_value(input_schema)?;
+    let input_schema = serde_json::from_value::<JsonSchema>(serialized_input_schema)?;
+
+    Ok(ResponsesApiTool {
+        name: fully_qualified_name,
+        description: description.unwrap_or_default(),
+        strict: false,
+        parameters: input_schema,
     })
 }
+
+/// Returns a list of OpenAiTools based on the provided config and MCP tools.
+/// Note that the keys of mcp_tools should be fully qualified names. See
+/// [`McpConnectionManager`] for more details.
+pub(crate) fn get_openai_tools(
+    config: &ToolsConfig,
+    mcp_tools: Option<HashMap<String, mcp_types::Tool>>,
+) -> Vec<OpenAiTool> {
+    let mut tools: Vec<OpenAiTool> = Vec::new();
+
+    match &config.shell_type {
+        ConfigShellToolType::DefaultShell => {
+            tools.push(create_shell_tool());
+        }
+        ConfigShellToolType::ShellWithRequest { sandbox_policy } => {
+            tools.push(create_shell_tool_for_sandbox(sandbox_policy));
+        }
+        ConfigShellToolType::LocalShell => {
+            tools.push(OpenAiTool::LocalShell {});
+        }
+    }
+
+    if config.plan_tool {
+        tools.push(PLAN_TOOL.clone());
+    }
+
+    if let Some(mcp_tools) = mcp_tools {
+        for (name, tool) in mcp_tools {
+            match mcp_tool_to_openai_tool(name.clone(), tool.clone()) {
+                Ok(converted_tool) => tools.push(OpenAiTool::Function(converted_tool)),
+                Err(e) => {
+                    tracing::error!("Failed to convert {name:?} MCP tool to OpenAI tool: {e:?}");
+                }
+            }
+        }
+    }
+
+    tools
+}
+
+#[cfg(test)]
+#[allow(clippy::expect_used)]
+mod tests {
+    use crate::model_family::find_family_for_model;
+    use mcp_types::ToolInputSchema;
+
+    use super::*;
+
+    fn assert_eq_tool_names(tools: &[OpenAiTool], expected_names: &[&str]) {
+        let tool_names = tools
+            .iter()
+            .map(|tool| match tool {
+                OpenAiTool::Function(ResponsesApiTool { name, .. }) => name,
+                OpenAiTool::LocalShell {} => "local_shell",
+            })
+            .collect::<Vec<_>>();
+
+        assert_eq!(
+            tool_names.len(),
+            expected_names.len(),
+            "tool_name mismatch, {tool_names:?}, {expected_names:?}",
+        );
+        for (name, expected_name) in tool_names.iter().zip(expected_names.iter()) {
+            assert_eq!(
+                name, expected_name,
+                "tool_name mismatch, {name:?}, {expected_name:?}"
+            );
+        }
+    }
+
+    #[test]
+    fn test_get_openai_tools() {
+        let model_family = find_family_for_model("codex-mini-latest")
+            .expect("codex-mini-latest should be a valid model family");
+        let config = ToolsConfig::new(
+            &model_family,
+            AskForApproval::Never,
+            SandboxPolicy::ReadOnly,
+            true,
+        );
+        let tools = get_openai_tools(&config, Some(HashMap::new()));
+
+        assert_eq_tool_names(&tools, &["local_shell", "update_plan"]);
+    }
+
+    #[test]
+    fn test_get_openai_tools_default_shell() {
+        let model_family = find_family_for_model("o3").expect("o3 should be a valid model family");
+        let config = ToolsConfig::new(
+            &model_family,
+            AskForApproval::Never,
+            SandboxPolicy::ReadOnly,
+            true,
+        );
+        let tools = get_openai_tools(&config, Some(HashMap::new()));
+
+        assert_eq_tool_names(&tools, &["shell", "update_plan"]);
+    }
+
+    #[test]
+    fn test_get_openai_tools_mcp_tools() {
+        let model_family = find_family_for_model("o3").expect("o3 should be a valid model family");
+        let config = ToolsConfig::new(
+            &model_family,
+            AskForApproval::Never,
+            SandboxPolicy::ReadOnly,
+            false,
+        );
+        let tools = get_openai_tools(
+            &config,
+            Some(HashMap::from([(
+                "test_server/do_something_cool".to_string(),
+                mcp_types::Tool {
+                    name: "do_something_cool".to_string(),
+                    input_schema: ToolInputSchema {
+                        properties: Some(serde_json::json!({
+                            "string_argument": {
+                                "type": "string",
+                            },
+                            "number_argument": {
+                                "type": "number",
+                            },
+                            "object_argument": {
+                                "type": "object",
+                                "properties": {
+                                    "string_property": { "type": "string" },
+                                    "number_property": { "type": "number" },
+                                },
+                                "required": [
+                                    "string_property",
+                                    "number_property"
+                                ],
+                                "additionalProperties": Some(false),
+                            },
+                        })),
+                        required: None,
+                        r#type: "object".to_string(),
+                    },
+                    output_schema: None,
+                    title: None,
+                    annotations: None,
+                    description: Some("Do something cool".to_string()),
+                },
+            )])),
+        );
+
+        assert_eq_tool_names(&tools, &["shell", "test_server/do_something_cool"]);
+
+        assert_eq!(
+            tools[1],
+            OpenAiTool::Function(ResponsesApiTool {
+                name: "test_server/do_something_cool".to_string(),
+                parameters: JsonSchema::Object {
+                    properties: BTreeMap::from([
+                        (
+                            "string_argument".to_string(),
+                            JsonSchema::String { description: None }
+                        ),
+                        (
+                            "number_argument".to_string(),
+                            JsonSchema::Number { description: None }
+                        ),
+                        (
+                            "object_argument".to_string(),
+                            JsonSchema::Object {
+                                properties: BTreeMap::from([
+                                    (
+                                        "string_property".to_string(),
+                                        JsonSchema::String { description: None }
+                                    ),
+                                    (
+                                        "number_property".to_string(),
+                                        JsonSchema::Number { description: None }
+                                    ),
+                                ]),
+                                required: Some(vec![
+                                    "string_property".to_string(),
+                                    "number_property".to_string(),
+                                ]),
+                                additional_properties: Some(false),
+                            },
+                        ),
+                    ]),
+                    required: None,
+                    additional_properties: None,
+                },
+                description: "Do something cool".to_string(),
+                strict: false,
+            })
+        );
+    }
+}

codex-rs/core/src/plan_tool.rs

@@ -39,23 +39,30 @@ pub struct UpdatePlanArgs {
 
 pub(crate) static PLAN_TOOL: LazyLock<OpenAiTool> = LazyLock::new(|| {
     let mut plan_item_props = BTreeMap::new();
-    plan_item_props.insert("step".to_string(), JsonSchema::String);
-    plan_item_props.insert("status".to_string(), JsonSchema::String);
+    plan_item_props.insert("step".to_string(), JsonSchema::String { description: None });
+    plan_item_props.insert(
+        "status".to_string(),
+        JsonSchema::String { description: None },
+    );
 
     let plan_items_schema = JsonSchema::Array {
+        description: Some("The list of steps".to_string()),
         items: Box::new(JsonSchema::Object {
             properties: plan_item_props,
-            required: &["step", "status"],
-            additional_properties: false,
+            required: Some(vec!["step".to_string(), "status".to_string()]),
+            additional_properties: Some(false),
         }),
     };
 
     let mut properties = BTreeMap::new();
-    properties.insert("explanation".to_string(), JsonSchema::String);
+    properties.insert(
+        "explanation".to_string(),
+        JsonSchema::String { description: None },
+    );
     properties.insert("plan".to_string(), plan_items_schema);
 
     OpenAiTool::Function(ResponsesApiTool {
-        name: "update_plan",
+        name: "update_plan".to_string(),
         description: r#"Use the update_plan tool to keep the user updated on the current plan for the task.
 After understanding the user's task, call the update_plan tool with an initial plan. An example of a plan:
 1. Explore the codebase to find relevant files (status: in_progress)
@@ -66,12 +73,12 @@ Until all the steps are finished, there should always be exactly one in_progress
 Call the update_plan tool whenever you finish a step, marking the completed step as `completed` and marking the next step as `in_progress`.
 Before running a command, consider whether or not you have completed the previous step, and make sure to mark it as completed before moving on to the next step.
 Sometimes, you may need to change plans in the middle of a task: call `update_plan` with the updated plan and make sure to provide an `explanation` of the rationale when doing so.
-When all steps are completed, call update_plan one last time with all steps marked as `completed`."#,
+When all steps are completed, call update_plan one last time with all steps marked as `completed`."#.to_string(),
         strict: false,
         parameters: JsonSchema::Object {
             properties,
-            required: &["plan"],
-            additional_properties: false,
+            required: Some(vec!["plan".to_string()]),
+            additional_properties: Some(false),
         },
     })
 });

codex-rs/core/src/protocol.rs

@@ -150,13 +150,17 @@ pub enum AskForApproval {
     /// the user to approve execution without a sandbox.
     OnFailure,
 
+    /// The model decides when to ask the user for approval.
+    OnRequest,
+
     /// Never ask the user to approve commands. Failures are immediately returned
     /// to the model, and never escalated to the user for approval.
     Never,
 }
 
 /// Determines execution restrictions for model shell commands.
-#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
+#[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, Display)]
+#[strum(serialize_all = "kebab-case")]
 #[serde(tag = "mode", rename_all = "kebab-case")]
 pub enum SandboxPolicy {
     /// No restrictions whatsoever. Use with caution.
@@ -359,6 +363,12 @@ pub enum EventMsg {
     /// Agent reasoning delta event from agent.
     AgentReasoningDelta(AgentReasoningDeltaEvent),
 
+    /// Raw chain-of-thought from agent.
+    AgentReasoningRawContent(AgentReasoningRawContentEvent),
+
+    /// Agent reasoning content delta event from agent.
+    AgentReasoningRawContentDelta(AgentReasoningRawContentDeltaEvent),
+
     /// Ack the client's configure message.
     SessionConfigured(SessionConfiguredEvent),
 
@@ -464,6 +474,16 @@ pub struct AgentReasoningEvent {
     pub text: String,
 }
 
+#[derive(Debug, Clone, Deserialize, Serialize)]
+pub struct AgentReasoningRawContentEvent {
+    pub text: String,
+}
+
+#[derive(Debug, Clone, Deserialize, Serialize)]
+pub struct AgentReasoningRawContentDeltaEvent {
+    pub delta: String,
+}
+
 #[derive(Debug, Clone, Deserialize, Serialize)]
 pub struct AgentReasoningDeltaEvent {
     pub delta: String,

codex-rs/core/src/safety.rs

@@ -11,7 +11,7 @@ use crate::is_safe_command::is_known_safe_command;
 use crate::protocol::AskForApproval;
 use crate::protocol::SandboxPolicy;
 
-#[derive(Debug)]
+#[derive(Debug, PartialEq)]
 pub enum SafetyCheck {
     AutoApprove { sandbox_type: SandboxType },
     AskUser,
@@ -31,7 +31,7 @@ pub fn assess_patch_safety(
     }
 
     match policy {
-        AskForApproval::OnFailure | AskForApproval::Never => {
+        AskForApproval::OnFailure | AskForApproval::Never | AskForApproval::OnRequest => {
             // Continue to see if this can be auto-approved.
         }
         // TODO(ragona): I'm not sure this is actually correct? I believe in this case
@@ -76,6 +76,7 @@ pub fn assess_command_safety(
     approval_policy: AskForApproval,
     sandbox_policy: &SandboxPolicy,
     approved: &HashSet<Vec<String>>,
+    with_escalated_permissions: bool,
 ) -> SafetyCheck {
     // A command is "trusted" because either:
     // - it belongs to a set of commands we consider "safe" by default, or
@@ -96,12 +97,13 @@ pub fn assess_command_safety(
         };
     }
 
-    assess_safety_for_untrusted_command(approval_policy, sandbox_policy)
+    assess_safety_for_untrusted_command(approval_policy, sandbox_policy, with_escalated_permissions)
 }
 
 pub(crate) fn assess_safety_for_untrusted_command(
     approval_policy: AskForApproval,
     sandbox_policy: &SandboxPolicy,
+    with_escalated_permissions: bool,
 ) -> SafetyCheck {
     use AskForApproval::*;
     use SandboxPolicy::*;
@@ -113,9 +115,23 @@ pub(crate) fn assess_safety_for_untrusted_command(
             // commands.
             SafetyCheck::AskUser
         }
-        (OnFailure, DangerFullAccess) | (Never, DangerFullAccess) => SafetyCheck::AutoApprove {
+        (OnFailure, DangerFullAccess)
+        | (Never, DangerFullAccess)
+        | (OnRequest, DangerFullAccess) => SafetyCheck::AutoApprove {
             sandbox_type: SandboxType::None,
         },
+        (OnRequest, ReadOnly) | (OnRequest, WorkspaceWrite { .. }) => {
+            if with_escalated_permissions {
+                SafetyCheck::AskUser
+            } else {
+                match get_platform_sandbox() {
+                    Some(sandbox_type) => SafetyCheck::AutoApprove { sandbox_type },
+                    // Fall back to asking since the command is untrusted and
+                    // we do not have a sandbox available
+                    None => SafetyCheck::AskUser,
+                }
+            }
+        }
         (Never, ReadOnly)
         | (Never, WorkspaceWrite { .. })
         | (OnFailure, ReadOnly)
@@ -264,4 +280,47 @@ mod tests {
             &cwd,
         ))
     }
+
+    #[test]
+    fn test_request_escalated_privileges() {
+        // Should not be a trusted command
+        let command = vec!["git commit".to_string()];
+        let approval_policy = AskForApproval::OnRequest;
+        let sandbox_policy = SandboxPolicy::ReadOnly;
+        let approved: HashSet<Vec<String>> = HashSet::new();
+        let request_escalated_privileges = true;
+
+        let safety_check = assess_command_safety(
+            &command,
+            approval_policy,
+            &sandbox_policy,
+            &approved,
+            request_escalated_privileges,
+        );
+
+        assert_eq!(safety_check, SafetyCheck::AskUser);
+    }
+
+    #[test]
+    fn test_request_escalated_privileges_no_sandbox_fallback() {
+        let command = vec!["git".to_string(), "commit".to_string()];
+        let approval_policy = AskForApproval::OnRequest;
+        let sandbox_policy = SandboxPolicy::ReadOnly;
+        let approved: HashSet<Vec<String>> = HashSet::new();
+        let request_escalated_privileges = false;
+
+        let safety_check = assess_command_safety(
+            &command,
+            approval_policy,
+            &sandbox_policy,
+            &approved,
+            request_escalated_privileges,
+        );
+
+        let expected = match get_platform_sandbox() {
+            Some(sandbox_type) => SafetyCheck::AutoApprove { sandbox_type },
+            None => SafetyCheck::AskUser,
+        };
+        assert_eq!(safety_check, expected);
+    }
 }

codex-rs/core/src/shell.rs

@@ -215,6 +215,8 @@ mod tests {
                         "HOME".to_string(),
                         temp_home.path().to_str().unwrap().to_string(),
                     )]),
+                    with_escalated_permissions: None,
+                    justification: None,
                 },
                 SandboxType::None,
                 Arc::new(Notify::new()),

codex-rs/core/tests/client.rs

@@ -1,9 +1,12 @@
+#![allow(clippy::expect_used)]
+#![allow(clippy::unwrap_used)]
 use std::path::PathBuf;
 
 use chrono::Utc;
 use codex_core::Codex;
 use codex_core::CodexSpawnOk;
 use codex_core::ModelProviderInfo;
+use codex_core::WireApi;
 use codex_core::built_in_model_providers;
 use codex_core::protocol::EventMsg;
 use codex_core::protocol::InputItem;
@@ -21,14 +24,42 @@ use tempfile::TempDir;
 use wiremock::Mock;
 use wiremock::MockServer;
 use wiremock::ResponseTemplate;
+use wiremock::matchers::header_regex;
 use wiremock::matchers::method;
 use wiremock::matchers::path;
+use wiremock::matchers::query_param;
 
 /// Build minimal SSE stream with completed marker using the JSON fixture.
 fn sse_completed(id: &str) -> String {
     load_sse_fixture_with_id("tests/fixtures/completed_template.json", id)
 }
 
+fn assert_message_role(request_body: &serde_json::Value, role: &str) {
+    assert_eq!(request_body["role"].as_str().unwrap(), role);
+}
+
+fn assert_message_starts_with(request_body: &serde_json::Value, text: &str) {
+    let content = request_body["content"][0]["text"]
+        .as_str()
+        .expect("invalid message content");
+
+    assert!(
+        content.starts_with(text),
+        "expected message content '{content}' to start with '{text}'"
+    );
+}
+
+fn assert_message_ends_with(request_body: &serde_json::Value, text: &str) {
+    let content = request_body["content"][0]["text"]
+        .as_str()
+        .expect("invalid message content");
+
+    assert!(
+        content.ends_with(text),
+        "expected message content '{content}' to end with '{text}'"
+    );
+}
+
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn includes_session_id_and_model_headers_in_request() {
     #![allow(clippy::unwrap_used)]
@@ -368,14 +399,88 @@ async fn includes_user_instructions_message_in_request() {
             .unwrap()
             .contains("be nice")
     );
-    assert_eq!(request_body["input"][0]["role"], "user");
-    assert!(
-        request_body["input"][0]["content"][0]["text"]
-            .as_str()
-            .unwrap()
-            .starts_with("be nice")
-    );
+    assert_message_role(&request_body["input"][0], "user");
+    assert_message_starts_with(&request_body["input"][0], "<environment_context>\n\n");
+    assert_message_ends_with(&request_body["input"][0], "</environment_context>");
+    assert_message_role(&request_body["input"][1], "user");
+    assert_message_starts_with(&request_body["input"][1], "<user_instructions>\n\n");
+    assert_message_ends_with(&request_body["input"][1], "</user_instructions>");
 }
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn azure_overrides_assign_properties_used_for_responses_url() {
+    #![allow(clippy::unwrap_used)]
+
+    let existing_env_var_with_random_value = if cfg!(windows) { "USERNAME" } else { "USER" };
+
+    // Mock server
+    let server = MockServer::start().await;
+
+    // First request – must NOT include `previous_response_id`.
+    let first = ResponseTemplate::new(200)
+        .insert_header("content-type", "text/event-stream")
+        .set_body_raw(sse_completed("resp1"), "text/event-stream");
+
+    // Expect POST to /openai/responses with api-version query param
+    Mock::given(method("POST"))
+        .and(path("/openai/responses"))
+        .and(query_param("api-version", "2025-04-01-preview"))
+        .and(header_regex("Custom-Header", "Value"))
+        .and(header_regex(
+            "Authorization",
+            format!(
+                "Bearer {}",
+                std::env::var(existing_env_var_with_random_value).unwrap()
+            )
+            .as_str(),
+        ))
+        .respond_with(first)
+        .expect(1)
+        .mount(&server)
+        .await;
+
+    let provider = ModelProviderInfo {
+        name: "custom".to_string(),
+        base_url: Some(format!("{}/openai", server.uri())),
+        // Reuse the existing environment variable to avoid using unsafe code
+        env_key: Some(existing_env_var_with_random_value.to_string()),
+        query_params: Some(std::collections::HashMap::from([(
+            "api-version".to_string(),
+            "2025-04-01-preview".to_string(),
+        )])),
+        env_key_instructions: None,
+        wire_api: WireApi::Responses,
+        http_headers: Some(std::collections::HashMap::from([(
+            "Custom-Header".to_string(),
+            "Value".to_string(),
+        )])),
+        env_http_headers: None,
+        request_max_retries: None,
+        stream_max_retries: None,
+        stream_idle_timeout_ms: None,
+        requires_auth: false,
+    };
+
+    // Init session
+    let codex_home = TempDir::new().unwrap();
+    let mut config = load_default_config_for_test(&codex_home);
+    config.model_provider = provider;
+
+    let ctrl_c = std::sync::Arc::new(tokio::sync::Notify::new());
+    let CodexSpawnOk { codex, .. } = Codex::spawn(config, None, ctrl_c.clone()).await.unwrap();
+
+    codex
+        .submit(Op::UserInput {
+            items: vec![InputItem::Text {
+                text: "hello".into(),
+            }],
+        })
+        .await
+        .unwrap();
+
+    wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
+}
+
 fn auth_from_token(id_token: String) -> CodexAuth {
     CodexAuth::new(
         None,

codex-rs/core/tests/exec.rs

@@ -28,6 +28,8 @@ async fn run_test_cmd(tmp: TempDir, cmd: Vec<&str>, should_be_ok: bool) {
         cwd: tmp.path().to_path_buf(),
         timeout_ms: Some(1000),
         env: HashMap::new(),
+        with_escalated_permissions: None,
+        justification: None,
     };
 
     let ctrl_c = Arc::new(Notify::new());

codex-rs/core/tests/exec_stream_events.rs

@@ -53,6 +53,8 @@ async fn test_exec_stdout_stream_events_echo() {
         cwd: std::env::current_dir().unwrap_or_else(|_| PathBuf::from(".")),
         timeout_ms: Some(5_000),
         env: HashMap::new(),
+        with_escalated_permissions: None,
+        justification: None,
     };
 
     let ctrl_c = Arc::new(Notify::new());
@@ -103,6 +105,8 @@ async fn test_exec_stderr_stream_events_echo() {
         cwd: std::env::current_dir().unwrap_or_else(|_| PathBuf::from(".")),
         timeout_ms: Some(5_000),
         env: HashMap::new(),
+        with_escalated_permissions: None,
+        justification: None,
     };
 
     let ctrl_c = Arc::new(Notify::new());

codex-rs/exec/Cargo.toml

@@ -25,6 +25,7 @@ codex-common = { path = "../common", features = [
     "sandbox_summary",
 ] }
 codex-core = { path = "../core" }
+codex-ollama = { path = "../ollama" }
 owo-colors = "4.2.0"
 serde_json = "1"
 shlex = "1.3.0"

codex-rs/exec/src/cli.rs

@@ -14,6 +14,9 @@ pub struct Cli {
     #[arg(long, short = 'm')]
     pub model: Option<String>,
 
+    #[arg(long = "oss", default_value_t = false)]
+    pub oss: bool,
+
     /// Select the sandbox policy to use when executing model-generated shell
     /// commands.
     #[arg(long = "sandbox", short = 's')]

codex-rs/exec/src/event_processor.rs

@@ -1,9 +1,6 @@
 use std::path::Path;
 
-use codex_common::summarize_sandbox_policy;
-use codex_core::WireApi;
 use codex_core::config::Config;
-use codex_core::model_supports_reasoning_summaries;
 use codex_core::protocol::Event;
 
 pub(crate) enum CodexStatus {
@@ -20,44 +17,14 @@ pub(crate) trait EventProcessor {
     fn process_event(&mut self, event: Event) -> CodexStatus;
 }
 
-pub(crate) fn create_config_summary_entries(config: &Config) -> Vec<(&'static str, String)> {
-    let mut entries = vec![
-        ("workdir", config.cwd.display().to_string()),
-        ("model", config.model.clone()),
-        ("provider", config.model_provider_id.clone()),
-        ("approval", config.approval_policy.to_string()),
-        ("sandbox", summarize_sandbox_policy(&config.sandbox_policy)),
-    ];
-    if config.model_provider.wire_api == WireApi::Responses
-        && model_supports_reasoning_summaries(config)
-    {
-        entries.push((
-            "reasoning effort",
-            config.model_reasoning_effort.to_string(),
-        ));
-        entries.push((
-            "reasoning summaries",
-            config.model_reasoning_summary.to_string(),
-        ));
-    }
-
-    entries
-}
-
-pub(crate) fn handle_last_message(
-    last_agent_message: Option<&str>,
-    last_message_path: Option<&Path>,
-) {
-    match (last_message_path, last_agent_message) {
-        (Some(path), Some(msg)) => write_last_message_file(msg, Some(path)),
-        (Some(path), None) => {
-            write_last_message_file("", Some(path));
-            eprintln!(
-                "Warning: no last agent message; wrote empty content to {}",
-                path.display()
-            );
-        }
-        (None, _) => eprintln!("Warning: no file to write last message to."),
+pub(crate) fn handle_last_message(last_agent_message: Option<&str>, output_file: &Path) {
+    let message = last_agent_message.unwrap_or_default();
+    write_last_message_file(message, Some(output_file));
+    if last_agent_message.is_none() {
+        eprintln!(
+            "Warning: no last agent message; wrote empty content to {}",
+            output_file.display()
+        );
     }
 }
 

codex-rs/exec/src/event_processor_with_human_output.rs

@@ -5,6 +5,8 @@ use codex_core::plan_tool::UpdatePlanArgs;
 use codex_core::protocol::AgentMessageDeltaEvent;
 use codex_core::protocol::AgentMessageEvent;
 use codex_core::protocol::AgentReasoningDeltaEvent;
+use codex_core::protocol::AgentReasoningRawContentDeltaEvent;
+use codex_core::protocol::AgentReasoningRawContentEvent;
 use codex_core::protocol::BackgroundEventEvent;
 use codex_core::protocol::ErrorEvent;
 use codex_core::protocol::Event;
@@ -31,8 +33,8 @@ use std::time::Instant;
 
 use crate::event_processor::CodexStatus;
 use crate::event_processor::EventProcessor;
-use crate::event_processor::create_config_summary_entries;
 use crate::event_processor::handle_last_message;
+use codex_common::create_config_summary_entries;
 
 /// This should be configurable. When used in CI, users may not want to impose
 /// a limit so they can see the full transcript.
@@ -55,8 +57,10 @@ pub(crate) struct EventProcessorWithHumanOutput {
 
     /// Whether to include `AgentReasoning` events in the output.
     show_agent_reasoning: bool,
+    show_raw_agent_reasoning: bool,
     answer_started: bool,
     reasoning_started: bool,
+    raw_reasoning_started: bool,
     last_message_path: Option<PathBuf>,
 }
 
@@ -81,8 +85,10 @@ impl EventProcessorWithHumanOutput {
                 green: Style::new().green(),
                 cyan: Style::new().cyan(),
                 show_agent_reasoning: !config.hide_agent_reasoning,
+                show_raw_agent_reasoning: config.show_raw_agent_reasoning,
                 answer_started: false,
                 reasoning_started: false,
+                raw_reasoning_started: false,
                 last_message_path,
             }
         } else {
@@ -97,8 +103,10 @@ impl EventProcessorWithHumanOutput {
                 green: Style::new(),
                 cyan: Style::new(),
                 show_agent_reasoning: !config.hide_agent_reasoning,
+                show_raw_agent_reasoning: config.show_raw_agent_reasoning,
                 answer_started: false,
                 reasoning_started: false,
+                raw_reasoning_started: false,
                 last_message_path,
             }
         }
@@ -170,10 +178,9 @@ impl EventProcessor for EventProcessorWithHumanOutput {
                 // Ignore.
             }
             EventMsg::TaskComplete(TaskCompleteEvent { last_agent_message }) => {
-                handle_last_message(
-                    last_agent_message.as_deref(),
-                    self.last_message_path.as_deref(),
-                );
+                if let Some(output_file) = self.last_message_path.as_deref() {
+                    handle_last_message(last_agent_message.as_deref(), output_file);
+                }
                 return CodexStatus::InitiateShutdown;
             }
             EventMsg::TokenCount(TokenUsage { total_tokens, .. }) => {
@@ -204,6 +211,32 @@ impl EventProcessor for EventProcessorWithHumanOutput {
                 #[allow(clippy::expect_used)]
                 std::io::stdout().flush().expect("could not flush stdout");
             }
+            EventMsg::AgentReasoningRawContent(AgentReasoningRawContentEvent { text }) => {
+                if !self.show_raw_agent_reasoning {
+                    return CodexStatus::Running;
+                }
+                if !self.raw_reasoning_started {
+                    print!("{text}");
+                    #[allow(clippy::expect_used)]
+                    std::io::stdout().flush().expect("could not flush stdout");
+                } else {
+                    println!();
+                    self.raw_reasoning_started = false;
+                }
+            }
+            EventMsg::AgentReasoningRawContentDelta(AgentReasoningRawContentDeltaEvent {
+                delta,
+            }) => {
+                if !self.show_raw_agent_reasoning {
+                    return CodexStatus::Running;
+                }
+                if !self.raw_reasoning_started {
+                    self.raw_reasoning_started = true;
+                }
+                print!("{delta}");
+                #[allow(clippy::expect_used)]
+                std::io::stdout().flush().expect("could not flush stdout");
+            }
             EventMsg::AgentMessage(AgentMessageEvent { message }) => {
                 // if answer_started is false, this means we haven't received any
                 // delta. Thus, we need to print the message as a new answer.

codex-rs/exec/src/event_processor_with_json_output.rs

@@ -9,8 +9,8 @@ use serde_json::json;
 
 use crate::event_processor::CodexStatus;
 use crate::event_processor::EventProcessor;
-use crate::event_processor::create_config_summary_entries;
 use crate::event_processor::handle_last_message;
+use codex_common::create_config_summary_entries;
 
 pub(crate) struct EventProcessorWithJsonOutput {
     last_message_path: Option<PathBuf>,
@@ -46,10 +46,9 @@ impl EventProcessor for EventProcessorWithJsonOutput {
                 CodexStatus::Running
             }
             EventMsg::TaskComplete(TaskCompleteEvent { last_agent_message }) => {
-                handle_last_message(
-                    last_agent_message.as_deref(),
-                    self.last_message_path.as_deref(),
-                );
+                if let Some(output_file) = self.last_message_path.as_deref() {
+                    handle_last_message(last_agent_message.as_deref(), output_file);
+                }
                 CodexStatus::InitiateShutdown
             }
             EventMsg::ShutdownComplete => CodexStatus::Shutdown,

codex-rs/exec/src/lib.rs

@@ -9,6 +9,7 @@ use std::path::PathBuf;
 use std::sync::Arc;
 
 pub use cli::Cli;
+use codex_core::BUILT_IN_OSS_MODEL_PROVIDER_ID;
 use codex_core::codex_wrapper::CodexConversation;
 use codex_core::codex_wrapper::{self};
 use codex_core::config::Config;
@@ -21,6 +22,7 @@ use codex_core::protocol::InputItem;
 use codex_core::protocol::Op;
 use codex_core::protocol::TaskCompleteEvent;
 use codex_core::util::is_inside_git_repo;
+use codex_ollama::DEFAULT_OSS_MODEL;
 use event_processor_with_human_output::EventProcessorWithHumanOutput;
 use event_processor_with_json_output::EventProcessorWithJsonOutput;
 use tracing::debug;
@@ -34,7 +36,8 @@ use crate::event_processor::EventProcessor;
 pub async fn run_main(cli: Cli, codex_linux_sandbox_exe: Option<PathBuf>) -> anyhow::Result<()> {
     let Cli {
         images,
-        model,
+        model: model_cli_arg,
+        oss,
         config_profile,
         full_auto,
         dangerously_bypass_approvals_and_sandbox,
@@ -114,6 +117,23 @@ pub async fn run_main(cli: Cli, codex_linux_sandbox_exe: Option<PathBuf>) -> any
         sandbox_mode_cli_arg.map(Into::<SandboxMode>::into)
     };
 
+    // When using `--oss`, let the bootstrapper pick the model (defaulting to
+    // gpt-oss:20b) and ensure it is present locally. Also, force the built‑in
+    // `oss` model provider.
+    let model = if let Some(model) = model_cli_arg {
+        Some(model)
+    } else if oss {
+        Some(DEFAULT_OSS_MODEL.to_owned())
+    } else {
+        None // No model specified, will use the default.
+    };
+
+    let model_provider = if oss {
+        Some(BUILT_IN_OSS_MODEL_PROVIDER_ID.to_string())
+    } else {
+        None // No specific model provider override.
+    };
+
     // Load configuration and determine approval policy
     let overrides = ConfigOverrides {
         model,
@@ -123,10 +143,12 @@ pub async fn run_main(cli: Cli, codex_linux_sandbox_exe: Option<PathBuf>) -> any
         approval_policy: Some(AskForApproval::Never),
         sandbox_mode,
         cwd: cwd.map(|p| p.canonicalize().unwrap_or(p)),
-        model_provider: None,
+        model_provider,
         codex_linux_sandbox_exe,
         base_instructions: None,
         include_plan_tool: None,
+        disable_response_storage: oss.then_some(true),
+        show_raw_agent_reasoning: oss.then_some(true),
     };
     // Parse `-c` overrides.
     let cli_kv_overrides = match config_overrides.parse_overrides() {
@@ -148,6 +170,12 @@ pub async fn run_main(cli: Cli, codex_linux_sandbox_exe: Option<PathBuf>) -> any
         ))
     };
 
+    if oss {
+        codex_ollama::ensure_oss_ready(&config)
+            .await
+            .map_err(|e| anyhow::anyhow!("OSS setup failed: {e}"))?;
+    }
+
     // Print the effective configuration and prompt so users can see what Codex
     // is using.
     event_processor.print_config_summary(&config, &prompt);
@@ -188,10 +216,16 @@ pub async fn run_main(cli: Cli, codex_linux_sandbox_exe: Option<PathBuf>) -> any
                     res = codex.next_event() => match res {
                         Ok(event) => {
                             debug!("Received event: {event:?}");
+
+                            let is_shutdown_complete = matches!(event.msg, EventMsg::ShutdownComplete);
                             if let Err(e) = tx.send(event) {
                                 error!("Error sending event: {e:?}");
                                 break;
                             }
+                            if is_shutdown_complete {
+                                info!("Received shutdown event, exiting event loop.");
+                                break;
+                            }
                         },
                         Err(e) => {
                             error!("Error receiving event: {e:?}");

codex-rs/execpolicy/Cargo.toml

@@ -26,7 +26,7 @@ multimap = "0.10.0"
 path-absolutize = "3.1.1"
 regex-lite = "0.1"
 serde = { version = "1.0.194", features = ["derive"] }
-serde_json = "1.0.110"
+serde_json = "1.0.142"
 serde_with = { version = "3", features = ["macros"] }
 
 [dev-dependencies]

codex-rs/file-search/Cargo.toml

@@ -17,5 +17,5 @@ clap = { version = "4", features = ["derive"] }
 ignore = "0.4.23"
 nucleo-matcher = "0.3.1"
 serde = { version = "1", features = ["derive"] }
-serde_json = "1.0.110"
+serde_json = "1.0.142"
 tokio = { version = "1", features = ["full"] }

codex-rs/linux-sandbox/tests/landlock.rs

@@ -44,6 +44,8 @@ async fn run_cmd(cmd: &[&str], writable_roots: &[PathBuf], timeout_ms: u64) {
         cwd: std::env::current_dir().expect("cwd should exist"),
         timeout_ms: Some(timeout_ms),
         env: create_env_from_core_vars(),
+        with_escalated_permissions: None,
+        justification: None,
     };
 
     let sandbox_policy = SandboxPolicy::WorkspaceWrite {
@@ -139,6 +141,8 @@ async fn assert_network_blocked(cmd: &[&str]) {
         // do not stall the suite.
         timeout_ms: Some(NETWORK_TIMEOUT_MS),
         env: create_env_from_core_vars(),
+        with_escalated_permissions: None,
+        justification: None,
     };
 
     let sandbox_policy = SandboxPolicy::new_read_only_policy();

codex-rs/login/src/lib.rs

@@ -22,7 +22,7 @@ const SOURCE_FOR_PYTHON_SERVER: &str = include_str!("./login_with_chatgpt.py");
 const CLIENT_ID: &str = "app_EMoamEEZ73f0CkXaXp7hrann";
 pub const OPENAI_API_KEY_ENV_VAR: &str = "OPENAI_API_KEY";
 
-#[derive(Clone, Debug, PartialEq)]
+#[derive(Clone, Debug, PartialEq, Copy)]
 pub enum AuthMode {
     ApiKey,
     ChatGPT,

codex-rs/login/src/login_with_chatgpt.py

@@ -458,6 +458,7 @@ class _ApiKeyHTTPServer(http.server.HTTPServer):
             "code_challenge": self.pkce.code_challenge,
             "code_challenge_method": "S256",
             "id_token_add_organizations": "true",
+            "codex_cli_simplified_flow": "true",
             "state": self.state,
         }
         return f"{self.issuer}/oauth/authorize?" + urllib.parse.urlencode(params)

codex-rs/mcp-server/src/codex_tool_config.rs

@@ -158,6 +158,8 @@ impl CodexToolCallParam {
             codex_linux_sandbox_exe,
             base_instructions,
             include_plan_tool,
+            disable_response_storage: None,
+            show_raw_agent_reasoning: None,
         };
 
         let cli_overrides = cli_overrides

codex-rs/mcp-server/src/codex_tool_runner.rs

@@ -252,7 +252,9 @@ async fn run_codex_tool_session_inner(
                     EventMsg::AgentMessage(AgentMessageEvent { .. }) => {
                         // TODO: think how we want to support this in the MCP
                     }
-                    EventMsg::TaskStarted
+                    EventMsg::AgentReasoningRawContent(_)
+                    | EventMsg::AgentReasoningRawContentDelta(_)
+                    | EventMsg::TaskStarted
                     | EventMsg::TokenCount(_)
                     | EventMsg::AgentReasoning(_)
                     | EventMsg::McpToolCallBegin(_)

codex-rs/mcp-server/src/conversation_loop.rs

@@ -90,7 +90,9 @@ pub async fn run_conversation_loop(
                     EventMsg::AgentMessage(AgentMessageEvent { .. }) => {
                         // TODO: think how we want to support this in the MCP
                     }
-                    EventMsg::TaskStarted
+                    EventMsg::AgentReasoningRawContent(_)
+                    | EventMsg::AgentReasoningRawContentDelta(_)
+                    | EventMsg::TaskStarted
                     | EventMsg::TokenCount(_)
                     | EventMsg::AgentReasoning(_)
                     | EventMsg::McpToolCallBegin(_)

codex-rs/mcp-server/src/tool_handlers/create_conversation.rs

@@ -59,6 +59,8 @@ pub(crate) async fn handle_create_conversation(
         codex_linux_sandbox_exe: None,
         base_instructions,
         include_plan_tool: None,
+        disable_response_storage: None,
+        show_raw_agent_reasoning: None,
     };
 
     let cfg: CodexConfig = match CodexConfig::load_with_cli_overrides(cli_overrides, overrides) {

codex-rs/mcp-server/tests/send_message.rs

@@ -18,7 +18,7 @@ use tokio::time::timeout;
 
 const DEFAULT_READ_TIMEOUT: std::time::Duration = std::time::Duration::from_secs(10);
 
-#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+#[tokio::test]
 async fn test_send_message_success() {
     // Spin up a mock completions server that immediately ends the Codex turn.
     // Two Codex turns hit the mock model (session start + send-user-message). Provide two SSE responses.
@@ -99,13 +99,13 @@ async fn test_send_message_success() {
         response
     );
     // wait for the server to hear the user message
-    sleep(Duration::from_secs(1));
+    sleep(Duration::from_secs(5));
 
     // Ensure the server and tempdir live until end of test
     drop(server);
 }
 
-#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+#[tokio::test]
 async fn test_send_message_session_not_found() {
     // Start MCP without creating a Codex session
     let codex_home = TempDir::new().expect("tempdir");

codex-rs/ollama/Cargo.toml

@@ -0,0 +1,32 @@
+[package]
+edition = "2024"
+name = "codex-ollama"
+version = { workspace = true }
+
+[lib]
+name = "codex_ollama"
+path = "src/lib.rs"
+
+[lints]
+workspace = true
+
+[dependencies]
+async-stream = "0.3"
+bytes = "1.10.1"
+codex-core = { path = "../core" }
+futures = "0.3"
+reqwest = { version = "0.12", features = ["json", "stream"] }
+serde_json = "1"
+tokio = { version = "1", features = [
+    "io-std",
+    "macros",
+    "process",
+    "rt-multi-thread",
+    "signal",
+] }
+toml = "0.9.2"
+tracing = { version = "0.1.41", features = ["log"] }
+wiremock = "0.6"
+
+[dev-dependencies]
+tempfile = "3"

codex-rs/ollama/src/client.rs

@@ -0,0 +1,351 @@
+use bytes::BytesMut;
+use futures::StreamExt;
+use futures::stream::BoxStream;
+use serde_json::Value as JsonValue;
+use std::collections::VecDeque;
+use std::io;
+
+use crate::parser::pull_events_from_value;
+use crate::pull::PullEvent;
+use crate::pull::PullProgressReporter;
+use crate::url::base_url_to_host_root;
+use crate::url::is_openai_compatible_base_url;
+use codex_core::BUILT_IN_OSS_MODEL_PROVIDER_ID;
+use codex_core::ModelProviderInfo;
+use codex_core::WireApi;
+use codex_core::config::Config;
+
+const OLLAMA_CONNECTION_ERROR: &str = "No running Ollama server detected. Start it with: `ollama serve` (after installing). Install instructions: https://github.com/ollama/ollama?tab=readme-ov-file#ollama";
+
+/// Client for interacting with a local Ollama instance.
+pub struct OllamaClient {
+    client: reqwest::Client,
+    host_root: String,
+    uses_openai_compat: bool,
+}
+
+impl OllamaClient {
+    /// Construct a client for the built‑in open‑source ("oss") model provider
+    /// and verify that a local Ollama server is reachable. If no server is
+    /// detected, returns an error with helpful installation/run instructions.
+    pub async fn try_from_oss_provider(config: &Config) -> io::Result<Self> {
+        // Note that we must look up the provider from the Config to ensure that
+        // any overrides the user has in their config.toml are taken into
+        // account.
+        let provider = config
+            .model_providers
+            .get(BUILT_IN_OSS_MODEL_PROVIDER_ID)
+            .ok_or_else(|| {
+                io::Error::new(
+                    io::ErrorKind::NotFound,
+                    format!("Built-in provider {BUILT_IN_OSS_MODEL_PROVIDER_ID} not found",),
+                )
+            })?;
+
+        Self::try_from_provider(provider).await
+    }
+
+    #[cfg(test)]
+    async fn try_from_provider_with_base_url(base_url: &str) -> io::Result<Self> {
+        let provider = codex_core::create_oss_provider_with_base_url(base_url);
+        Self::try_from_provider(&provider).await
+    }
+
+    /// Build a client from a provider definition and verify the server is reachable.
+    async fn try_from_provider(provider: &ModelProviderInfo) -> io::Result<Self> {
+        #![allow(clippy::expect_used)]
+        let base_url = provider
+            .base_url
+            .as_ref()
+            .expect("oss provider must have a base_url");
+        let uses_openai_compat = is_openai_compatible_base_url(base_url)
+            || matches!(provider.wire_api, WireApi::Chat)
+                && is_openai_compatible_base_url(base_url);
+        let host_root = base_url_to_host_root(base_url);
+        let client = reqwest::Client::builder()
+            .connect_timeout(std::time::Duration::from_secs(5))
+            .build()
+            .unwrap_or_else(|_| reqwest::Client::new());
+        let client = Self {
+            client,
+            host_root,
+            uses_openai_compat,
+        };
+        client.probe_server().await?;
+        Ok(client)
+    }
+
+    /// Probe whether the server is reachable by hitting the appropriate health endpoint.
+    async fn probe_server(&self) -> io::Result<()> {
+        let url = if self.uses_openai_compat {
+            format!("{}/v1/models", self.host_root.trim_end_matches('/'))
+        } else {
+            format!("{}/api/tags", self.host_root.trim_end_matches('/'))
+        };
+        let resp = self.client.get(url).send().await.map_err(|err| {
+            tracing::warn!("Failed to connect to Ollama server: {err:?}");
+            io::Error::other(OLLAMA_CONNECTION_ERROR)
+        })?;
+        if resp.status().is_success() {
+            Ok(())
+        } else {
+            tracing::warn!(
+                "Failed to probe server at {}: HTTP {}",
+                self.host_root,
+                resp.status()
+            );
+            Err(io::Error::other(OLLAMA_CONNECTION_ERROR))
+        }
+    }
+
+    /// Return the list of model names known to the local Ollama instance.
+    pub async fn fetch_models(&self) -> io::Result<Vec<String>> {
+        let tags_url = format!("{}/api/tags", self.host_root.trim_end_matches('/'));
+        let resp = self
+            .client
+            .get(tags_url)
+            .send()
+            .await
+            .map_err(io::Error::other)?;
+        if !resp.status().is_success() {
+            return Ok(Vec::new());
+        }
+        let val = resp.json::<JsonValue>().await.map_err(io::Error::other)?;
+        let names = val
+            .get("models")
+            .and_then(|m| m.as_array())
+            .map(|arr| {
+                arr.iter()
+                    .filter_map(|v| v.get("name").and_then(|n| n.as_str()))
+                    .map(|s| s.to_string())
+                    .collect::<Vec<_>>()
+            })
+            .unwrap_or_default();
+        Ok(names)
+    }
+
+    /// Start a model pull and emit streaming events. The returned stream ends when
+    /// a Success event is observed or the server closes the connection.
+    pub async fn pull_model_stream(
+        &self,
+        model: &str,
+    ) -> io::Result<BoxStream<'static, PullEvent>> {
+        let url = format!("{}/api/pull", self.host_root.trim_end_matches('/'));
+        let resp = self
+            .client
+            .post(url)
+            .json(&serde_json::json!({"model": model, "stream": true}))
+            .send()
+            .await
+            .map_err(io::Error::other)?;
+        if !resp.status().is_success() {
+            return Err(io::Error::other(format!(
+                "failed to start pull: HTTP {}",
+                resp.status()
+            )));
+        }
+
+        let mut stream = resp.bytes_stream();
+        let mut buf = BytesMut::new();
+        let _pending: VecDeque<PullEvent> = VecDeque::new();
+
+        // Using an async stream adaptor backed by unfold-like manual loop.
+        let s = async_stream::stream! {
+            while let Some(chunk) = stream.next().await {
+                match chunk {
+                    Ok(bytes) => {
+                        buf.extend_from_slice(&bytes);
+                        while let Some(pos) = buf.iter().position(|b| *b == b'\n') {
+                            let line = buf.split_to(pos + 1);
+                            if let Ok(text) = std::str::from_utf8(&line) {
+                                let text = text.trim();
+                                if text.is_empty() { continue; }
+                                if let Ok(value) = serde_json::from_str::<JsonValue>(text) {
+                                    for ev in pull_events_from_value(&value) { yield ev; }
+                                    if let Some(err_msg) = value.get("error").and_then(|e| e.as_str()) {
+                                        yield PullEvent::Error(err_msg.to_string());
+                                        return;
+                                    }
+                                    if let Some(status) = value.get("status").and_then(|s| s.as_str()) {
+                                        if status == "success" { yield PullEvent::Success; return; }
+                                    }
+                                }
+                            }
+                        }
+                    }
+                    Err(_) => {
+                        // Connection error: end the stream.
+                        return;
+                    }
+                }
+            }
+        };
+
+        Ok(Box::pin(s))
+    }
+
+    /// High-level helper to pull a model and drive a progress reporter.
+    pub async fn pull_with_reporter(
+        &self,
+        model: &str,
+        reporter: &mut dyn PullProgressReporter,
+    ) -> io::Result<()> {
+        reporter.on_event(&PullEvent::Status(format!("Pulling model {model}...")))?;
+        let mut stream = self.pull_model_stream(model).await?;
+        while let Some(event) = stream.next().await {
+            reporter.on_event(&event)?;
+            match event {
+                PullEvent::Success => {
+                    return Ok(());
+                }
+                PullEvent::Error(err) => {
+                    // Empirically, ollama returns a 200 OK response even when
+                    // the output stream includes an error message. Verify with:
+                    //
+                    // `curl -i http://localhost:11434/api/pull -d '{ "model": "foobarbaz" }'`
+                    //
+                    // As such, we have to check the event stream, not the
+                    // HTTP response status, to determine whether to return Err.
+                    return Err(io::Error::other(format!("Pull failed: {err}")));
+                }
+                PullEvent::ChunkProgress { .. } | PullEvent::Status(_) => {
+                    continue;
+                }
+            }
+        }
+        Err(io::Error::other(
+            "Pull stream ended unexpectedly without success.",
+        ))
+    }
+
+    /// Low-level constructor given a raw host root, e.g. "http://localhost:11434".
+    #[cfg(test)]
+    fn from_host_root(host_root: impl Into<String>) -> Self {
+        let client = reqwest::Client::builder()
+            .connect_timeout(std::time::Duration::from_secs(5))
+            .build()
+            .unwrap_or_else(|_| reqwest::Client::new());
+        Self {
+            client,
+            host_root: host_root.into(),
+            uses_openai_compat: false,
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    #![allow(clippy::expect_used, clippy::unwrap_used)]
+    use super::*;
+
+    // Happy-path tests using a mock HTTP server; skip if sandbox network is disabled.
+    #[tokio::test]
+    async fn test_fetch_models_happy_path() {
+        if std::env::var(codex_core::spawn::CODEX_SANDBOX_NETWORK_DISABLED_ENV_VAR).is_ok() {
+            tracing::info!(
+                "{} is set; skipping test_fetch_models_happy_path",
+                codex_core::spawn::CODEX_SANDBOX_NETWORK_DISABLED_ENV_VAR
+            );
+            return;
+        }
+
+        let server = wiremock::MockServer::start().await;
+        wiremock::Mock::given(wiremock::matchers::method("GET"))
+            .and(wiremock::matchers::path("/api/tags"))
+            .respond_with(
+                wiremock::ResponseTemplate::new(200).set_body_raw(
+                    serde_json::json!({
+                        "models": [ {"name": "llama3.2:3b"}, {"name":"mistral"} ]
+                    })
+                    .to_string(),
+                    "application/json",
+                ),
+            )
+            .mount(&server)
+            .await;
+
+        let client = OllamaClient::from_host_root(server.uri());
+        let models = client.fetch_models().await.expect("fetch models");
+        assert!(models.contains(&"llama3.2:3b".to_string()));
+        assert!(models.contains(&"mistral".to_string()));
+    }
+
+    #[tokio::test]
+    async fn test_probe_server_happy_path_openai_compat_and_native() {
+        if std::env::var(codex_core::spawn::CODEX_SANDBOX_NETWORK_DISABLED_ENV_VAR).is_ok() {
+            tracing::info!(
+                "{} set; skipping test_probe_server_happy_path_openai_compat_and_native",
+                codex_core::spawn::CODEX_SANDBOX_NETWORK_DISABLED_ENV_VAR
+            );
+            return;
+        }
+
+        let server = wiremock::MockServer::start().await;
+
+        // Native endpoint
+        wiremock::Mock::given(wiremock::matchers::method("GET"))
+            .and(wiremock::matchers::path("/api/tags"))
+            .respond_with(wiremock::ResponseTemplate::new(200))
+            .mount(&server)
+            .await;
+        let native = OllamaClient::from_host_root(server.uri());
+        native.probe_server().await.expect("probe native");
+
+        // OpenAI compatibility endpoint
+        wiremock::Mock::given(wiremock::matchers::method("GET"))
+            .and(wiremock::matchers::path("/v1/models"))
+            .respond_with(wiremock::ResponseTemplate::new(200))
+            .mount(&server)
+            .await;
+        let ollama_client =
+            OllamaClient::try_from_provider_with_base_url(&format!("{}/v1", server.uri()))
+                .await
+                .expect("probe OpenAI compat");
+        ollama_client
+            .probe_server()
+            .await
+            .expect("probe OpenAI compat");
+    }
+
+    #[tokio::test]
+    async fn test_try_from_oss_provider_ok_when_server_running() {
+        if std::env::var(codex_core::spawn::CODEX_SANDBOX_NETWORK_DISABLED_ENV_VAR).is_ok() {
+            tracing::info!(
+                "{} set; skipping test_try_from_oss_provider_ok_when_server_running",
+                codex_core::spawn::CODEX_SANDBOX_NETWORK_DISABLED_ENV_VAR
+            );
+            return;
+        }
+
+        let server = wiremock::MockServer::start().await;
+
+        // OpenAI‑compat models endpoint responds OK.
+        wiremock::Mock::given(wiremock::matchers::method("GET"))
+            .and(wiremock::matchers::path("/v1/models"))
+            .respond_with(wiremock::ResponseTemplate::new(200))
+            .mount(&server)
+            .await;
+
+        OllamaClient::try_from_provider_with_base_url(&format!("{}/v1", server.uri()))
+            .await
+            .expect("client should be created when probe succeeds");
+    }
+
+    #[tokio::test]
+    async fn test_try_from_oss_provider_err_when_server_missing() {
+        if std::env::var(codex_core::spawn::CODEX_SANDBOX_NETWORK_DISABLED_ENV_VAR).is_ok() {
+            tracing::info!(
+                "{} set; skipping test_try_from_oss_provider_err_when_server_missing",
+                codex_core::spawn::CODEX_SANDBOX_NETWORK_DISABLED_ENV_VAR
+            );
+            return;
+        }
+
+        let server = wiremock::MockServer::start().await;
+        let err = OllamaClient::try_from_provider_with_base_url(&format!("{}/v1", server.uri()))
+            .await
+            .err()
+            .expect("expected error");
+        assert_eq!(OLLAMA_CONNECTION_ERROR, err.to_string());
+    }
+}

codex-rs/ollama/src/lib.rs

@@ -0,0 +1,44 @@
+mod client;
+mod parser;
+mod pull;
+mod url;
+
+pub use client::OllamaClient;
+use codex_core::config::Config;
+pub use pull::CliProgressReporter;
+pub use pull::PullEvent;
+pub use pull::PullProgressReporter;
+pub use pull::TuiProgressReporter;
+
+/// Default OSS model to use when `--oss` is passed without an explicit `-m`.
+pub const DEFAULT_OSS_MODEL: &str = "gpt-oss:20b";
+
+/// Prepare the local OSS environment when `--oss` is selected.
+///
+/// - Ensures a local Ollama server is reachable.
+/// - Checks if the model exists locally and pulls it if missing.
+pub async fn ensure_oss_ready(config: &Config) -> std::io::Result<()> {
+    // Only download when the requested model is the default OSS model (or when -m is not provided).
+    let model = config.model.as_ref();
+
+    // Verify local Ollama is reachable.
+    let ollama_client = crate::OllamaClient::try_from_oss_provider(config).await?;
+
+    // If the model is not present locally, pull it.
+    match ollama_client.fetch_models().await {
+        Ok(models) => {
+            if !models.iter().any(|m| m == model) {
+                let mut reporter = crate::CliProgressReporter::new();
+                ollama_client
+                    .pull_with_reporter(model, &mut reporter)
+                    .await?;
+            }
+        }
+        Err(err) => {
+            // Not fatal; higher layers may still proceed and surface errors later.
+            tracing::warn!("Failed to query local models from Ollama: {}.", err);
+        }
+    }
+
+    Ok(())
+}

codex-rs/ollama/src/parser.rs

@@ -0,0 +1,82 @@
+use serde_json::Value as JsonValue;
+
+use crate::pull::PullEvent;
+
+// Convert a single JSON object representing a pull update into one or more events.
+pub(crate) fn pull_events_from_value(value: &JsonValue) -> Vec<PullEvent> {
+    let mut events = Vec::new();
+    if let Some(status) = value.get("status").and_then(|s| s.as_str()) {
+        events.push(PullEvent::Status(status.to_string()));
+        if status == "success" {
+            events.push(PullEvent::Success);
+        }
+    }
+    let digest = value
+        .get("digest")
+        .and_then(|d| d.as_str())
+        .unwrap_or("")
+        .to_string();
+    let total = value.get("total").and_then(|t| t.as_u64());
+    let completed = value.get("completed").and_then(|t| t.as_u64());
+    if total.is_some() || completed.is_some() {
+        events.push(PullEvent::ChunkProgress {
+            digest,
+            total,
+            completed,
+        });
+    }
+    events
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn test_pull_events_decoder_status_and_success() {
+        let v: JsonValue = serde_json::json!({"status":"verifying"});
+        let events = pull_events_from_value(&v);
+        assert!(matches!(events.as_slice(), [PullEvent::Status(s)] if s == "verifying"));
+
+        let v2: JsonValue = serde_json::json!({"status":"success"});
+        let events2 = pull_events_from_value(&v2);
+        assert_eq!(events2.len(), 2);
+        assert!(matches!(events2[0], PullEvent::Status(ref s) if s == "success"));
+        assert!(matches!(events2[1], PullEvent::Success));
+    }
+
+    #[test]
+    fn test_pull_events_decoder_progress() {
+        let v: JsonValue = serde_json::json!({"digest":"sha256:abc","total":100});
+        let events = pull_events_from_value(&v);
+        assert_eq!(events.len(), 1);
+        match &events[0] {
+            PullEvent::ChunkProgress {
+                digest,
+                total,
+                completed,
+            } => {
+                assert_eq!(digest, "sha256:abc");
+                assert_eq!(*total, Some(100));
+                assert_eq!(*completed, None);
+            }
+            _ => panic!("expected ChunkProgress"),
+        }
+
+        let v2: JsonValue = serde_json::json!({"digest":"sha256:def","completed":42});
+        let events2 = pull_events_from_value(&v2);
+        assert_eq!(events2.len(), 1);
+        match &events2[0] {
+            PullEvent::ChunkProgress {
+                digest,
+                total,
+                completed,
+            } => {
+                assert_eq!(digest, "sha256:def");
+                assert_eq!(*total, None);
+                assert_eq!(*completed, Some(42));
+            }
+            _ => panic!("expected ChunkProgress"),
+        }
+    }
+}

codex-rs/ollama/src/pull.rs

@@ -0,0 +1,147 @@
+use std::collections::HashMap;
+use std::io;
+use std::io::Write;
+
+/// Events emitted while pulling a model from Ollama.
+#[derive(Debug, Clone)]
+pub enum PullEvent {
+    /// A human-readable status message (e.g., "verifying", "writing").
+    Status(String),
+    /// Byte-level progress update for a specific layer digest.
+    ChunkProgress {
+        digest: String,
+        total: Option<u64>,
+        completed: Option<u64>,
+    },
+    /// The pull finished successfully.
+    Success,
+
+    /// Error event with a message.
+    Error(String),
+}
+
+/// A simple observer for pull progress events. Implementations decide how to
+/// render progress (CLI, TUI, logs, ...).
+pub trait PullProgressReporter {
+    fn on_event(&mut self, event: &PullEvent) -> io::Result<()>;
+}
+
+/// A minimal CLI reporter that writes inline progress to stderr.
+pub struct CliProgressReporter {
+    printed_header: bool,
+    last_line_len: usize,
+    last_completed_sum: u64,
+    last_instant: std::time::Instant,
+    totals_by_digest: HashMap<String, (u64, u64)>,
+}
+
+impl Default for CliProgressReporter {
+    fn default() -> Self {
+        Self::new()
+    }
+}
+
+impl CliProgressReporter {
+    pub fn new() -> Self {
+        Self {
+            printed_header: false,
+            last_line_len: 0,
+            last_completed_sum: 0,
+            last_instant: std::time::Instant::now(),
+            totals_by_digest: HashMap::new(),
+        }
+    }
+}
+
+impl PullProgressReporter for CliProgressReporter {
+    fn on_event(&mut self, event: &PullEvent) -> io::Result<()> {
+        let mut out = std::io::stderr();
+        match event {
+            PullEvent::Status(status) => {
+                // Avoid noisy manifest messages; otherwise show status inline.
+                if status.eq_ignore_ascii_case("pulling manifest") {
+                    return Ok(());
+                }
+                let pad = self.last_line_len.saturating_sub(status.len());
+                let line = format!("\r{status}{}", " ".repeat(pad));
+                self.last_line_len = status.len();
+                out.write_all(line.as_bytes())?;
+                out.flush()
+            }
+            PullEvent::ChunkProgress {
+                digest,
+                total,
+                completed,
+            } => {
+                if let Some(t) = *total {
+                    self.totals_by_digest
+                        .entry(digest.clone())
+                        .or_insert((0, 0))
+                        .0 = t;
+                }
+                if let Some(c) = *completed {
+                    self.totals_by_digest
+                        .entry(digest.clone())
+                        .or_insert((0, 0))
+                        .1 = c;
+                }
+
+                let (sum_total, sum_completed) = self
+                    .totals_by_digest
+                    .values()
+                    .fold((0u64, 0u64), |acc, (t, c)| (acc.0 + *t, acc.1 + *c));
+                if sum_total > 0 {
+                    if !self.printed_header {
+                        let gb = (sum_total as f64) / (1024.0 * 1024.0 * 1024.0);
+                        let header = format!("Downloading model: total {gb:.2} GB\n");
+                        out.write_all(b"\r\x1b[2K")?;
+                        out.write_all(header.as_bytes())?;
+                        self.printed_header = true;
+                    }
+                    let now = std::time::Instant::now();
+                    let dt = now
+                        .duration_since(self.last_instant)
+                        .as_secs_f64()
+                        .max(0.001);
+                    let dbytes = sum_completed.saturating_sub(self.last_completed_sum) as f64;
+                    let speed_mb_s = dbytes / (1024.0 * 1024.0) / dt;
+                    self.last_completed_sum = sum_completed;
+                    self.last_instant = now;
+
+                    let done_gb = (sum_completed as f64) / (1024.0 * 1024.0 * 1024.0);
+                    let total_gb = (sum_total as f64) / (1024.0 * 1024.0 * 1024.0);
+                    let pct = (sum_completed as f64) * 100.0 / (sum_total as f64);
+                    let text =
+                        format!("{done_gb:.2}/{total_gb:.2} GB ({pct:.1}%) {speed_mb_s:.1} MB/s");
+                    let pad = self.last_line_len.saturating_sub(text.len());
+                    let line = format!("\r{text}{}", " ".repeat(pad));
+                    self.last_line_len = text.len();
+                    out.write_all(line.as_bytes())?;
+                    out.flush()
+                } else {
+                    Ok(())
+                }
+            }
+            PullEvent::Error(_) => {
+                // This will be handled by the caller, so we don't do anything
+                // here or the error will be printed twice.
+                Ok(())
+            }
+            PullEvent::Success => {
+                out.write_all(b"\n")?;
+                out.flush()
+            }
+        }
+    }
+}
+
+/// For now the TUI reporter delegates to the CLI reporter. This keeps UI and
+/// CLI behavior aligned until a dedicated TUI integration is implemented.
+#[derive(Default)]
+pub struct TuiProgressReporter(CliProgressReporter);
+
+impl PullProgressReporter for TuiProgressReporter {
+    fn on_event(&mut self, event: &PullEvent) -> io::Result<()> {
+        self.0.on_event(event)
+    }
+}

codex-rs/ollama/src/url.rs

@@ -0,0 +1,39 @@
+/// Identify whether a base_url points at an OpenAI-compatible root (".../v1").
+pub(crate) fn is_openai_compatible_base_url(base_url: &str) -> bool {
+    base_url.trim_end_matches('/').ends_with("/v1")
+}
+
+/// Convert a provider base_url into the native Ollama host root.
+/// For example, "http://localhost:11434/v1" -> "http://localhost:11434".
+pub fn base_url_to_host_root(base_url: &str) -> String {
+    let trimmed = base_url.trim_end_matches('/');
+    if trimmed.ends_with("/v1") {
+        trimmed
+            .trim_end_matches("/v1")
+            .trim_end_matches('/')
+            .to_string()
+    } else {
+        trimmed.to_string()
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn test_base_url_to_host_root() {
+        assert_eq!(
+            base_url_to_host_root("http://localhost:11434/v1"),
+            "http://localhost:11434"
+        );
+        assert_eq!(
+            base_url_to_host_root("http://localhost:11434"),
+            "http://localhost:11434"
+        );
+        assert_eq!(
+            base_url_to_host_root("http://localhost:11434/"),
+            "http://localhost:11434"
+        );
+    }
+}

codex-rs/scripts/create_github_release.sh

@@ -19,7 +19,33 @@ if ! git diff --quiet || ! git diff --cached --quiet || [ -n "$(git ls-files --o
 fi
 
 # Fail if in a detached HEAD state.
-CURRENT_BRANCH=$(git symbolic-ref --short -q HEAD)
+CURRENT_BRANCH=$(git symbolic-ref --short -q HEAD 2>/dev/null || true)
+if [ -z "${CURRENT_BRANCH:-}" ]; then
+  echo "ERROR: Could not determine the current branch (detached HEAD?)." >&2
+  echo "       Please run this script from a checked-out branch." >&2
+  exit 1
+fi
+
+# Ensure we are on the 'main' branch before proceeding.
+if [ "${CURRENT_BRANCH}" != "main" ]; then
+  echo "ERROR: Releases must be created from the 'main' branch (current: '${CURRENT_BRANCH}')." >&2
+  echo "       Please switch to 'main' and try again." >&2
+  exit 1
+fi
+
+# Ensure the current local commit on 'main' is present on 'origin/main'.
+# This guarantees we only create releases from commits that are already on
+# the canonical repository (https://github.com/openai/codex).
+if ! git fetch --quiet origin main; then
+  echo "ERROR: Failed to fetch 'origin/main'. Ensure the 'origin' remote is configured and reachable." >&2
+  exit 1
+fi
+
+if ! git merge-base --is-ancestor HEAD origin/main; then
+  echo "ERROR: Your local 'main' HEAD commit is not present on 'origin/main'." >&2
+  echo "       Please push your commits first (git push origin main) or check out a commit on 'origin/main'." >&2
+  exit 1
+fi
 
 # Create a new branch for the release and make a commit with the new version.
 if [ $# -ge 1 ]; then

codex-rs/tui/Cargo.toml

@@ -11,6 +11,10 @@ path = "src/main.rs"
 name = "codex_tui"
 path = "src/lib.rs"
 
+[features]
+# Enable vt100-based tests (emulator) when running with `--features vt100-tests`.
+vt100-tests = []
+
 [lints]
 workspace = true
 
@@ -29,6 +33,7 @@ codex-common = { path = "../common", features = [
 codex-core = { path = "../core" }
 codex-file-search = { path = "../file-search" }
 codex-login = { path = "../login" }
+codex-ollama = { path = "../ollama" }
 color-eyre = "0.6.3"
 crossterm = { version = "0.28.1", features = ["bracketed-paste"] }
 image = { version = "^0.25.6", default-features = false, features = ["jpeg"] }
@@ -73,3 +78,4 @@ insta = "1.43.1"
 pretty_assertions = "1"
 rand = "0.8"
 chrono = { version = "0.4", features = ["serde"] }
+vt100 = "0.16.2"

codex-rs/tui/src/app.rs

@@ -233,7 +233,8 @@ impl App<'_> {
                                     widget.on_ctrl_c();
                                 }
                                 AppState::GitWarning { .. } => {
-                                    // No-op.
+                                    // Allow exiting the app with Ctrl+C from the warning screen.
+                                    self.app_event_tx.send(AppEvent::ExitRequest);
                                 }
                             }
                         }
@@ -299,6 +300,13 @@ impl App<'_> {
                         self.app_state = AppState::Chat { widget: new_widget };
                         self.app_event_tx.send(AppEvent::RequestRedraw);
                     }
+                    SlashCommand::Init => {
+                        // Guard: do not run if a task is active.
+                        if let AppState::Chat { widget } = &mut self.app_state {
+                            const INIT_PROMPT: &str = include_str!("../../../INIT.md");
+                            widget.submit_text_message(INIT_PROMPT.to_string());
+                        }
+                    }
                     SlashCommand::Compact => {
                         if let AppState::Chat { widget } = &mut self.app_state {
                             widget.clear_token_usage();
@@ -329,6 +337,11 @@ impl App<'_> {
                             widget.add_diff_output(text);
                         }
                     }
+                    SlashCommand::Status => {
+                        if let AppState::Chat { widget } = &mut self.app_state {
+                            widget.add_status_output();
+                        }
+                    }
                     #[cfg(debug_assertions)]
                     SlashCommand::TestApproval => {
                         use std::collections::HashMap;
@@ -415,7 +428,7 @@ impl App<'_> {
         let size = terminal.size()?;
         let desired_height = match &self.app_state {
             AppState::Chat { widget } => widget.desired_height(size.width),
-            AppState::GitWarning { .. } => 10,
+            AppState::GitWarning { .. } => size.height,
         };
 
         let mut area = terminal.viewport_area;

codex-rs/tui/src/bottom_pane/chat_composer.rs

@@ -729,6 +729,7 @@ impl WidgetRef for &ChatComposer {
 
 #[cfg(test)]
 mod tests {
+    use crate::app_event::AppEvent;
     use crate::bottom_pane::AppEventSender;
     use crate::bottom_pane::ChatComposer;
     use crate::bottom_pane::InputResult;
@@ -1004,6 +1005,49 @@ mod tests {
         }
     }
 
+    #[test]
+    fn slash_init_dispatches_command_and_does_not_submit_literal_text() {
+        use crossterm::event::KeyCode;
+        use crossterm::event::KeyEvent;
+        use crossterm::event::KeyModifiers;
+        use std::sync::mpsc::TryRecvError;
+
+        let (tx, rx) = std::sync::mpsc::channel();
+        let sender = AppEventSender::new(tx);
+        let mut composer = ChatComposer::new(true, sender, false);
+
+        // Type the slash command.
+        for ch in [
+            '/', 'i', 'n', 'i', 't', // "/init"
+        ] {
+            let _ = composer.handle_key_event(KeyEvent::new(KeyCode::Char(ch), KeyModifiers::NONE));
+        }
+
+        // Press Enter to dispatch the selected command.
+        let (result, _needs_redraw) =
+            composer.handle_key_event(KeyEvent::new(KeyCode::Enter, KeyModifiers::NONE));
+
+        // When a slash command is dispatched, the composer should not submit
+        // literal text and should clear its textarea.
+        match result {
+            InputResult::None => {}
+            InputResult::Submitted(text) => {
+                panic!("expected command dispatch, but composer submitted literal text: {text}")
+            }
+        }
+        assert!(composer.textarea.is_empty(), "composer should be cleared");
+
+        // Verify a DispatchCommand event for the "init" command was sent.
+        match rx.try_recv() {
+            Ok(AppEvent::DispatchCommand(cmd)) => {
+                assert_eq!(cmd.command(), "init");
+            }
+            Ok(_other) => panic!("unexpected app event"),
+            Err(TryRecvError::Empty) => panic!("expected a DispatchCommand event for '/init'"),
+            Err(TryRecvError::Disconnected) => panic!("app event channel disconnected"),
+        }
+    }
+
     #[test]
     fn test_multiple_pastes_submission() {
         use crossterm::event::KeyCode;

codex-rs/tui/src/bottom_pane/command_popup.rs

@@ -188,3 +188,38 @@ impl WidgetRef for CommandPopup {
         table.render(area, buf);
     }
 }
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn filter_includes_init_when_typing_prefix() {
+        let mut popup = CommandPopup::new();
+        // Simulate the composer line starting with '/in' so the popup filters
+        // matching commands by prefix.
+        popup.on_composer_text_change("/in".to_string());
+
+        // Access the filtered list via the selected command and ensure that
+        // one of the matches is the new "init" command.
+        let matches = popup.filtered_commands();
+        assert!(
+            matches.iter().any(|cmd| cmd.command() == "init"),
+            "expected '/init' to appear among filtered commands"
+        );
+    }
+
+    #[test]
+    fn selecting_init_by_exact_match() {
+        let mut popup = CommandPopup::new();
+        popup.on_composer_text_change("/init".to_string());
+
+        // When an exact match exists, the selected command should be that
+        // command by default.
+        let selected = popup.selected_command();
+        match selected {
+            Some(cmd) => assert_eq!(cmd.command(), "init"),
+            None => panic!("expected a selected command for exact match"),
+        }
+    }
+}

codex-rs/tui/src/bottom_pane/live_ring_widget.rs

@@ -0,0 +1,45 @@
+use ratatui::buffer::Buffer;
+use ratatui::layout::Rect;
+use ratatui::text::Line;
+use ratatui::widgets::Paragraph;
+use ratatui::widgets::WidgetRef;
+
+/// Minimal rendering-only widget for the transient ring rows.
+pub(crate) struct LiveRingWidget {
+    max_rows: u16,
+    rows: Vec<Line<'static>>, // newest at the end
+}
+
+impl LiveRingWidget {
+    pub fn new() -> Self {
+        Self {
+            max_rows: 3,
+            rows: Vec::new(),
+        }
+    }
+
+    pub fn set_max_rows(&mut self, n: u16) {
+        self.max_rows = n.max(1);
+    }
+
+    pub fn set_rows(&mut self, rows: Vec<Line<'static>>) {
+        self.rows = rows;
+    }
+
+    pub fn desired_height(&self, _width: u16) -> u16 {
+        let len = self.rows.len() as u16;
+        len.min(self.max_rows)
+    }
+}
+
+impl WidgetRef for LiveRingWidget {
+    fn render_ref(&self, area: Rect, buf: &mut Buffer) {
+        if area.height == 0 {
+            return;
+        }
+        let visible = self.rows.len().saturating_sub(self.max_rows as usize);
+        let slice = &self.rows[visible..];
+        let para = Paragraph::new(slice.to_vec());
+        para.render_ref(area, buf);
+    }
+}

codex-rs/tui/src/bottom_pane/mod.rs

@@ -4,12 +4,12 @@ use crate::app_event::AppEvent;
 use crate::app_event_sender::AppEventSender;
 use crate::user_approval_widget::ApprovalRequest;
 use bottom_pane_view::BottomPaneView;
-use bottom_pane_view::ConditionalUpdate;
 use codex_core::protocol::TokenUsage;
 use codex_file_search::FileMatch;
 use crossterm::event::KeyEvent;
 use ratatui::buffer::Buffer;
 use ratatui::layout::Rect;
+use ratatui::text::Line;
 use ratatui::widgets::WidgetRef;
 
 mod approval_modal_view;
@@ -18,6 +18,7 @@ mod chat_composer;
 mod chat_composer_history;
 mod command_popup;
 mod file_search_popup;
+mod live_ring_widget;
 mod status_indicator_view;
 mod textarea;
 
@@ -30,6 +31,7 @@ pub(crate) enum CancellationEvent {
 pub(crate) use chat_composer::ChatComposer;
 pub(crate) use chat_composer::InputResult;
 
+use crate::status_indicator_widget::StatusIndicatorWidget;
 use approval_modal_view::ApprovalModalView;
 use status_indicator_view::StatusIndicatorView;
 
@@ -46,6 +48,19 @@ pub(crate) struct BottomPane<'a> {
     has_input_focus: bool,
     is_task_running: bool,
     ctrl_c_quit_hint: bool,
+
+    /// Optional live, multi‑line status/"live cell" rendered directly above
+    /// the composer while a task is running. Unlike `active_view`, this does
+    /// not replace the composer; it augments it.
+    live_status: Option<StatusIndicatorWidget>,
+
+    /// Optional transient ring shown above the composer. This is a rendering-only
+    /// container used during development before we wire it to ChatWidget events.
+    live_ring: Option<live_ring_widget::LiveRingWidget>,
+
+    /// True if the active view is the StatusIndicatorView that replaces the
+    /// composer during a running task.
+    status_view_active: bool,
 }
 
 pub(crate) struct BottomPaneParams {
@@ -55,6 +70,7 @@ pub(crate) struct BottomPaneParams {
 }
 
 impl BottomPane<'_> {
+    const BOTTOM_PAD_LINES: u16 = 2;
     pub fn new(params: BottomPaneParams) -> Self {
         let enhanced_keys_supported = params.enhanced_keys_supported;
         Self {
@@ -68,14 +84,40 @@ impl BottomPane<'_> {
             has_input_focus: params.has_input_focus,
             is_task_running: false,
             ctrl_c_quit_hint: false,
+            live_status: None,
+            live_ring: None,
+            status_view_active: false,
         }
     }
 
     pub fn desired_height(&self, width: u16) -> u16 {
-        self.active_view
+        let overlay_status_h = self
+            .live_status
             .as_ref()
-            .map(|v| v.desired_height(width))
-            .unwrap_or(self.composer.desired_height(width))
+            .map(|s| s.desired_height(width))
+            .unwrap_or(0);
+        let ring_h = self
+            .live_ring
+            .as_ref()
+            .map(|r| r.desired_height(width))
+            .unwrap_or(0);
+
+        let view_height = if let Some(view) = self.active_view.as_ref() {
+            // Add a single blank spacer line between live ring and status view when active.
+            let spacer = if self.live_ring.is_some() && self.status_view_active {
+                1
+            } else {
+                0
+            };
+            spacer + view.desired_height(width)
+        } else {
+            self.composer.desired_height(width)
+        };
+
+        overlay_status_h
+            .saturating_add(ring_h)
+            .saturating_add(view_height)
+            .saturating_add(Self::BOTTOM_PAD_LINES)
     }
 
     pub fn cursor_pos(&self, area: Rect) -> Option<(u16, u16)> {
@@ -97,9 +139,10 @@ impl BottomPane<'_> {
             if !view.is_complete() {
                 self.active_view = Some(view);
             } else if self.is_task_running {
-                self.active_view = Some(Box::new(StatusIndicatorView::new(
-                    self.app_event_tx.clone(),
-                )));
+                let mut v = StatusIndicatorView::new(self.app_event_tx.clone());
+                v.update_text("waiting for model".to_string());
+                self.active_view = Some(Box::new(v));
+                self.status_view_active = true;
             }
             self.request_redraw();
             InputResult::None
@@ -126,9 +169,11 @@ impl BottomPane<'_> {
                 if !view.is_complete() {
                     self.active_view = Some(view);
                 } else if self.is_task_running {
-                    self.active_view = Some(Box::new(StatusIndicatorView::new(
-                        self.app_event_tx.clone(),
-                    )));
+                    // Modal aborted but task still running – restore status indicator.
+                    let mut v = StatusIndicatorView::new(self.app_event_tx.clone());
+                    v.update_text("waiting for model".to_string());
+                    self.active_view = Some(Box::new(v));
+                    self.status_view_active = true;
                 }
                 self.show_ctrl_c_quit_hint();
             }
@@ -148,19 +193,42 @@ impl BottomPane<'_> {
         }
     }
 
-    /// Update the status indicator text (only when the `StatusIndicatorView` is
-    /// active).
+    /// Update the status indicator text. Prefer replacing the composer with
+    /// the StatusIndicatorView so the input pane shows a single-line status
+    /// like: `▌ Working waiting for model`.
     pub(crate) fn update_status_text(&mut self, text: String) {
-        if let Some(view) = &mut self.active_view {
-            match view.update_status_text(text) {
-                ConditionalUpdate::NeedsRedraw => {
-                    self.request_redraw();
-                }
-                ConditionalUpdate::NoRedraw => {
-                    // No redraw needed.
-                }
+        let mut handled_by_view = false;
+        if let Some(view) = self.active_view.as_mut() {
+            if matches!(
+                view.update_status_text(text.clone()),
+                bottom_pane_view::ConditionalUpdate::NeedsRedraw
+            ) {
+                handled_by_view = true;
             }
+        } else {
+            let mut v = StatusIndicatorView::new(self.app_event_tx.clone());
+            v.update_text(text.clone());
+            self.active_view = Some(Box::new(v));
+            self.status_view_active = true;
+            handled_by_view = true;
         }
+
+        // Fallback: if the current active view did not consume status updates
+        // and no modal view is active, present an overlay above the composer.
+        // If a modal is active, do NOT render the overlay to avoid drawing
+        // over the dialog.
+        if !handled_by_view && self.active_view.is_none() {
+            if self.live_status.is_none() {
+                self.live_status = Some(StatusIndicatorWidget::new(self.app_event_tx.clone()));
+            }
+            if let Some(status) = &mut self.live_status {
+                status.update_text(text);
+            }
+        } else if !handled_by_view {
+            // Ensure any previous overlay is cleared when a modal becomes active.
+            self.live_status = None;
+        }
+        self.request_redraw();
     }
 
     pub(crate) fn show_ctrl_c_quit_hint(&mut self) {
@@ -186,27 +254,23 @@ impl BottomPane<'_> {
     pub fn set_task_running(&mut self, running: bool) {
         self.is_task_running = running;
 
-        match (running, self.active_view.is_some()) {
-            (true, false) => {
-                // Show status indicator overlay.
+        if running {
+            if self.active_view.is_none() {
                 self.active_view = Some(Box::new(StatusIndicatorView::new(
                     self.app_event_tx.clone(),
                 )));
-                self.request_redraw();
+                self.status_view_active = true;
             }
-            (false, true) => {
-                if let Some(mut view) = self.active_view.take() {
-                    if view.should_hide_when_task_is_done() {
-                        // Leave self.active_view as None.
-                        self.request_redraw();
-                    } else {
-                        // Preserve the view.
-                        self.active_view = Some(view);
-                    }
+            self.request_redraw();
+        } else {
+            self.live_status = None;
+            // Drop the status view when a task completes, but keep other
+            // modal views (e.g. approval dialogs).
+            if let Some(mut view) = self.active_view.take() {
+                if !view.should_hide_when_task_is_done() {
+                    self.active_view = Some(view);
                 }
-            }
-            _ => {
-                // No change.
+                self.status_view_active = false;
             }
         }
     }
@@ -248,6 +312,9 @@ impl BottomPane<'_> {
         // Otherwise create a new approval modal overlay.
         let modal = ApprovalModalView::new(request, self.app_event_tx.clone());
         self.active_view = Some(Box::new(modal));
+        // Hide any overlay status while a modal is visible.
+        self.live_status = None;
+        self.status_view_active = false;
         self.request_redraw()
     }
 
@@ -281,15 +348,82 @@ impl BottomPane<'_> {
         self.composer.on_file_search_result(query, matches);
         self.request_redraw();
     }
+
+    /// Set the rows and cap for the transient live ring overlay.
+    pub(crate) fn set_live_ring_rows(&mut self, max_rows: u16, rows: Vec<Line<'static>>) {
+        let mut w = live_ring_widget::LiveRingWidget::new();
+        w.set_max_rows(max_rows);
+        w.set_rows(rows);
+        self.live_ring = Some(w);
+    }
+
+    pub(crate) fn clear_live_ring(&mut self) {
+        self.live_ring = None;
+    }
+
+    // Removed restart_live_status_with_text – no longer used by the current streaming UI.
 }
 
 impl WidgetRef for &BottomPane<'_> {
     fn render_ref(&self, area: Rect, buf: &mut Buffer) {
-        // Show BottomPaneView if present.
-        if let Some(ov) = &self.active_view {
-            ov.render(area, buf);
-        } else {
-            (&self.composer).render_ref(area, buf);
+        let mut y_offset = 0u16;
+        if let Some(ring) = &self.live_ring {
+            let live_h = ring.desired_height(area.width).min(area.height);
+            if live_h > 0 {
+                let live_rect = Rect {
+                    x: area.x,
+                    y: area.y,
+                    width: area.width,
+                    height: live_h,
+                };
+                ring.render_ref(live_rect, buf);
+                y_offset = live_h;
+            }
+        }
+        // Spacer between live ring and status view when active
+        if self.live_ring.is_some() && self.status_view_active && y_offset < area.height {
+            // Leave one empty line
+            y_offset = y_offset.saturating_add(1);
+        }
+        if let Some(status) = &self.live_status {
+            let live_h = status
+                .desired_height(area.width)
+                .min(area.height.saturating_sub(y_offset));
+            if live_h > 0 {
+                let live_rect = Rect {
+                    x: area.x,
+                    y: area.y + y_offset,
+                    width: area.width,
+                    height: live_h,
+                };
+                status.render_ref(live_rect, buf);
+                y_offset = y_offset.saturating_add(live_h);
+            }
+        }
+
+        if let Some(view) = &self.active_view {
+            if y_offset < area.height {
+                // Reserve bottom padding lines; keep at least 1 line for the view.
+                let avail = area.height - y_offset;
+                let pad = BottomPane::BOTTOM_PAD_LINES.min(avail.saturating_sub(1));
+                let view_rect = Rect {
+                    x: area.x,
+                    y: area.y + y_offset,
+                    width: area.width,
+                    height: avail - pad,
+                };
+                view.render(view_rect, buf);
+            }
+        } else if y_offset < area.height {
+            let composer_rect = Rect {
+                x: area.x,
+                y: area.y + y_offset,
+                width: area.width,
+                // Reserve bottom padding
+                height: (area.height - y_offset)
+                    - BottomPane::BOTTOM_PAD_LINES.min((area.height - y_offset).saturating_sub(1)),
+            };
+            (&self.composer).render_ref(composer_rect, buf);
         }
     }
 }
@@ -298,6 +432,9 @@ impl WidgetRef for &BottomPane<'_> {
 mod tests {
     use super::*;
     use crate::app_event::AppEvent;
+    use ratatui::buffer::Buffer;
+    use ratatui::layout::Rect;
+    use ratatui::text::Line;
     use std::path::PathBuf;
     use std::sync::mpsc::channel;
 
@@ -324,4 +461,316 @@ mod tests {
         assert!(pane.ctrl_c_quit_hint_visible());
         assert_eq!(CancellationEvent::Ignored, pane.on_ctrl_c());
     }
+
+    #[test]
+    fn live_ring_renders_above_composer() {
+        let (tx_raw, _rx) = channel::<AppEvent>();
+        let tx = AppEventSender::new(tx_raw);
+        let mut pane = BottomPane::new(BottomPaneParams {
+            app_event_tx: tx,
+            has_input_focus: true,
+            enhanced_keys_supported: false,
+        });
+
+        // Provide 4 rows with max_rows=3; only the last 3 should be visible.
+        pane.set_live_ring_rows(
+            3,
+            vec![
+                Line::from("one".to_string()),
+                Line::from("two".to_string()),
+                Line::from("three".to_string()),
+                Line::from("four".to_string()),
+            ],
+        );
+
+        let area = Rect::new(0, 0, 10, 5);
+        let mut buf = Buffer::empty(area);
+        (&pane).render_ref(area, &mut buf);
+
+        // Extract the first 3 rows and assert they contain the last three lines.
+        let mut lines: Vec<String> = Vec::new();
+        for y in 0..3 {
+            let mut s = String::new();
+            for x in 0..area.width {
+                s.push(buf[(x, y)].symbol().chars().next().unwrap_or(' '));
+            }
+            lines.push(s.trim_end().to_string());
+        }
+        assert_eq!(lines, vec!["two", "three", "four"]);
+    }
+
+    #[test]
+    fn status_indicator_visible_with_live_ring() {
+        let (tx_raw, _rx) = channel::<AppEvent>();
+        let tx = AppEventSender::new(tx_raw);
+        let mut pane = BottomPane::new(BottomPaneParams {
+            app_event_tx: tx,
+            has_input_focus: true,
+            enhanced_keys_supported: false,
+        });
+
+        // Simulate task running which replaces composer with the status indicator.
+        pane.set_task_running(true);
+        pane.update_status_text("waiting for model".to_string());
+
+        // Provide 2 rows in the live ring (e.g., streaming CoT) and ensure the
+        // status indicator remains visible below them.
+        pane.set_live_ring_rows(
+            2,
+            vec![
+                Line::from("cot1".to_string()),
+                Line::from("cot2".to_string()),
+            ],
+        );
+
+        // Allow some frames so the dot animation is present.
+        std::thread::sleep(std::time::Duration::from_millis(120));
+
+        // Height should include both ring rows, 1 spacer, and the 1-line status.
+        let area = Rect::new(0, 0, 30, 4);
+        let mut buf = Buffer::empty(area);
+        (&pane).render_ref(area, &mut buf);
+
+        // Top two rows are the live ring.
+        let mut r0 = String::new();
+        let mut r1 = String::new();
+        for x in 0..area.width {
+            r0.push(buf[(x, 0)].symbol().chars().next().unwrap_or(' '));
+            r1.push(buf[(x, 1)].symbol().chars().next().unwrap_or(' '));
+        }
+        assert!(r0.contains("cot1"), "expected first live row: {r0:?}");
+        assert!(r1.contains("cot2"), "expected second live row: {r1:?}");
+
+        // Row 2 is the spacer (blank)
+        let mut r2 = String::new();
+        for x in 0..area.width {
+            r2.push(buf[(x, 2)].symbol().chars().next().unwrap_or(' '));
+        }
+        assert!(r2.trim().is_empty(), "expected blank spacer line: {r2:?}");
+
+        // Bottom row is the status line; it should contain the left bar and "Working".
+        let mut r3 = String::new();
+        for x in 0..area.width {
+            r3.push(buf[(x, 3)].symbol().chars().next().unwrap_or(' '));
+        }
+        assert_eq!(buf[(0, 3)].symbol().chars().next().unwrap_or(' '), '▌');
+        assert!(
+            r3.contains("Working"),
+            "expected Working header in status line: {r3:?}"
+        );
+    }
+
+    #[test]
+    fn overlay_not_shown_above_approval_modal() {
+        let (tx_raw, _rx) = channel::<AppEvent>();
+        let tx = AppEventSender::new(tx_raw);
+        let mut pane = BottomPane::new(BottomPaneParams {
+            app_event_tx: tx,
+            has_input_focus: true,
+            enhanced_keys_supported: false,
+        });
+
+        // Create an approval modal (active view).
+        pane.push_approval_request(exec_request());
+        // Attempt to update status; this should NOT create an overlay while modal is visible.
+        pane.update_status_text("running command".to_string());
+
+        // Render and verify the top row does not include the Working header overlay.
+        let area = Rect::new(0, 0, 60, 6);
+        let mut buf = Buffer::empty(area);
+        (&pane).render_ref(area, &mut buf);
+
+        let mut r0 = String::new();
+        for x in 0..area.width {
+            r0.push(buf[(x, 0)].symbol().chars().next().unwrap_or(' '));
+        }
+        assert!(
+            !r0.contains("Working"),
+            "overlay Working header should not render above modal"
+        );
+    }
+
+    #[test]
+    fn composer_not_shown_after_denied_if_task_running() {
+        let (tx_raw, rx) = channel::<AppEvent>();
+        let tx = AppEventSender::new(tx_raw);
+        let mut pane = BottomPane::new(BottomPaneParams {
+            app_event_tx: tx.clone(),
+            has_input_focus: true,
+            enhanced_keys_supported: false,
+        });
+
+        // Start a running task so the status indicator replaces the composer.
+        pane.set_task_running(true);
+        pane.update_status_text("waiting for model".to_string());
+
+        // Push an approval modal (e.g., command approval) which should hide the status view.
+        pane.push_approval_request(exec_request());
+
+        // Simulate pressing 'n' (deny) on the modal.
+        use crossterm::event::KeyCode;
+        use crossterm::event::KeyEvent;
+        use crossterm::event::KeyModifiers;
+        pane.handle_key_event(KeyEvent::new(KeyCode::Char('n'), KeyModifiers::NONE));
+
+        // After denial, since the task is still running, the status indicator
+        // should be restored as the active view; the composer should NOT be visible.
+        assert!(
+            pane.status_view_active,
+            "status view should be active after denial"
+        );
+        assert!(pane.active_view.is_some(), "active view should be present");
+
+        // Render and ensure the top row includes the Working header instead of the composer.
+        // Give the animation thread a moment to tick.
+        std::thread::sleep(std::time::Duration::from_millis(120));
+        let area = Rect::new(0, 0, 40, 3);
+        let mut buf = Buffer::empty(area);
+        (&pane).render_ref(area, &mut buf);
+        let mut row0 = String::new();
+        for x in 0..area.width {
+            row0.push(buf[(x, 0)].symbol().chars().next().unwrap_or(' '));
+        }
+        assert!(
+            row0.contains("Working"),
+            "expected Working header after denial: {row0:?}"
+        );
+
+        // Drain the channel to avoid unused warnings.
+        drop(rx);
+    }
+
+    #[test]
+    fn status_indicator_visible_during_command_execution() {
+        let (tx_raw, _rx) = channel::<AppEvent>();
+        let tx = AppEventSender::new(tx_raw);
+        let mut pane = BottomPane::new(BottomPaneParams {
+            app_event_tx: tx,
+            has_input_focus: true,
+            enhanced_keys_supported: false,
+        });
+
+        // Begin a task: show initial status.
+        pane.set_task_running(true);
+        pane.update_status_text("waiting for model".to_string());
+
+        // As a long-running command begins (post-approval), ensure the status
+        // indicator is visible while we wait for the command to run.
+        pane.update_status_text("running command".to_string());
+
+        // Allow some frames so the animation thread ticks.
+        std::thread::sleep(std::time::Duration::from_millis(120));
+
+        // Render and confirm the line contains the "Working" header.
+        let area = Rect::new(0, 0, 40, 3);
+        let mut buf = Buffer::empty(area);
+        (&pane).render_ref(area, &mut buf);
+
+        let mut row0 = String::new();
+        for x in 0..area.width {
+            row0.push(buf[(x, 0)].symbol().chars().next().unwrap_or(' '));
+        }
+        assert!(
+            row0.contains("Working"),
+            "expected Working header: {row0:?}"
+        );
+    }
+
+    #[test]
+    fn bottom_padding_present_for_status_view() {
+        let (tx_raw, _rx) = channel::<AppEvent>();
+        let tx = AppEventSender::new(tx_raw);
+        let mut pane = BottomPane::new(BottomPaneParams {
+            app_event_tx: tx,
+            has_input_focus: true,
+            enhanced_keys_supported: false,
+        });
+
+        // Activate spinner (status view replaces composer) with no live ring.
+        pane.set_task_running(true);
+        pane.update_status_text("waiting for model".to_string());
+
+        // Use height == desired_height; expect 1 status row at top and 2 bottom padding rows.
+        let height = pane.desired_height(30);
+        assert!(
+            height >= 3,
+            "expected at least 3 rows with bottom padding; got {height}"
+        );
+        let area = Rect::new(0, 0, 30, height);
+        let mut buf = Buffer::empty(area);
+        (&pane).render_ref(area, &mut buf);
+
+        // Top row contains the status header
+        let mut top = String::new();
+        for x in 0..area.width {
+            top.push(buf[(x, 0)].symbol().chars().next().unwrap_or(' '));
+        }
+        assert_eq!(buf[(0, 0)].symbol().chars().next().unwrap_or(' '), '▌');
+        assert!(
+            top.contains("Working"),
+            "expected Working header on top row: {top:?}"
+        );
+
+        // Bottom two rows are blank padding
+        let mut r_last = String::new();
+        let mut r_last2 = String::new();
+        for x in 0..area.width {
+            r_last.push(buf[(x, height - 1)].symbol().chars().next().unwrap_or(' '));
+            r_last2.push(buf[(x, height - 2)].symbol().chars().next().unwrap_or(' '));
+        }
+        assert!(
+            r_last.trim().is_empty(),
+            "expected last row blank: {r_last:?}"
+        );
+        assert!(
+            r_last2.trim().is_empty(),
+            "expected second-to-last row blank: {r_last2:?}"
+        );
+    }
+
+    #[test]
+    fn bottom_padding_shrinks_when_tiny() {
+        let (tx_raw, _rx) = channel::<AppEvent>();
+        let tx = AppEventSender::new(tx_raw);
+        let mut pane = BottomPane::new(BottomPaneParams {
+            app_event_tx: tx,
+            has_input_focus: true,
+            enhanced_keys_supported: false,
+        });
+
+        pane.set_task_running(true);
+        pane.update_status_text("waiting for model".to_string());
+
+        // Height=2 → pad shrinks to 1; bottom row is blank, top row has spinner.
+        let area2 = Rect::new(0, 0, 20, 2);
+        let mut buf2 = Buffer::empty(area2);
+        (&pane).render_ref(area2, &mut buf2);
+        let mut row0 = String::new();
+        let mut row1 = String::new();
+        for x in 0..area2.width {
+            row0.push(buf2[(x, 0)].symbol().chars().next().unwrap_or(' '));
+            row1.push(buf2[(x, 1)].symbol().chars().next().unwrap_or(' '));
+        }
+        assert!(
+            row0.contains("Working"),
+            "expected Working header on row 0: {row0:?}"
+        );
+        assert!(
+            row1.trim().is_empty(),
+            "expected bottom padding on row 1: {row1:?}"
+        );
+
+        // Height=1 → no padding; single row is the spinner.
+        let area1 = Rect::new(0, 0, 20, 1);
+        let mut buf1 = Buffer::empty(area1);
+        (&pane).render_ref(area1, &mut buf1);
+        let mut only = String::new();
+        for x in 0..area1.width {
+            only.push(buf1[(x, 0)].symbol().chars().next().unwrap_or(' '));
+        }
+        assert!(
+            only.contains("Working"),
+            "expected Working header with no padding: {only:?}"
+        );
+    }
 }

codex-rs/tui/src/bottom_pane/textarea.rs

@@ -210,7 +210,7 @@ impl TextArea {
                 ..
             } => self.insert_str(&c.to_string()),
             KeyEvent {
-                code: KeyCode::Char('j'),
+                code: KeyCode::Char('j' | 'm'),
                 modifiers: KeyModifiers::CONTROL,
                 ..
             }
@@ -220,11 +220,22 @@ impl TextArea {
             } => self.insert_str("\n"),
             KeyEvent {
                 code: KeyCode::Backspace,
+                modifiers: KeyModifiers::ALT,
+                ..
+            } => self.delete_backward_word(),
+            KeyEvent {
+                code: KeyCode::Backspace,
+                modifiers: KeyModifiers::NONE,
                 ..
             } => self.delete_backward(1),
             KeyEvent {
                 code: KeyCode::Delete,
                 ..
+            }
+            | KeyEvent {
+                code: KeyCode::Char('d'),
+                modifiers: KeyModifiers::CONTROL,
+                ..
             } => self.delete_forward(1),
 
             KeyEvent {
@@ -303,14 +314,14 @@ impl TextArea {
             }
             KeyEvent {
                 code: KeyCode::Left,
-                modifiers: KeyModifiers::CONTROL,
+                modifiers: KeyModifiers::CONTROL | KeyModifiers::ALT,
                 ..
             } => {
                 self.set_cursor(self.beginning_of_previous_word());
             }
             KeyEvent {
                 code: KeyCode::Right,
-                modifiers: KeyModifiers::CONTROL,
+                modifiers: KeyModifiers::CONTROL | KeyModifiers::ALT,
                 ..
             } => {
                 self.set_cursor(self.end_of_next_word());

codex-rs/tui/src/chatwidget.rs

@@ -9,7 +9,10 @@ use codex_core::protocol::AgentMessageDeltaEvent;
 use codex_core::protocol::AgentMessageEvent;
 use codex_core::protocol::AgentReasoningDeltaEvent;
 use codex_core::protocol::AgentReasoningEvent;
+use codex_core::protocol::AgentReasoningRawContentDeltaEvent;
+use codex_core::protocol::AgentReasoningRawContentEvent;
 use codex_core::protocol::ApplyPatchApprovalRequestEvent;
+use codex_core::protocol::BackgroundEventEvent;
 use codex_core::protocol::ErrorEvent;
 use codex_core::protocol::Event;
 use codex_core::protocol::EventMsg;
@@ -23,6 +26,7 @@ use codex_core::protocol::Op;
 use codex_core::protocol::PatchApplyBeginEvent;
 use codex_core::protocol::TaskCompleteEvent;
 use codex_core::protocol::TokenUsage;
+use codex_core::protocol::TurnDiffEvent;
 use crossterm::event::KeyEvent;
 use crossterm::event::KeyEventKind;
 use ratatui::buffer::Buffer;
@@ -31,6 +35,7 @@ use ratatui::widgets::Widget;
 use ratatui::widgets::WidgetRef;
 use tokio::sync::mpsc::UnboundedSender;
 use tokio::sync::mpsc::unbounded_channel;
+use tracing::info;
 
 use crate::app_event::AppEvent;
 use crate::app_event_sender::AppEventSender;
@@ -42,8 +47,10 @@ use crate::exec_command::strip_bash_lc_and_escape;
 use crate::history_cell::CommandOutput;
 use crate::history_cell::HistoryCell;
 use crate::history_cell::PatchEventType;
+use crate::live_wrap::RowBuilder;
 use crate::user_approval_widget::ApprovalRequest;
 use codex_file_search::FileMatch;
+use ratatui::style::Stylize;
 
 struct RunningCommand {
     command: Vec<String>,
@@ -59,11 +66,16 @@ pub(crate) struct ChatWidget<'a> {
     initial_user_message: Option<UserMessage>,
     token_usage: TokenUsage,
     reasoning_buffer: String,
+    content_buffer: String,
     // Buffer for streaming assistant answer text; we do not surface partial
     // We wait for the final AgentMessage event and then emit the full text
     // at once into scrollback so the history contains a single message.
     answer_buffer: String,
     running_commands: HashMap<String, RunningCommand>,
+    live_builder: RowBuilder,
+    current_stream: Option<StreamKind>,
+    stream_header_emitted: bool,
+    live_max_rows: u16,
 }
 
 struct UserMessage {
@@ -71,6 +83,12 @@ struct UserMessage {
     image_paths: Vec<PathBuf>,
 }
 
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+enum StreamKind {
+    Answer,
+    Reasoning,
+}
+
 impl From<String> for UserMessage {
     fn from(text: String) -> Self {
         Self {
@@ -89,6 +107,24 @@ fn create_initial_user_message(text: String, image_paths: Vec<PathBuf>) -> Optio
 }
 
 impl ChatWidget<'_> {
+    fn emit_stream_header(&mut self, kind: StreamKind) {
+        use ratatui::text::Line as RLine;
+        if self.stream_header_emitted {
+            return;
+        }
+        let header = match kind {
+            StreamKind::Reasoning => RLine::from("thinking".magenta().italic()),
+            StreamKind::Answer => RLine::from("codex".magenta().bold()),
+        };
+        self.app_event_tx
+            .send(AppEvent::InsertHistory(vec![header]));
+        self.stream_header_emitted = true;
+    }
+    fn finalize_active_stream(&mut self) {
+        if let Some(kind) = self.current_stream {
+            self.finalize_stream(kind);
+        }
+    }
     pub(crate) fn new(
         config: Config,
         app_event_tx: AppEventSender,
@@ -149,8 +185,13 @@ impl ChatWidget<'_> {
             ),
             token_usage: TokenUsage::default(),
             reasoning_buffer: String::new(),
+            content_buffer: String::new(),
             answer_buffer: String::new(),
             running_commands: HashMap::new(),
+            live_builder: RowBuilder::new(80),
+            current_stream: None,
+            stream_header_emitted: false,
+            live_max_rows: 3,
         }
     }
 
@@ -234,58 +275,59 @@ impl ChatWidget<'_> {
 
                 self.request_redraw();
             }
-            EventMsg::AgentMessage(AgentMessageEvent { message }) => {
-                // Final assistant answer. Prefer the fully provided message
-                // from the event; if it is empty fall back to any accumulated
-                // delta buffer (some providers may only stream deltas and send
-                // an empty final message).
-                let full = if message.is_empty() {
-                    std::mem::take(&mut self.answer_buffer)
-                } else {
-                    self.answer_buffer.clear();
-                    message
-                };
-                if !full.is_empty() {
-                    self.add_to_history(HistoryCell::new_agent_message(&self.config, full));
-                }
+            EventMsg::AgentMessage(AgentMessageEvent { message: _ }) => {
+                // Final assistant answer: commit all remaining rows and close with
+                // a blank line. Use the final text if provided, otherwise rely on
+                // streamed deltas already in the builder.
+                self.finalize_stream(StreamKind::Answer);
                 self.request_redraw();
             }
             EventMsg::AgentMessageDelta(AgentMessageDeltaEvent { delta }) => {
-                // Buffer only – do not emit partial lines. This avoids cases
-                // where long responses appear truncated if the terminal
-                // wrapped early. The full message is emitted on
-                // AgentMessage.
+                self.begin_stream(StreamKind::Answer);
                 self.answer_buffer.push_str(&delta);
+                self.stream_push_and_maybe_commit(&delta);
+                self.request_redraw();
             }
             EventMsg::AgentReasoningDelta(AgentReasoningDeltaEvent { delta }) => {
-                // Buffer only – disable incremental reasoning streaming so we
-                // avoid truncated intermediate lines. Full text emitted on
-                // AgentReasoning.
+                // Stream CoT into the live pane; keep input visible and commit
+                // overflow rows incrementally to scrollback.
+                self.begin_stream(StreamKind::Reasoning);
                 self.reasoning_buffer.push_str(&delta);
+                self.stream_push_and_maybe_commit(&delta);
+                self.request_redraw();
             }
-            EventMsg::AgentReasoning(AgentReasoningEvent { text }) => {
-                // Emit full reasoning text once. Some providers might send
-                // final event with empty text if only deltas were used.
-                let full = if text.is_empty() {
-                    std::mem::take(&mut self.reasoning_buffer)
-                } else {
-                    self.reasoning_buffer.clear();
-                    text
-                };
-                if !full.is_empty() {
-                    self.add_to_history(HistoryCell::new_agent_reasoning(&self.config, full));
-                }
+            EventMsg::AgentReasoning(AgentReasoningEvent { text: _ }) => {
+                // Final reasoning: commit remaining rows and close with a blank.
+                self.finalize_stream(StreamKind::Reasoning);
+                self.request_redraw();
+            }
+            EventMsg::AgentReasoningRawContentDelta(AgentReasoningRawContentDeltaEvent {
+                delta,
+            }) => {
+                // Treat raw reasoning content the same as summarized reasoning for UI flow.
+                self.begin_stream(StreamKind::Reasoning);
+                self.reasoning_buffer.push_str(&delta);
+                self.stream_push_and_maybe_commit(&delta);
+                self.request_redraw();
+            }
+            EventMsg::AgentReasoningRawContent(AgentReasoningRawContentEvent { text: _ }) => {
+                // Finalize the raw reasoning stream just like the summarized reasoning event.
+                self.finalize_stream(StreamKind::Reasoning);
                 self.request_redraw();
             }
             EventMsg::TaskStarted => {
                 self.bottom_pane.clear_ctrl_c_quit_hint();
                 self.bottom_pane.set_task_running(true);
+                // Replace composer with single-line spinner while waiting.
+                self.bottom_pane
+                    .update_status_text("waiting for model".to_string());
                 self.request_redraw();
             }
             EventMsg::TaskComplete(TaskCompleteEvent {
                 last_agent_message: _,
             }) => {
                 self.bottom_pane.set_task_running(false);
+                self.bottom_pane.clear_live_ring();
                 self.request_redraw();
             }
             EventMsg::TokenCount(token_usage) => {
@@ -296,10 +338,18 @@ impl ChatWidget<'_> {
             EventMsg::Error(ErrorEvent { message }) => {
                 self.add_to_history(HistoryCell::new_error_event(message.clone()));
                 self.bottom_pane.set_task_running(false);
+                self.bottom_pane.clear_live_ring();
+                self.live_builder = RowBuilder::new(self.live_builder.width());
+                self.current_stream = None;
+                self.stream_header_emitted = false;
+                self.answer_buffer.clear();
+                self.reasoning_buffer.clear();
+                self.content_buffer.clear();
+                self.request_redraw();
             }
             EventMsg::PlanUpdate(update) => {
+                // Commit plan updates directly to history (no status-line preview).
                 self.add_to_history(HistoryCell::new_plan_update(update));
-                self.request_redraw();
             }
             EventMsg::ExecApprovalRequest(ExecApprovalRequestEvent {
                 call_id: _,
@@ -307,8 +357,8 @@ impl ChatWidget<'_> {
                 cwd,
                 reason,
             }) => {
-                // Print the command to the history so it is visible in the
-                // transcript *before* the modal asks for approval.
+                self.finalize_active_stream();
+                // Log a background summary immediately so the history is chronological.
                 let cmdline = strip_bash_lc_and_escape(&command);
                 let text = format!(
                     "command requires approval:\n$ {cmdline}{reason}",
@@ -334,6 +384,7 @@ impl ChatWidget<'_> {
                 reason,
                 grant_root,
             }) => {
+                self.finalize_active_stream();
                 // ------------------------------------------------------------------
                 // Before we even prompt the user for approval we surface the patch
                 // summary in the main conversation so that the dialog appears in a
@@ -344,7 +395,6 @@ impl ChatWidget<'_> {
                 // approval dialog) and avoids surprising the user with a modal
                 // prompt before they have seen *what* is being requested.
                 // ------------------------------------------------------------------
-
                 self.add_to_history(HistoryCell::new_patch_event(
                     PatchEventType::ApprovalRequest,
                     changes,
@@ -364,6 +414,10 @@ impl ChatWidget<'_> {
                 command,
                 cwd,
             }) => {
+                self.finalize_active_stream();
+                // Ensure the status indicator is visible while the command runs.
+                self.bottom_pane
+                    .update_status_text("running command".to_string());
                 self.running_commands.insert(
                     call_id,
                     RunningCommand {
@@ -379,13 +433,14 @@ impl ChatWidget<'_> {
                 auto_approved,
                 changes,
             }) => {
-                // Even when a patch is auto‑approved we still display the
-                // summary so the user can follow along.
                 self.add_to_history(HistoryCell::new_patch_event(
                     PatchEventType::ApplyBegin { auto_approved },
                     changes,
                 ));
             }
+            EventMsg::PatchApplyEnd(patch_apply_end_event) => {
+                self.add_to_history(HistoryCell::new_patch_end_event(patch_apply_end_event));
+            }
             EventMsg::ExecCommandEnd(ExecCommandEndEvent {
                 call_id,
                 exit_code,
@@ -393,6 +448,7 @@ impl ChatWidget<'_> {
                 stdout,
                 stderr,
             }) => {
+                // Compute summary before moving stdout into the history cell.
                 let cmd = self.running_commands.remove(&call_id);
                 self.add_to_history(HistoryCell::new_completed_exec_command(
                     cmd.map(|cmd| cmd.command).unwrap_or_else(|| vec![call_id]),
@@ -408,6 +464,7 @@ impl ChatWidget<'_> {
                 call_id: _,
                 invocation,
             }) => {
+                self.finalize_active_stream();
                 self.add_to_history(HistoryCell::new_active_mcp_tool_call(invocation));
             }
             EventMsg::McpToolCallEnd(McpToolCallEndEvent {
@@ -441,16 +498,20 @@ impl ChatWidget<'_> {
             EventMsg::ShutdownComplete => {
                 self.app_event_tx.send(AppEvent::ExitRequest);
             }
-            event => {
-                self.add_to_history(HistoryCell::new_background_event(format!("{event:?}")));
+            EventMsg::TurnDiff(TurnDiffEvent { unified_diff }) => {
+                info!("TurnDiffEvent: {unified_diff}");
+            }
+            EventMsg::BackgroundEvent(BackgroundEventEvent { message }) => {
+                info!("BackgroundEvent: {message}");
             }
         }
     }
 
     /// Update the live log preview while a task is running.
     pub(crate) fn update_latest_log(&mut self, line: String) {
-        // Forward only if we are currently showing the status indicator.
-        self.bottom_pane.update_status_text(line);
+        if self.bottom_pane.is_task_running() {
+            self.bottom_pane.update_status_text(line);
+        }
     }
 
     fn request_redraw(&mut self) {
@@ -461,6 +522,13 @@ impl ChatWidget<'_> {
         self.add_to_history(HistoryCell::new_diff_output(diff_output.clone()));
     }
 
+    pub(crate) fn add_status_output(&mut self) {
+        self.add_to_history(HistoryCell::new_status_output(
+            &self.config,
+            &self.token_usage,
+        ));
+    }
+
     /// Forward file-search results to the bottom pane.
     pub(crate) fn apply_file_search_result(&mut self, query: String, matches: Vec<FileMatch>) {
         self.bottom_pane.on_file_search_result(query, matches);
@@ -477,8 +545,15 @@ impl ChatWidget<'_> {
         if self.bottom_pane.is_task_running() {
             self.bottom_pane.clear_ctrl_c_quit_hint();
             self.submit_op(Op::Interrupt);
+            self.bottom_pane.set_task_running(false);
+            self.bottom_pane.clear_live_ring();
+            self.live_builder = RowBuilder::new(self.live_builder.width());
+            self.current_stream = None;
+            self.stream_header_emitted = false;
             self.answer_buffer.clear();
             self.reasoning_buffer.clear();
+            self.content_buffer.clear();
+            self.request_redraw();
             CancellationEvent::Ignored
         } else if self.bottom_pane.ctrl_c_quit_hint_visible() {
             self.submit_op(Op::Shutdown);
@@ -500,6 +575,16 @@ impl ChatWidget<'_> {
         }
     }
 
+    /// Programmatically submit a user text message as if typed in the
+    /// composer. The text will be added to conversation history and sent to
+    /// the agent.
+    pub(crate) fn submit_text_message(&mut self, text: String) {
+        if text.is_empty() {
+            return;
+        }
+        self.submit_user_message(text.into());
+    }
+
     pub(crate) fn token_usage(&self) -> &TokenUsage {
         &self.token_usage
     }
@@ -515,6 +600,104 @@ impl ChatWidget<'_> {
     }
 }
 
+impl ChatWidget<'_> {
+    fn begin_stream(&mut self, kind: StreamKind) {
+        if let Some(current) = self.current_stream {
+            if current != kind {
+                self.finalize_stream(current);
+            }
+        }
+
+        if self.current_stream != Some(kind) {
+            self.current_stream = Some(kind);
+            self.stream_header_emitted = false;
+            // Clear any previous live content; we're starting a new stream.
+            self.live_builder = RowBuilder::new(self.live_builder.width());
+            // Ensure the waiting status is visible (composer replaced).
+            self.bottom_pane
+                .update_status_text("waiting for model".to_string());
+            self.emit_stream_header(kind);
+        }
+    }
+
+    fn stream_push_and_maybe_commit(&mut self, delta: &str) {
+        self.live_builder.push_fragment(delta);
+
+        // Commit overflow rows (small batches) while keeping the last N rows visible.
+        let drained = self
+            .live_builder
+            .drain_commit_ready(self.live_max_rows as usize);
+        if !drained.is_empty() {
+            let mut lines: Vec<ratatui::text::Line<'static>> = Vec::new();
+            if !self.stream_header_emitted {
+                match self.current_stream {
+                    Some(StreamKind::Reasoning) => {
+                        lines.push(ratatui::text::Line::from("thinking".magenta().italic()));
+                    }
+                    Some(StreamKind::Answer) => {
+                        lines.push(ratatui::text::Line::from("codex".magenta().bold()));
+                    }
+                    None => {}
+                }
+                self.stream_header_emitted = true;
+            }
+            for r in drained {
+                lines.push(ratatui::text::Line::from(r.text));
+            }
+            self.app_event_tx.send(AppEvent::InsertHistory(lines));
+        }
+
+        // Update the live ring overlay lines (text-only, newest at bottom).
+        let rows = self
+            .live_builder
+            .display_rows()
+            .into_iter()
+            .map(|r| ratatui::text::Line::from(r.text))
+            .collect::<Vec<_>>();
+        self.bottom_pane
+            .set_live_ring_rows(self.live_max_rows, rows);
+    }
+
+    fn finalize_stream(&mut self, kind: StreamKind) {
+        if self.current_stream != Some(kind) {
+            // Nothing to do; either already finalized or not the active stream.
+            return;
+        }
+        // Flush any partial line as a full row, then drain all remaining rows.
+        self.live_builder.end_line();
+        let remaining = self.live_builder.drain_rows();
+        // TODO: Re-add markdown rendering for assistant answers and reasoning.
+        // When finalizing, pass the accumulated text through `markdown::append_markdown`
+        // to build styled `Line<'static>` entries instead of raw plain text lines.
+        if !remaining.is_empty() || !self.stream_header_emitted {
+            let mut lines: Vec<ratatui::text::Line<'static>> = Vec::new();
+            if !self.stream_header_emitted {
+                match kind {
+                    StreamKind::Reasoning => {
+                        lines.push(ratatui::text::Line::from("thinking".magenta().italic()));
+                    }
+                    StreamKind::Answer => {
+                        lines.push(ratatui::text::Line::from("codex".magenta().bold()));
+                    }
+                }
+                self.stream_header_emitted = true;
+            }
+            for r in remaining {
+                lines.push(ratatui::text::Line::from(r.text));
+            }
+            // Close the block with a blank line for readability.
+            lines.push(ratatui::text::Line::from(""));
+            self.app_event_tx.send(AppEvent::InsertHistory(lines));
+        }
+
+        // Clear the live overlay and reset state for the next stream.
+        self.live_builder = RowBuilder::new(self.live_builder.width());
+        self.bottom_pane.clear_live_ring();
+        self.current_stream = None;
+        self.stream_header_emitted = false;
+    }
+}
+
 impl WidgetRef for &ChatWidget<'_> {
     fn render_ref(&self, area: Rect, buf: &mut Buffer) {
         // In the hybrid inline viewport mode we only draw the interactive

codex-rs/tui/src/cli.rs

@@ -17,6 +17,12 @@ pub struct Cli {
     #[arg(long, short = 'm')]
     pub model: Option<String>,
 
+    /// Convenience flag to select the local open source model provider.
+    /// Equivalent to -c model_provider=oss; verifies a local Ollama server is
+    /// running.
+    #[arg(long = "oss", default_value_t = false)]
+    pub oss: bool,
+
     /// Configuration profile from config.toml to specify default options.
     #[arg(long = "profile", short = 'p')]
     pub config_profile: Option<String>,

codex-rs/tui/src/history_cell.rs

@@ -1,20 +1,19 @@
 use crate::exec_command::strip_bash_lc_and_escape;
-use crate::markdown::append_markdown;
 use crate::text_block::TextBlock;
 use crate::text_formatting::format_and_truncate_tool_result;
 use base64::Engine;
 use codex_ansi_escape::ansi_escape_line;
+use codex_common::create_config_summary_entries;
 use codex_common::elapsed::format_duration;
-use codex_common::summarize_sandbox_policy;
-use codex_core::WireApi;
 use codex_core::config::Config;
-use codex_core::model_supports_reasoning_summaries;
 use codex_core::plan_tool::PlanItemArg;
 use codex_core::plan_tool::StepStatus;
 use codex_core::plan_tool::UpdatePlanArgs;
 use codex_core::protocol::FileChange;
 use codex_core::protocol::McpInvocation;
+use codex_core::protocol::PatchApplyEndEvent;
 use codex_core::protocol::SessionConfiguredEvent;
+use codex_core::protocol::TokenUsage;
 use image::DynamicImage;
 use image::ImageReader;
 use mcp_types::EmbeddedResourceResource;
@@ -63,28 +62,35 @@ fn line_to_static(line: &Line) -> Line<'static> {
 /// scrollable list.
 pub(crate) enum HistoryCell {
     /// Welcome message.
-    WelcomeMessage { view: TextBlock },
+    WelcomeMessage {
+        view: TextBlock,
+    },
 
     /// Message from the user.
-    UserPrompt { view: TextBlock },
-
-    /// Message from the agent.
-    AgentMessage { view: TextBlock },
-
-    /// Reasoning event from the agent.
-    AgentReasoning { view: TextBlock },
+    UserPrompt {
+        view: TextBlock,
+    },
 
+    // AgentMessage and AgentReasoning variants were unused and have been removed.
     /// An exec tool call that has not finished yet.
-    ActiveExecCommand { view: TextBlock },
+    ActiveExecCommand {
+        view: TextBlock,
+    },
 
     /// Completed exec tool call.
-    CompletedExecCommand { view: TextBlock },
+    CompletedExecCommand {
+        view: TextBlock,
+    },
 
     /// An MCP tool call that has not finished yet.
-    ActiveMcpToolCall { view: TextBlock },
+    ActiveMcpToolCall {
+        view: TextBlock,
+    },
 
     /// Completed MCP tool call where we show the result serialized as JSON.
-    CompletedMcpToolCall { view: TextBlock },
+    CompletedMcpToolCall {
+        view: TextBlock,
+    },
 
     /// Completed MCP tool call where the result is an image.
     /// Admittedly, [mcp_types::CallToolResult] can have multiple content types,
@@ -94,28 +100,51 @@ pub(crate) enum HistoryCell {
     // resized version avoids doing the potentially expensive rescale twice
     // because the scroll-view first calls `height()` for layouting and then
     // `render_window()` for painting.
-    CompletedMcpToolCallWithImageOutput { _image: DynamicImage },
+    CompletedMcpToolCallWithImageOutput {
+        _image: DynamicImage,
+    },
 
     /// Background event.
-    BackgroundEvent { view: TextBlock },
+    BackgroundEvent {
+        view: TextBlock,
+    },
 
     /// Output from the `/diff` command.
-    GitDiffOutput { view: TextBlock },
+    GitDiffOutput {
+        view: TextBlock,
+    },
+
+    /// Output from the `/status` command.
+    StatusOutput {
+        view: TextBlock,
+    },
 
     /// Error event from the backend.
-    ErrorEvent { view: TextBlock },
+    ErrorEvent {
+        view: TextBlock,
+    },
 
     /// Info describing the newly-initialized session.
-    SessionInfo { view: TextBlock },
+    SessionInfo {
+        view: TextBlock,
+    },
 
     /// A pending code patch that is awaiting user approval. Mirrors the
     /// behaviour of `ActiveExecCommand` so the user sees *what* patch the
     /// model wants to apply before being prompted to approve or deny it.
-    PendingPatch { view: TextBlock },
+    PendingPatch {
+        view: TextBlock,
+    },
+
+    PatchEventEnd {
+        view: TextBlock,
+    },
 
     /// A human‑friendly rendering of the model's current plan and step
     /// statuses provided via the `update_plan` tool.
-    PlanUpdate { view: TextBlock },
+    PlanUpdate {
+        view: TextBlock,
+    },
 }
 
 const TOOL_CALL_MAX_LINES: usize = 5;
@@ -128,15 +157,15 @@ impl HistoryCell {
         match self {
             HistoryCell::WelcomeMessage { view }
             | HistoryCell::UserPrompt { view }
-            | HistoryCell::AgentMessage { view }
-            | HistoryCell::AgentReasoning { view }
             | HistoryCell::BackgroundEvent { view }
             | HistoryCell::GitDiffOutput { view }
+            | HistoryCell::StatusOutput { view }
             | HistoryCell::ErrorEvent { view }
             | HistoryCell::SessionInfo { view }
             | HistoryCell::CompletedExecCommand { view }
             | HistoryCell::CompletedMcpToolCall { view }
             | HistoryCell::PendingPatch { view }
+            | HistoryCell::PatchEventEnd { view }
             | HistoryCell::PlanUpdate { view }
             | HistoryCell::ActiveExecCommand { view, .. }
             | HistoryCell::ActiveMcpToolCall { view, .. } => {
@@ -177,26 +206,7 @@ impl HistoryCell {
                 ]),
             ];
 
-            let mut entries = vec![
-                ("workdir", config.cwd.display().to_string()),
-                ("model", config.model.clone()),
-                ("provider", config.model_provider_id.clone()),
-                ("approval", config.approval_policy.to_string()),
-                ("sandbox", summarize_sandbox_policy(&config.sandbox_policy)),
-            ];
-            if config.model_provider.wire_api == WireApi::Responses
-                && model_supports_reasoning_summaries(config)
-            {
-                entries.push((
-                    "reasoning effort",
-                    config.model_reasoning_effort.to_string(),
-                ));
-                entries.push((
-                    "reasoning summaries",
-                    config.model_reasoning_summary.to_string(),
-                ));
-            }
-            for (key, value) in entries {
+            for (key, value) in create_config_summary_entries(config) {
                 lines.push(Line::from(vec![format!("{key}: ").bold(), value.into()]));
             }
             lines.push(Line::from(""));
@@ -231,28 +241,6 @@ impl HistoryCell {
         }
     }
 
-    pub(crate) fn new_agent_message(config: &Config, message: String) -> Self {
-        let mut lines: Vec<Line<'static>> = Vec::new();
-        lines.push(Line::from("codex".magenta().bold()));
-        append_markdown(&message, &mut lines, config);
-        lines.push(Line::from(""));
-
-        HistoryCell::AgentMessage {
-            view: TextBlock::new(lines),
-        }
-    }
-
-    pub(crate) fn new_agent_reasoning(config: &Config, text: String) -> Self {
-        let mut lines: Vec<Line<'static>> = Vec::new();
-        lines.push(Line::from("thinking".magenta().italic()));
-        append_markdown(&text, &mut lines, config);
-        lines.push(Line::from(""));
-
-        HistoryCell::AgentReasoning {
-            view: TextBlock::new(lines),
-        }
-    }
-
     pub(crate) fn new_active_exec_command(command: Vec<String>) -> Self {
         let command_escaped = strip_bash_lc_and_escape(&command);
 
@@ -475,6 +463,49 @@ impl HistoryCell {
         }
     }
 
+    pub(crate) fn new_status_output(config: &Config, usage: &TokenUsage) -> Self {
+        let mut lines: Vec<Line<'static>> = Vec::new();
+        lines.push(Line::from("/status".magenta()));
+
+        // Config
+        for (key, value) in create_config_summary_entries(config) {
+            lines.push(Line::from(vec![format!("{key}: ").bold(), value.into()]));
+        }
+
+        // Token usage
+        lines.push(Line::from(""));
+        lines.push(Line::from("token usage".bold()));
+        lines.push(Line::from(vec![
+            "  input: ".bold(),
+            usage.input_tokens.to_string().into(),
+        ]));
+        lines.push(Line::from(vec![
+            "  cached input: ".bold(),
+            usage.cached_input_tokens.unwrap_or(0).to_string().into(),
+        ]));
+        lines.push(Line::from(vec![
+            "  output: ".bold(),
+            usage.output_tokens.to_string().into(),
+        ]));
+        lines.push(Line::from(vec![
+            "  reasoning output: ".bold(),
+            usage
+                .reasoning_output_tokens
+                .unwrap_or(0)
+                .to_string()
+                .into(),
+        ]));
+        lines.push(Line::from(vec![
+            "  total: ".bold(),
+            usage.total_tokens.to_string().into(),
+        ]));
+
+        lines.push(Line::from(""));
+        HistoryCell::StatusOutput {
+            view: TextBlock::new(lines),
+        }
+    }
+
     pub(crate) fn new_error_event(message: String) -> Self {
         let lines: Vec<Line<'static>> = vec![
             vec!["ERROR: ".red().bold(), message.into()].into(),
@@ -629,6 +660,28 @@ impl HistoryCell {
             view: TextBlock::new(lines),
         }
     }
+
+    pub(crate) fn new_patch_end_event(patch_apply_end_event: PatchApplyEndEvent) -> Self {
+        let PatchApplyEndEvent {
+            call_id: _,
+            stdout: _,
+            stderr,
+            success,
+        } = patch_apply_end_event;
+
+        let mut lines: Vec<Line<'static>> = if success {
+            vec![Line::from("patch applied successfully".italic())]
+        } else {
+            let mut lines = vec![Line::from("patch failed".italic())];
+            lines.extend(stderr.lines().map(|l| Line::from(l.to_string())));
+            lines
+        };
+        lines.push(Line::from(""));
+
+        HistoryCell::PatchEventEnd {
+            view: TextBlock::new(lines),
+        }
+    }
 }
 
 fn create_diff_summary(changes: HashMap<PathBuf, FileChange>) -> Vec<String> {

codex-rs/tui/src/insert_history.rs

@@ -14,7 +14,6 @@ use crossterm::style::SetBackgroundColor;
 use crossterm::style::SetColors;
 use crossterm::style::SetForegroundColor;
 use ratatui::layout::Size;
-use ratatui::prelude::Backend;
 use ratatui::style::Color;
 use ratatui::style::Modifier;
 use ratatui::text::Line;
@@ -22,6 +21,20 @@ use ratatui::text::Span;
 
 /// Insert `lines` above the viewport.
 pub(crate) fn insert_history_lines(terminal: &mut tui::Tui, lines: Vec<Line>) {
+    let mut out = std::io::stdout();
+    insert_history_lines_to_writer(terminal, &mut out, lines);
+}
+
+/// Like `insert_history_lines`, but writes ANSI to the provided writer. This
+/// is intended for testing where a capture buffer is used instead of stdout.
+pub fn insert_history_lines_to_writer<B, W>(
+    terminal: &mut crate::custom_terminal::Terminal<B>,
+    writer: &mut W,
+    lines: Vec<Line>,
+) where
+    B: ratatui::backend::Backend,
+    W: Write,
+{
     let screen_size = terminal.backend().size().unwrap_or(Size::new(0, 0));
     let cursor_pos = terminal.get_cursor_position().ok();
 
@@ -32,10 +45,22 @@ pub(crate) fn insert_history_lines(terminal: &mut tui::Tui, lines: Vec<Line>) {
         // If the viewport is not at the bottom of the screen, scroll it down to make room.
         // Don't scroll it past the bottom of the screen.
         let scroll_amount = wrapped_lines.min(screen_size.height - area.bottom());
-        terminal
-            .backend_mut()
-            .scroll_region_down(area.top()..screen_size.height, scroll_amount)
-            .ok();
+
+        // Emit ANSI to scroll the lower region (from the top of the viewport to the bottom
+        // of the screen) downward by `scroll_amount` lines. We do this by:
+        //   1) Limiting the scroll region to [area.top()+1 .. screen_height] (1-based bounds)
+        //   2) Placing the cursor at the top margin of that region
+        //   3) Emitting Reverse Index (RI, ESC M) `scroll_amount` times
+        //   4) Resetting the scroll region back to full screen
+        let top_1based = area.top() + 1; // Convert 0-based row to 1-based for DECSTBM
+        queue!(writer, SetScrollRegion(top_1based..screen_size.height)).ok();
+        queue!(writer, MoveTo(0, area.top())).ok();
+        for _ in 0..scroll_amount {
+            // Reverse Index (RI): ESC M
+            queue!(writer, Print("\x1bM")).ok();
+        }
+        queue!(writer, ResetScrollRegion).ok();
+
         let cursor_top = area.top().saturating_sub(1);
         area.y += scroll_amount;
         terminal.set_viewport_area(area);
@@ -59,23 +84,23 @@ pub(crate) fn insert_history_lines(terminal: &mut tui::Tui, lines: Vec<Line>) {
     // ││                            ││
     // │╰────────────────────────────╯│
     // └──────────────────────────────┘
-    queue!(std::io::stdout(), SetScrollRegion(1..area.top())).ok();
+    queue!(writer, SetScrollRegion(1..area.top())).ok();
 
     // NB: we are using MoveTo instead of set_cursor_position here to avoid messing with the
     // terminal's last_known_cursor_position, which hopefully will still be accurate after we
     // fetch/restore the cursor position. insert_history_lines should be cursor-position-neutral :)
-    queue!(std::io::stdout(), MoveTo(0, cursor_top)).ok();
+    queue!(writer, MoveTo(0, cursor_top)).ok();
 
     for line in lines {
-        queue!(std::io::stdout(), Print("\r\n")).ok();
-        write_spans(&mut std::io::stdout(), line.iter()).ok();
+        queue!(writer, Print("\r\n")).ok();
+        write_spans(writer, line.iter()).ok();
     }
 
-    queue!(std::io::stdout(), ResetScrollRegion).ok();
+    queue!(writer, ResetScrollRegion).ok();
 
     // Restore the cursor position to where it was before we started.
     if let Some(cursor_pos) = cursor_pos {
-        queue!(std::io::stdout(), MoveTo(cursor_pos.x, cursor_pos.y)).ok();
+        queue!(writer, MoveTo(cursor_pos.x, cursor_pos.y)).ok();
     }
 }
 
@@ -88,19 +113,25 @@ fn wrapped_line_count(lines: &[Line], width: u16) -> u16 {
 }
 
 fn line_height(line: &Line, width: u16) -> u16 {
-    use unicode_width::UnicodeWidthStr;
-    // get the total display width of the line, accounting for double-width chars
-    let total_width = line
+    // Use the same visible-width slicing semantics as the live row builder so
+    // our pre-scroll estimation matches how rows will actually wrap.
+    let w = width.max(1) as usize;
+    let mut rows = 0u16;
+    let mut remaining = line
         .spans
         .iter()
-        .map(|span| span.content.width())
-        .sum::<usize>();
-    // divide by width to get the number of lines, rounding up
-    if width == 0 {
-        1
-    } else {
-        (total_width as u16).div_ceil(width).max(1)
+        .map(|s| s.content.as_ref())
+        .collect::<Vec<_>>()
+        .join("");
+    while !remaining.is_empty() {
+        let (_prefix, suffix, taken) = crate::live_wrap::take_prefix_by_width(&remaining, w);
+        rows = rows.saturating_add(1);
+        if taken >= remaining.len() {
+            break;
+        }
+        remaining = suffix.to_string();
     }
+    rows.max(1)
 }
 
 #[derive(Debug, Clone, PartialEq, Eq)]
@@ -283,4 +314,12 @@ mod tests {
             String::from_utf8(expected).unwrap()
         );
     }
+
+    #[test]
+    fn line_height_counts_double_width_emoji() {
+        let line = Line::from("😀😀😀"); // each emoji ~ width 2
+        assert_eq!(line_height(&line, 4), 2);
+        assert_eq!(line_height(&line, 2), 3);
+        assert_eq!(line_height(&line, 6), 1);
+    }
 }

codex-rs/tui/src/lib.rs

@@ -3,12 +3,14 @@
 // alternate‑screen mode starts; that file opts‑out locally via `allow`.
 #![deny(clippy::print_stdout, clippy::print_stderr)]
 use app::App;
+use codex_core::BUILT_IN_OSS_MODEL_PROVIDER_ID;
 use codex_core::config::Config;
 use codex_core::config::ConfigOverrides;
 use codex_core::config_types::SandboxMode;
 use codex_core::protocol::AskForApproval;
 use codex_core::util::is_inside_git_repo;
 use codex_login::load_auth;
+use codex_ollama::DEFAULT_OSS_MODEL;
 use log_layer::TuiLogLayer;
 use std::fs::OpenOptions;
 use std::io::Write;
@@ -25,13 +27,14 @@ mod bottom_pane;
 mod chatwidget;
 mod citation_regex;
 mod cli;
-mod custom_terminal;
+pub mod custom_terminal;
 mod exec_command;
 mod file_search;
 mod get_git_diff;
 mod git_warning_screen;
 mod history_cell;
-mod insert_history;
+pub mod insert_history;
+pub mod live_wrap;
 mod log_layer;
 mod markdown;
 mod slash_command;
@@ -69,18 +72,37 @@ pub async fn run_main(
         )
     };
 
+    // When using `--oss`, let the bootstrapper pick the model (defaulting to
+    // gpt-oss:20b) and ensure it is present locally. Also, force the built‑in
+    // `oss` model provider.
+    let model = if let Some(model) = &cli.model {
+        Some(model.clone())
+    } else if cli.oss {
+        Some(DEFAULT_OSS_MODEL.to_owned())
+    } else {
+        None // No model specified, will use the default.
+    };
+
+    let model_provider_override = if cli.oss {
+        Some(BUILT_IN_OSS_MODEL_PROVIDER_ID.to_owned())
+    } else {
+        None
+    };
+
     let config = {
         // Load configuration and support CLI overrides.
         let overrides = ConfigOverrides {
-            model: cli.model.clone(),
+            model,
             approval_policy,
             sandbox_mode,
             cwd: cli.cwd.clone().map(|p| p.canonicalize().unwrap_or(p)),
-            model_provider: None,
+            model_provider: model_provider_override,
             config_profile: cli.config_profile.clone(),
             codex_linux_sandbox_exe,
             base_instructions: None,
             include_plan_tool: Some(true),
+            disable_response_storage: cli.oss.then_some(true),
+            show_raw_agent_reasoning: cli.oss.then_some(true),
         };
         // Parse `-c` overrides from the CLI.
         let cli_kv_overrides = match cli.config_overrides.parse_overrides() {
@@ -135,6 +157,12 @@ pub async fn run_main(
         .with_target(false)
         .with_filter(env_filter());
 
+    if cli.oss {
+        codex_ollama::ensure_oss_ready(&config)
+            .await
+            .map_err(|e| std::io::Error::other(format!("OSS setup failed: {e}")))?;
+    }
+
     // Channel that carries formatted log lines to the UI.
     let (log_tx, log_rx) = tokio::sync::mpsc::unbounded_channel::<String>();
     let tui_layer = TuiLogLayer::new(log_tx.clone(), 120).with_filter(env_filter());

codex-rs/tui/src/live_wrap.rs

@@ -0,0 +1,290 @@
+use unicode_width::UnicodeWidthChar;
+use unicode_width::UnicodeWidthStr;
+
+/// A single visual row produced by RowBuilder.
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct Row {
+    pub text: String,
+    /// True if this row ends with an explicit line break (as opposed to a hard wrap).
+    pub explicit_break: bool,
+}
+
+impl Row {
+    pub fn width(&self) -> usize {
+        self.text.width()
+    }
+}
+
+/// Incrementally wraps input text into visual rows of at most `width` cells.
+///
+/// Step 1: plain-text only. ANSI-carry and styled spans will be added later.
+pub struct RowBuilder {
+    target_width: usize,
+    /// Buffer for the current logical line (until a '\n' is seen).
+    current_line: String,
+    /// Output rows built so far for the current logical line and previous ones.
+    rows: Vec<Row>,
+}
+
+impl RowBuilder {
+    pub fn new(target_width: usize) -> Self {
+        Self {
+            target_width: target_width.max(1),
+            current_line: String::new(),
+            rows: Vec::new(),
+        }
+    }
+
+    pub fn width(&self) -> usize {
+        self.target_width
+    }
+
+    pub fn set_width(&mut self, width: usize) {
+        self.target_width = width.max(1);
+        // Rewrap everything we have (simple approach for Step 1).
+        let mut all = String::new();
+        for row in self.rows.drain(..) {
+            all.push_str(&row.text);
+            if row.explicit_break {
+                all.push('\n');
+            }
+        }
+        all.push_str(&self.current_line);
+        self.current_line.clear();
+        self.push_fragment(&all);
+    }
+
+    /// Push an input fragment. May contain newlines.
+    pub fn push_fragment(&mut self, fragment: &str) {
+        if fragment.is_empty() {
+            return;
+        }
+        let mut start = 0usize;
+        for (i, ch) in fragment.char_indices() {
+            if ch == '\n' {
+                // Flush anything pending before the newline.
+                if start < i {
+                    self.current_line.push_str(&fragment[start..i]);
+                }
+                self.flush_current_line(true);
+                start = i + ch.len_utf8();
+            }
+        }
+        if start < fragment.len() {
+            self.current_line.push_str(&fragment[start..]);
+            self.wrap_current_line();
+        }
+    }
+
+    /// Mark the end of the current logical line (equivalent to pushing a '\n').
+    pub fn end_line(&mut self) {
+        self.flush_current_line(true);
+    }
+
+    /// Drain and return all produced rows.
+    pub fn drain_rows(&mut self) -> Vec<Row> {
+        std::mem::take(&mut self.rows)
+    }
+
+    /// Return a snapshot of produced rows (non-draining).
+    pub fn rows(&self) -> &[Row] {
+        &self.rows
+    }
+
+    /// Rows suitable for display, including the current partial line if any.
+    pub fn display_rows(&self) -> Vec<Row> {
+        let mut out = self.rows.clone();
+        if !self.current_line.is_empty() {
+            out.push(Row {
+                text: self.current_line.clone(),
+                explicit_break: false,
+            });
+        }
+        out
+    }
+
+    /// Drain the oldest rows that exceed `max_keep` display rows (including the
+    /// current partial line, if any). Returns the drained rows in order.
+    pub fn drain_commit_ready(&mut self, max_keep: usize) -> Vec<Row> {
+        let display_count = self.rows.len() + if self.current_line.is_empty() { 0 } else { 1 };
+        if display_count <= max_keep {
+            return Vec::new();
+        }
+        let to_commit = display_count - max_keep;
+        let commit_count = to_commit.min(self.rows.len());
+        let mut drained = Vec::with_capacity(commit_count);
+        for _ in 0..commit_count {
+            drained.push(self.rows.remove(0));
+        }
+        drained
+    }
+
+    fn flush_current_line(&mut self, explicit_break: bool) {
+        // Wrap any remaining content in the current line and then finalize with explicit_break.
+        self.wrap_current_line();
+        // If the current line ended exactly on a width boundary and is non-empty, represent
+        // the explicit break as an empty explicit row so that fragmentation invariance holds.
+        if explicit_break {
+            if self.current_line.is_empty() {
+                // We ended on a boundary previously; add an empty explicit row.
+                self.rows.push(Row {
+                    text: String::new(),
+                    explicit_break: true,
+                });
+            } else {
+                // There is leftover content that did not wrap yet; push it now with the explicit flag.
+                let mut s = String::new();
+                std::mem::swap(&mut s, &mut self.current_line);
+                self.rows.push(Row {
+                    text: s,
+                    explicit_break: true,
+                });
+            }
+        }
+        // Reset current line buffer for next logical line.
+        self.current_line.clear();
+    }
+
+    fn wrap_current_line(&mut self) {
+        // While the current_line exceeds width, cut a prefix.
+        loop {
+            if self.current_line.is_empty() {
+                break;
+            }
+            let (prefix, suffix, taken) =
+                take_prefix_by_width(&self.current_line, self.target_width);
+            if taken == 0 {
+                // Avoid infinite loop on pathological inputs; take one scalar and continue.
+                if let Some((i, ch)) = self.current_line.char_indices().next() {
+                    let len = i + ch.len_utf8();
+                    let p = self.current_line[..len].to_string();
+                    self.rows.push(Row {
+                        text: p,
+                        explicit_break: false,
+                    });
+                    self.current_line = self.current_line[len..].to_string();
+                    continue;
+                }
+                break;
+            }
+            if suffix.is_empty() {
+                // Fits entirely; keep in buffer (do not push yet) so we can append more later.
+                break;
+            } else {
+                // Emit wrapped prefix as a non-explicit row and continue with the remainder.
+                self.rows.push(Row {
+                    text: prefix,
+                    explicit_break: false,
+                });
+                self.current_line = suffix.to_string();
+            }
+        }
+    }
+}
+
+/// Take a prefix of `text` whose visible width is at most `max_cols`.
+/// Returns (prefix, suffix, prefix_width).
+pub fn take_prefix_by_width(text: &str, max_cols: usize) -> (String, &str, usize) {
+    if max_cols == 0 || text.is_empty() {
+        return (String::new(), text, 0);
+    }
+    let mut cols = 0usize;
+    let mut end_idx = 0usize;
+    for (i, ch) in text.char_indices() {
+        let ch_width = UnicodeWidthChar::width(ch).unwrap_or(0);
+        if cols.saturating_add(ch_width) > max_cols {
+            break;
+        }
+        cols += ch_width;
+        end_idx = i + ch.len_utf8();
+        if cols == max_cols {
+            break;
+        }
+    }
+    let prefix = text[..end_idx].to_string();
+    let suffix = &text[end_idx..];
+    (prefix, suffix, cols)
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use pretty_assertions::assert_eq;
+
+    #[test]
+    fn rows_do_not_exceed_width_ascii() {
+        let mut rb = RowBuilder::new(10);
+        rb.push_fragment("hello whirl this is a test");
+        let rows = rb.rows().to_vec();
+        assert_eq!(
+            rows,
+            vec![
+                Row {
+                    text: "hello whir".to_string(),
+                    explicit_break: false
+                },
+                Row {
+                    text: "l this is ".to_string(),
+                    explicit_break: false
+                }
+            ]
+        );
+    }
+
+    #[test]
+    fn rows_do_not_exceed_width_emoji_cjk() {
+        // 😀 is width 2; 你/好 are width 2.
+        let mut rb = RowBuilder::new(6);
+        rb.push_fragment("😀😀 你好");
+        let rows = rb.rows().to_vec();
+        // At width 6, we expect the first row to fit exactly two emojis and a space
+        // (2 + 2 + 1 = 5) plus one more column for the first CJK char (2 would overflow),
+        // so only the two emojis and the space fit; the rest remains buffered.
+        assert_eq!(
+            rows,
+            vec![Row {
+                text: "😀😀 ".to_string(),
+                explicit_break: false
+            }]
+        );
+    }
+
+    #[test]
+    fn fragmentation_invariance_long_token() {
+        let s = "ABCDEFGHIJKLMNOPQRSTUVWXYZ"; // 26 chars
+        let mut rb_all = RowBuilder::new(7);
+        rb_all.push_fragment(s);
+        let all_rows = rb_all.rows().to_vec();
+
+        let mut rb_chunks = RowBuilder::new(7);
+        for i in (0..s.len()).step_by(3) {
+            let end = (i + 3).min(s.len());
+            rb_chunks.push_fragment(&s[i..end]);
+        }
+        let chunk_rows = rb_chunks.rows().to_vec();
+
+        assert_eq!(all_rows, chunk_rows);
+    }
+
+    #[test]
+    fn newline_splits_rows() {
+        let mut rb = RowBuilder::new(10);
+        rb.push_fragment("hello\nworld");
+        let rows = rb.display_rows();
+        assert!(rows.iter().any(|r| r.explicit_break));
+        assert_eq!(rows[0].text, "hello");
+        // Second row should begin with 'world'
+        assert!(rows.iter().any(|r| r.text.starts_with("world")));
+    }
+
+    #[test]
+    fn rewrap_on_width_change() {
+        let mut rb = RowBuilder::new(10);
+        rb.push_fragment("abcdefghijK");
+        assert!(!rb.rows().is_empty());
+        rb.set_width(5);
+        for r in rb.rows() {
+            assert!(r.width() <= 5);
+        }
+    }
+}

codex-rs/tui/src/markdown.rs

@@ -1,3 +1,4 @@
+use crate::citation_regex::CITATION_REGEX;
 use codex_core::config::Config;
 use codex_core::config_types::UriBasedFileOpener;
 use ratatui::text::Line;
@@ -5,8 +6,7 @@ use ratatui::text::Span;
 use std::borrow::Cow;
 use std::path::Path;
 
-use crate::citation_regex::CITATION_REGEX;
-
+#[allow(dead_code)]
 pub(crate) fn append_markdown(
     markdown_source: &str,
     lines: &mut Vec<Line<'static>>,
@@ -15,6 +15,7 @@ pub(crate) fn append_markdown(
     append_markdown_with_opener_and_cwd(markdown_source, lines, config.file_opener, &config.cwd);
 }
 
+#[allow(dead_code)]
 fn append_markdown_with_opener_and_cwd(
     markdown_source: &str,
     lines: &mut Vec<Line<'static>>,
@@ -60,6 +61,7 @@ fn append_markdown_with_opener_and_cwd(
 /// ```text
 /// <scheme>://file<ABS_PATH>:<LINE>
 /// ```
+#[allow(dead_code)]
 fn rewrite_file_citations<'a>(
     src: &'a str,
     file_opener: UriBasedFileOpener,

codex-rs/tui/src/slash_command.rs

@@ -13,8 +13,10 @@ pub enum SlashCommand {
     // DO NOT ALPHA-SORT! Enum order is presentation order in the popup, so
     // more frequently used commands should be listed first.
     New,
+    Init,
     Compact,
     Diff,
+    Status,
     Quit,
     #[cfg(debug_assertions)]
     TestApproval,
@@ -24,12 +26,12 @@ impl SlashCommand {
     /// User-visible description shown in the popup.
     pub fn description(self) -> &'static str {
         match self {
-            SlashCommand::New => "Start a new chat.",
-            SlashCommand::Compact => "Compact the chat history.",
-            SlashCommand::Quit => "Exit the application.",
-            SlashCommand::Diff => {
-                "Show git diff of the working directory (including untracked files)"
-            }
+            SlashCommand::New => "Start a new chat",
+            SlashCommand::Init => "Create an AGENTS.md file with instructions for Codex.",
+            SlashCommand::Compact => "Compact the chat history",
+            SlashCommand::Quit => "Exit the application",
+            SlashCommand::Diff => "Show git diff (including untracked files)",
+            SlashCommand::Status => "Show current session configuration and token usage",
             #[cfg(debug_assertions)]
             SlashCommand::TestApproval => "Test approval request",
         }

codex-rs/tui/src/status_indicator_widget.rs

@@ -9,24 +9,22 @@ use std::thread;
 use std::time::Duration;
 
 use ratatui::buffer::Buffer;
-use ratatui::layout::Alignment;
 use ratatui::layout::Rect;
 use ratatui::style::Color;
 use ratatui::style::Modifier;
 use ratatui::style::Style;
-use ratatui::style::Stylize;
 use ratatui::text::Line;
 use ratatui::text::Span;
-use ratatui::widgets::Block;
-use ratatui::widgets::BorderType;
-use ratatui::widgets::Borders;
-use ratatui::widgets::Padding;
 use ratatui::widgets::Paragraph;
 use ratatui::widgets::WidgetRef;
+use unicode_width::UnicodeWidthStr;
 
 use crate::app_event::AppEvent;
 use crate::app_event_sender::AppEventSender;
 
+// We render the live text using markdown so it visually matches the history
+// cells. Before rendering we strip any ANSI escape sequences to avoid writing
+// raw control bytes into the back buffer.
 use codex_ansi_escape::ansi_escape_line;
 
 pub(crate) struct StatusIndicatorWidget {
@@ -34,6 +32,14 @@ pub(crate) struct StatusIndicatorWidget {
     /// time).
     text: String,
 
+    /// Animation state: reveal target `text` progressively like a typewriter.
+    /// We compute the currently visible prefix length based on the current
+    /// frame index and a constant typing speed.  The `base_frame` and
+    /// `reveal_len_at_base` form the anchor from which we advance.
+    last_target_len: usize,
+    base_frame: usize,
+    reveal_len_at_base: usize,
+
     frame_idx: Arc<AtomicUsize>,
     running: Arc<AtomicBool>,
     // Keep one sender alive to prevent the channel from closing while the
@@ -66,9 +72,13 @@ impl StatusIndicatorWidget {
         }
 
         Self {
-            text: String::from("waiting for logs…"),
+            text: String::from("waiting for model"),
+            last_target_len: 0,
+            base_frame: 0,
+            reveal_len_at_base: 0,
             frame_idx,
             running,
+
             _app_event_tx: app_event_tx,
         }
     }
@@ -79,7 +89,67 @@ impl StatusIndicatorWidget {
 
     /// Update the line that is displayed in the widget.
     pub(crate) fn update_text(&mut self, text: String) {
-        self.text = text.replace(['\n', '\r'], " ");
+        // If the text hasn't changed, don't reset the baseline; let the
+        // animation continue advancing naturally.
+        if text == self.text {
+            return;
+        }
+        // Update the target text, preserving newlines so wrapping matches history cells.
+        // Strip ANSI escapes for the character count so the typewriter animation speed is stable.
+        let stripped = {
+            let line = ansi_escape_line(&text);
+            line.spans
+                .iter()
+                .map(|s| s.content.as_ref())
+                .collect::<Vec<_>>()
+                .join("")
+        };
+        let new_len = stripped.chars().count();
+
+        // Compute how many characters are currently revealed so we can carry
+        // this forward as the new baseline when target text changes.
+        let current_frame = self.frame_idx.load(std::sync::atomic::Ordering::Relaxed);
+        let shown_now = self.current_shown_len(current_frame);
+
+        self.text = text;
+        self.last_target_len = new_len;
+        self.base_frame = current_frame;
+        self.reveal_len_at_base = shown_now.min(new_len);
+    }
+
+    /// Reset the animation and start revealing `text` from the beginning.
+    #[cfg(test)]
+    pub(crate) fn restart_with_text(&mut self, text: String) {
+        let sanitized = text.replace(['\n', '\r'], " ");
+        let stripped = {
+            let line = ansi_escape_line(&sanitized);
+            line.spans
+                .iter()
+                .map(|s| s.content.as_ref())
+                .collect::<Vec<_>>()
+                .join("")
+        };
+
+        let new_len = stripped.chars().count();
+        let current_frame = self.frame_idx.load(std::sync::atomic::Ordering::Relaxed);
+
+        self.text = sanitized;
+        self.last_target_len = new_len;
+        self.base_frame = current_frame;
+        // Start from zero revealed characters for a fresh typewriter cycle.
+        self.reveal_len_at_base = 0;
+    }
+
+    /// Calculate how many characters should currently be visible given the
+    /// animation baseline and frame counter.
+    fn current_shown_len(&self, current_frame: usize) -> usize {
+        // Increase typewriter speed (~5x): reveal more characters per frame.
+        const TYPING_CHARS_PER_FRAME: usize = 7;
+        let frames = current_frame.saturating_sub(self.base_frame);
+        let advanced = self
+            .reveal_len_at_base
+            .saturating_add(frames.saturating_mul(TYPING_CHARS_PER_FRAME));
+        advanced.min(self.last_target_len)
     }
 }
 
@@ -92,26 +162,22 @@ impl Drop for StatusIndicatorWidget {
 
 impl WidgetRef for StatusIndicatorWidget {
     fn render_ref(&self, area: Rect, buf: &mut Buffer) {
-        let widget_style = Style::default();
-        let block = Block::default()
-            .padding(Padding::new(1, 0, 0, 0))
-            .borders(Borders::LEFT)
-            .border_type(BorderType::QuadrantOutside)
-            .border_style(widget_style.dim());
+        // Ensure minimal height
+        if area.height == 0 || area.width == 0 {
+            return;
+        }
+
+        // Build animated gradient header for the word "Working".
         let idx = self.frame_idx.load(std::sync::atomic::Ordering::Relaxed);
         let header_text = "Working";
         let header_chars: Vec<char> = header_text.chars().collect();
-
         let padding = 4usize; // virtual padding around the word for smoother loop
         let period = header_chars.len() + padding * 2;
         let pos = idx % period;
-
         let has_true_color = supports_color::on_cached(supports_color::Stream::Stdout)
             .map(|level| level.has_16m)
             .unwrap_or(false);
-
-        // Width of the bright band (in characters).
-        let band_half_width = 2.0;
+        let band_half_width = 2.0; // width of the bright band in characters
 
         let mut header_spans: Vec<Span<'static>> = Vec::new();
         for (i, ch) in header_chars.iter().enumerate() {
@@ -133,64 +199,46 @@ impl WidgetRef for StatusIndicatorWidget {
                     .fg(Color::Rgb(level, level, level))
                     .add_modifier(Modifier::BOLD)
             } else {
-                // Bold makes dark gray and gray look the same, so don't use it
-                // when true color is not supported.
+                // Bold makes dark gray and gray look the same, so don't use it when true color is not supported.
                 Style::default().fg(color_for_level(level))
             };
 
             header_spans.push(Span::styled(ch.to_string(), style));
         }
 
-        header_spans.push(Span::styled(
+        // Plain rendering: no borders or padding so the live cell is visually indistinguishable from terminal scrollback.
+        let inner_width = area.width as usize;
+
+        // Compose a single status line like: "▌ Working [•] waiting for model"
+        let mut spans: Vec<Span<'static>> = Vec::new();
+        spans.push(Span::styled("▌ ", Style::default().fg(Color::Cyan)));
+        // Gradient header
+        spans.extend(header_spans);
+        // Space after header
+        spans.push(Span::styled(
             " ",
             Style::default()
                 .fg(Color::White)
                 .add_modifier(Modifier::BOLD),
         ));
 
-        // Ensure we do not overflow width.
-        let inner_width = block.inner(area).width as usize;
-
-        // Sanitize and colour‑strip the potentially colourful log text.  This
-        // ensures that **no** raw ANSI escape sequences leak into the
-        // back‑buffer which would otherwise cause cursor jumps or stray
-        // artefacts when the terminal is resized.
-        let line = ansi_escape_line(&self.text);
-        let mut sanitized_tail: String = line
-            .spans
-            .iter()
-            .map(|s| s.content.as_ref())
-            .collect::<Vec<_>>()
-            .join("");
-
-        // Truncate *after* stripping escape codes so width calculation is
-        // accurate. See UTF‑8 boundary comments above.
-        let header_len: usize = header_spans.iter().map(|s| s.content.len()).sum();
-
-        if header_len + sanitized_tail.len() > inner_width {
-            let available_bytes = inner_width.saturating_sub(header_len);
-
-            if sanitized_tail.is_char_boundary(available_bytes) {
-                sanitized_tail.truncate(available_bytes);
+        // Truncate spans to fit the width.
+        let mut acc: Vec<Span<'static>> = Vec::new();
+        let mut used = 0usize;
+        for s in spans {
+            let w = s.content.width();
+            if used + w <= inner_width {
+                acc.push(s);
+                used += w;
             } else {
-                let mut idx = available_bytes;
-                while idx < sanitized_tail.len() && !sanitized_tail.is_char_boundary(idx) {
-                    idx += 1;
-                }
-                sanitized_tail.truncate(idx);
+                break;
             }
         }
+        let lines = vec![Line::from(acc)];
 
-        let mut spans = header_spans;
+        // No-op once full text is revealed; the app no longer reacts to a completion event.
 
-        // Re‑apply the DIM modifier so the tail appears visually subdued
-        // irrespective of the colour information preserved by
-        // `ansi_escape_line`.
-        spans.push(Span::styled(sanitized_tail, Style::default().dim()));
-
-        let paragraph = Paragraph::new(Line::from(spans))
-            .block(block)
-            .alignment(Alignment::Left);
+        let paragraph = Paragraph::new(lines);
         paragraph.render_ref(area, buf);
     }
 }
@@ -204,3 +252,50 @@ fn color_for_level(level: u8) -> Color {
         Color::White
     }
 }
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use crate::app_event::AppEvent;
+    use crate::app_event_sender::AppEventSender;
+    use std::sync::mpsc::channel;
+
+    #[test]
+    fn renders_without_left_border_or_padding() {
+        let (tx_raw, _rx) = channel::<AppEvent>();
+        let tx = AppEventSender::new(tx_raw);
+        let mut w = StatusIndicatorWidget::new(tx);
+        w.restart_with_text("Hello".to_string());
+
+        let area = ratatui::layout::Rect::new(0, 0, 30, 1);
+        // Allow a short delay so the typewriter reveals the first character.
+        std::thread::sleep(std::time::Duration::from_millis(120));
+        let mut buf = ratatui::buffer::Buffer::empty(area);
+        w.render_ref(area, &mut buf);
+
+        // Leftmost column has the left bar
+        let ch0 = buf[(0, 0)].symbol().chars().next().unwrap_or(' ');
+        assert_eq!(ch0, '▌', "expected left bar at col 0: {ch0:?}");
+    }
+
+    #[test]
+    fn working_header_is_present_on_last_line() {
+        let (tx_raw, _rx) = channel::<AppEvent>();
+        let tx = AppEventSender::new(tx_raw);
+        let mut w = StatusIndicatorWidget::new(tx);
+        w.restart_with_text("Hi".to_string());
+        // Ensure some frames elapse so we get a stable state.
+        std::thread::sleep(std::time::Duration::from_millis(120));
+
+        let area = ratatui::layout::Rect::new(0, 0, 30, 1);
+        let mut buf = ratatui::buffer::Buffer::empty(area);
+        w.render_ref(area, &mut buf);
+
+        // Single line; it should contain the animated "Working" header.
+        let mut row = String::new();
+        for x in 0..area.width {
+            row.push(buf[(x, 0)].symbol().chars().next().unwrap_or(' '));
+        }
+        assert!(row.contains("Working"), "expected Working header: {row:?}");
+    }
+}

codex-rs/tui/src/tui.rs

@@ -3,11 +3,14 @@ use std::io::Stdout;
 use std::io::stdout;
 
 use codex_core::config::Config;
+use crossterm::cursor::MoveTo;
 use crossterm::event::DisableBracketedPaste;
 use crossterm::event::EnableBracketedPaste;
 use crossterm::event::KeyboardEnhancementFlags;
 use crossterm::event::PopKeyboardEnhancementFlags;
 use crossterm::event::PushKeyboardEnhancementFlags;
+use crossterm::terminal::Clear;
+use crossterm::terminal::ClearType;
 use ratatui::backend::CrosstermBackend;
 use ratatui::crossterm::execute;
 use ratatui::crossterm::terminal::disable_raw_mode;
@@ -36,6 +39,9 @@ pub fn init(_config: &Config) -> Result<Tui> {
     )?;
     set_panic_hook();
 
+    // Clear screen and move cursor to top-left before drawing UI
+    execute!(stdout(), Clear(ClearType::All), MoveTo(0, 0))?;
+
     let backend = CrosstermBackend::new(stdout());
     let tui = Terminal::with_options(backend)?;
     Ok(tui)

codex-rs/tui/src/user_approval_widget.rs

@@ -47,6 +47,8 @@ pub(crate) enum ApprovalRequest {
 }
 
 /// Options displayed in the *select* mode.
+///
+/// The `key` is matched case-insensitively.
 struct SelectOption {
     label: Line<'static>,
     description: &'static str,
@@ -187,6 +189,16 @@ impl UserApprovalWidget<'_> {
         }
     }
 
+    /// Normalize a key for comparison.
+    /// - For `KeyCode::Char`, converts to lowercase for case-insensitive matching.
+    /// - Other key codes are returned unchanged.
+    fn normalize_keycode(code: KeyCode) -> KeyCode {
+        match code {
+            KeyCode::Char(c) => KeyCode::Char(c.to_ascii_lowercase()),
+            other => other,
+        }
+    }
+
     /// Handle Ctrl-C pressed by the user while the modal is visible.
     /// Behaves like pressing Escape: abort the request and close the modal.
     pub(crate) fn on_ctrl_c(&mut self) {
@@ -210,7 +222,12 @@ impl UserApprovalWidget<'_> {
                 self.send_decision(ReviewDecision::Abort);
             }
             other => {
-                if let Some(opt) = self.select_options.iter().find(|opt| opt.key == other) {
+                let normalized = Self::normalize_keycode(other);
+                if let Some(opt) = self
+                    .select_options
+                    .iter()
+                    .find(|opt| Self::normalize_keycode(opt.key) == normalized)
+                {
                     self.send_decision(opt.decision);
                 }
             }
@@ -330,3 +347,59 @@ impl WidgetRef for &UserApprovalWidget<'_> {
             );
     }
 }
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use crossterm::event::KeyCode;
+    use crossterm::event::KeyEvent;
+    use crossterm::event::KeyModifiers;
+    use std::path::PathBuf;
+    use std::sync::mpsc::channel;
+
+    #[test]
+    fn lowercase_shortcut_is_accepted() {
+        let (tx_raw, rx) = channel::<AppEvent>();
+        let tx = AppEventSender::new(tx_raw);
+        let req = ApprovalRequest::Exec {
+            id: "1".to_string(),
+            command: vec!["echo".to_string()],
+            cwd: PathBuf::new(),
+            reason: None,
+        };
+        let mut widget = UserApprovalWidget::new(req, tx);
+        widget.handle_key_event(KeyEvent::new(KeyCode::Char('y'), KeyModifiers::NONE));
+        assert!(widget.is_complete());
+        let events: Vec<AppEvent> = rx.try_iter().collect();
+        assert!(events.iter().any(|e| matches!(
+            e,
+            AppEvent::CodexOp(Op::ExecApproval {
+                decision: ReviewDecision::Approved,
+                ..
+            })
+        )));
+    }
+
+    #[test]
+    fn uppercase_shortcut_is_accepted() {
+        let (tx_raw, rx) = channel::<AppEvent>();
+        let tx = AppEventSender::new(tx_raw);
+        let req = ApprovalRequest::Exec {
+            id: "2".to_string(),
+            command: vec!["echo".to_string()],
+            cwd: PathBuf::new(),
+            reason: None,
+        };
+        let mut widget = UserApprovalWidget::new(req, tx);
+        widget.handle_key_event(KeyEvent::new(KeyCode::Char('Y'), KeyModifiers::NONE));
+        assert!(widget.is_complete());
+        let events: Vec<AppEvent> = rx.try_iter().collect();
+        assert!(events.iter().any(|e| matches!(
+            e,
+            AppEvent::CodexOp(Op::ExecApproval {
+                decision: ReviewDecision::Approved,
+                ..
+            })
+        )));
+    }
+}

codex-rs/tui/tests/vt100_history.rs

@@ -0,0 +1,214 @@
+#![cfg(feature = "vt100-tests")]
+#![expect(clippy::expect_used)]
+
+use ratatui::backend::TestBackend;
+use ratatui::layout::Rect;
+use ratatui::style::Color;
+use ratatui::style::Style;
+use ratatui::text::Line;
+use ratatui::text::Span;
+
+// Small helper macro to assert a collection contains an item with a clearer
+// failure message.
+macro_rules! assert_contains {
+    ($collection:expr, $item:expr $(,)?) => {
+        assert!(
+            $collection.contains(&$item),
+            "Expected {:?} to contain {:?}",
+            $collection,
+            $item
+        );
+    };
+    ($collection:expr, $item:expr, $($arg:tt)+) => {
+        assert!($collection.contains(&$item), $($arg)+);
+    };
+}
+
+struct TestScenario {
+    width: u16,
+    height: u16,
+    term: codex_tui::custom_terminal::Terminal<TestBackend>,
+}
+
+impl TestScenario {
+    fn new(width: u16, height: u16, viewport: Rect) -> Self {
+        let backend = TestBackend::new(width, height);
+        let mut term = codex_tui::custom_terminal::Terminal::with_options(backend)
+            .expect("failed to construct terminal");
+        term.set_viewport_area(viewport);
+        Self {
+            width,
+            height,
+            term,
+        }
+    }
+
+    fn run_insert(&mut self, lines: Vec<Line<'static>>) -> Vec<u8> {
+        let mut buf: Vec<u8> = Vec::new();
+        codex_tui::insert_history::insert_history_lines_to_writer(&mut self.term, &mut buf, lines);
+        buf
+    }
+
+    fn screen_rows_from_bytes(&self, bytes: &[u8]) -> Vec<String> {
+        let mut parser = vt100::Parser::new(self.height, self.width, 0);
+        parser.process(bytes);
+        let screen = parser.screen();
+
+        let mut rows: Vec<String> = Vec::with_capacity(self.height as usize);
+        for row in 0..self.height {
+            let mut s = String::with_capacity(self.width as usize);
+            for col in 0..self.width {
+                if let Some(cell) = screen.cell(row, col) {
+                    if let Some(ch) = cell.contents().chars().next() {
+                        s.push(ch);
+                    } else {
+                        s.push(' ');
+                    }
+                } else {
+                    s.push(' ');
+                }
+            }
+            rows.push(s.trim_end().to_string());
+        }
+        rows
+    }
+}
+
+#[test]
+fn hist_001_basic_insertion_no_wrap() {
+    // Screen of 20x6; viewport is the last row (height=1 at y=5)
+    let area = Rect::new(0, 5, 20, 1);
+    let mut scenario = TestScenario::new(20, 6, area);
+
+    let lines = vec![Line::from("first"), Line::from("second")];
+    let buf = scenario.run_insert(lines);
+    let rows = scenario.screen_rows_from_bytes(&buf);
+    assert_contains!(rows, String::from("first"));
+    assert_contains!(rows, String::from("second"));
+    let first_idx = rows
+        .iter()
+        .position(|r| r == "first")
+        .expect("expected 'first' row to be present");
+    let second_idx = rows
+        .iter()
+        .position(|r| r == "second")
+        .expect("expected 'second' row to be present");
+    assert_eq!(second_idx, first_idx + 1, "rows should be adjacent");
+}
+
+#[test]
+fn hist_002_long_token_wraps() {
+    let area = Rect::new(0, 5, 20, 1);
+    let mut scenario = TestScenario::new(20, 6, area);
+
+    let long = "A".repeat(45); // > 2 lines at width 20
+    let lines = vec![Line::from(long.clone())];
+    let buf = scenario.run_insert(lines);
+    let mut parser = vt100::Parser::new(6, 20, 0);
+    parser.process(&buf);
+    let screen = parser.screen();
+
+    // Count total A's on the screen
+    let mut count_a = 0usize;
+    for row in 0..6 {
+        for col in 0..20 {
+            if let Some(cell) = screen.cell(row, col) {
+                if let Some(ch) = cell.contents().chars().next() {
+                    if ch == 'A' {
+                        count_a += 1;
+                    }
+                }
+            }
+        }
+    }
+
+    assert_eq!(
+        count_a,
+        long.len(),
+        "wrapped content did not preserve all characters"
+    );
+}
+
+#[test]
+fn hist_003_emoji_and_cjk() {
+    let area = Rect::new(0, 5, 20, 1);
+    let mut scenario = TestScenario::new(20, 6, area);
+
+    let text = String::from("😀😀😀😀😀 你好世界");
+    let lines = vec![Line::from(text.clone())];
+    let buf = scenario.run_insert(lines);
+    let rows = scenario.screen_rows_from_bytes(&buf);
+    let reconstructed: String = rows.join("").chars().filter(|c| *c != ' ').collect();
+    for ch in text.chars().filter(|c| !c.is_whitespace()) {
+        assert!(
+            reconstructed.contains(ch),
+            "missing character {ch:?} in reconstructed screen"
+        );
+    }
+}
+
+#[test]
+fn hist_004_mixed_ansi_spans() {
+    let area = Rect::new(0, 5, 20, 1);
+    let mut scenario = TestScenario::new(20, 6, area);
+
+    let line = Line::from(vec![
+        Span::styled("red", Style::default().fg(Color::Red)),
+        Span::raw("+plain"),
+    ]);
+    let buf = scenario.run_insert(vec![line]);
+    let rows = scenario.screen_rows_from_bytes(&buf);
+    assert_contains!(rows, String::from("red+plain"));
+}
+
+#[test]
+fn hist_006_cursor_restoration() {
+    let area = Rect::new(0, 5, 20, 1);
+    let mut scenario = TestScenario::new(20, 6, area);
+
+    let lines = vec![Line::from("x")];
+    let buf = scenario.run_insert(lines);
+    let s = String::from_utf8_lossy(&buf);
+    // CUP to 1;1 (ANSI: ESC[1;1H)
+    assert!(
+        s.contains("\u{1b}[1;1H"),
+        "expected final CUP to 1;1 in output, got: {s:?}"
+    );
+    // Reset scroll region
+    assert!(
+        s.contains("\u{1b}[r"),
+        "expected reset scroll region in output, got: {s:?}"
+    );
+}
+
+#[test]
+fn hist_005_pre_scroll_region_down() {
+    // Viewport not at bottom: y=3 (0-based), height=1
+    let area = Rect::new(0, 3, 20, 1);
+    let mut scenario = TestScenario::new(20, 6, area);
+
+    let lines = vec![Line::from("first"), Line::from("second")];
+    let buf = scenario.run_insert(lines);
+    let s = String::from_utf8_lossy(&buf);
+    // Expect we limited scroll region to [top+1 .. screen_height] => [4 .. 6] (1-based)
+    assert!(
+        s.contains("\u{1b}[4;6r"),
+        "expected pre-scroll SetScrollRegion 4..6, got: {s:?}"
+    );
+    // Expect we moved cursor to top of that region: row 3 (0-based) => CUP 4;1H
+    assert!(
+        s.contains("\u{1b}[4;1H"),
+        "expected cursor at top of pre-scroll region, got: {s:?}"
+    );
+    // Expect at least two Reverse Index commands (ESC M) for two inserted lines
+    let ri_count = s.matches("\u{1b}M").count();
+    assert!(
+        ri_count >= 1,
+        "expected at least one RI (ESC M), got: {s:?}"
+    );
+    // After pre-scroll, we set insertion scroll region to [1 .. new_top] => [1 .. 5]
+    assert!(
+        s.contains("\u{1b}[1;5r"),
+        "expected insertion SetScrollRegion 1..5, got: {s:?}"
+    );
+}

codex-rs/tui/tests/vt100_live_commit.rs

@@ -0,0 +1,101 @@
+#![cfg(feature = "vt100-tests")]
+
+use ratatui::backend::TestBackend;
+use ratatui::layout::Rect;
+use ratatui::text::Line;
+
+#[test]
+fn live_001_commit_on_overflow() {
+    let backend = TestBackend::new(20, 6);
+    let mut term = match codex_tui::custom_terminal::Terminal::with_options(backend) {
+        Ok(t) => t,
+        Err(e) => panic!("failed to construct terminal: {e}"),
+    };
+    let area = Rect::new(0, 5, 20, 1);
+    term.set_viewport_area(area);
+
+    // Build 5 explicit rows at width 20.
+    let mut rb = codex_tui::live_wrap::RowBuilder::new(20);
+    rb.push_fragment("one\n");
+    rb.push_fragment("two\n");
+    rb.push_fragment("three\n");
+    rb.push_fragment("four\n");
+    rb.push_fragment("five\n");
+
+    // Keep the last 3 in the live ring; commit the first 2.
+    let commit_rows = rb.drain_commit_ready(3);
+    let lines: Vec<Line<'static>> = commit_rows
+        .into_iter()
+        .map(|r| Line::from(r.text))
+        .collect();
+
+    let mut buf: Vec<u8> = Vec::new();
+    codex_tui::insert_history::insert_history_lines_to_writer(&mut term, &mut buf, lines);
+
+    let mut parser = vt100::Parser::new(6, 20, 0);
+    parser.process(&buf);
+    let screen = parser.screen();
+
+    // The words "one" and "two" should appear above the viewport.
+    let mut joined = String::new();
+    for row in 0..6 {
+        for col in 0..20 {
+            if let Some(cell) = screen.cell(row, col) {
+                if let Some(ch) = cell.contents().chars().next() {
+                    joined.push(ch);
+                } else {
+                    joined.push(' ');
+                }
+            }
+        }
+        joined.push('\n');
+    }
+    assert!(
+        joined.contains("one"),
+        "expected committed 'one' to be visible\n{joined}"
+    );
+    assert!(
+        joined.contains("two"),
+        "expected committed 'two' to be visible\n{joined}"
+    );
+    // The last three (three,four,five) remain in the live ring, not committed here.
+}
+
+#[test]
+fn live_002_pre_scroll_and_commit() {
+    let backend = TestBackend::new(20, 6);
+    let mut term = match codex_tui::custom_terminal::Terminal::with_options(backend) {
+        Ok(t) => t,
+        Err(e) => panic!("failed to construct terminal: {e}"),
+    };
+    // Viewport not at bottom: y=3
+    let area = Rect::new(0, 3, 20, 1);
+    term.set_viewport_area(area);
+
+    let mut rb = codex_tui::live_wrap::RowBuilder::new(20);
+    rb.push_fragment("alpha\n");
+    rb.push_fragment("beta\n");
+    rb.push_fragment("gamma\n");
+    rb.push_fragment("delta\n");
+
+    // Keep 3, commit 1.
+    let commit_rows = rb.drain_commit_ready(3);
+    let lines: Vec<Line<'static>> = commit_rows
+        .into_iter()
+        .map(|r| Line::from(r.text))
+        .collect();
+
+    let mut buf: Vec<u8> = Vec::new();
+    codex_tui::insert_history::insert_history_lines_to_writer(&mut term, &mut buf, lines);
+    let s = String::from_utf8_lossy(&buf);
+
+    // Expect a SetScrollRegion to [area.top()+1 .. screen_height] and a cursor move to top of that region.
+    assert!(
+        s.contains("\u{1b}[4;6r"),
+        "expected pre-scroll region 4..6, got: {s:?}"
+    );
+    assert!(
+        s.contains("\u{1b}[4;1H"),
+        "expected cursor CUP 4;1H, got: {s:?}"
+    );
+}