916 Commits

• Fix unified_exec on windows (#7620)

  Fix unified_exec on windows
  
  Requires removal of PSUEDOCONSOLE_INHERIT_CURSOR flag so child processed
  don't attempt to wait for cursor position response (and timeout).
  
  
  https://github.com/wezterm/wezterm/compare/main...pakrym:wezterm:PSUEDOCONSOLE_INHERIT_CURSOR?expand=1
  
  ---------
  
  Co-authored-by: pakrym-oai <pakrym@openai.com>

  Pavel Krymets · 2025-12-05 20:09:43 +00:00
  f48d88067e
• feat(core) Add login to shell_command tool (#6846)

  ## Summary
  Adds the `login` parameter to the `shell_command` tool - optional,
  defaults to true.
  
  ## Testing
  - [x] Tested locally

  Dylan Hurd · 2025-12-05 11:03:25 -08:00
  a8cbbdbc6e
• Wire with_remote_overrides to construct model families (#7621)

  - This PR wires `with_remote_overrides` and make the
  `construct_model_families` an async function
  - Moves getting model family a level above to keep the function `sync`
  - Updates the tests to local, offline, and `sync` helper for model
  families

  Ahmed Ibrahim · 2025-12-05 10:40:15 -08:00
  d08efb1743
• fix: ignore ghost snapshots in token consumption (#7638)
  jif-oai · 2025-12-05 13:57:24 +00:00
  e91bb6b947
• fix: taking plan type from usage endpoint instead of thru auth token (#7610)

  pull plan type from the usage endpoint, persist it in session state /
  tui state, and propagate through rate limit snapshots

  zhao-oai · 2025-12-04 23:34:13 -08:00
  b8eab7ce90
• feat: exec policy integration in shell mcp (#7609)

  adding execpolicy support into the `posix` mcp
  
  Co-authored-by: Michael Bolin <mbolin@openai.com>

  zhao-oai · 2025-12-04 21:55:54 -08:00
  b1c918d8f7
• Call models endpoint in models manager (#7616)

  - Introduce `with_remote_overrides` and update
  `refresh_available_models`
  - Put `auth_manager` instead of `auth_mode` on `models_manager`
  - Remove `ShellType` and `ReasoningLevel` to use already existing
  structs

  Ahmed Ibrahim · 2025-12-04 18:28:03 -08:00
  7b359c9c8e
• chore: refactor to move Arc<RwLock> concern outside exec_policy_for (#7615)

  The caller should decide whether wrapping the policy in `Arc<RwLock>` is
  necessary. This should make https://github.com/openai/codex/pull/7609 a
  bit smoother.
  
  - `exec_policy_for()` -> `load_exec_policy_for_features()`
  - introduce `load_exec_policy()` that does not take `Features` as an arg
  - both return `Result<Policy, ExecPolicyError>` instead of
  Result<Arc<RwLock<Policy>>, ExecPolicyError>`
  
  This simplifies the tests as they have no need for `Arc<RwLock>`.

  Michael Bolin · 2025-12-04 15:13:27 -08:00
  0972cd9404
• Add models endpoint (#7603)

  - Use the codex-api crate to introduce models endpoint. 
  - Add `models` to codex core tests helpers
  - Add `ModelsInfo` for the endpoint return type

  Ahmed Ibrahim · 2025-12-04 12:57:54 -08:00
  903b7774bc
• Inline response recording and remove process_items indirection (#7310)

  - Inline response recording during streaming: `run_turn` now records
  items as they arrive instead of building a `ProcessedResponseItem` list
  and post‑processing via `process_items`.
  - Simplify turn handling: `handle_output_item_done` returns the
  follow‑up signal + optional tool future; `needs_follow_up` is set only
  there, and in‑flight tool futures are drained once at the end (errors
  logged, no extra state writes).
  - Flattened stream loop: removed `process_items` indirection and the
  extra output queue
  - - Tests: relaxed `tool_parallelism::tool_results_grouped` to allow any
  completion order while still requiring matching call/output IDs.

  Ahmed Ibrahim · 2025-12-04 12:17:54 -08:00
  6e6338aa87
• remove model_family from `config (#7571)

  - Remove `model_family` from `config`
  - Make sure to still override config elements related to `model_family`
  like supporting reasoning

  Ahmed Ibrahim · 2025-12-04 11:57:58 -08:00
  9b2055586d
• chore(core): test apply_patch_cli on Windows (#7554)

  ## Summary
  These tests pass on windows, let's enable them.
  
  ## Testing
  - [x] These are more tests

  Dylan Hurd · 2025-12-04 10:39:45 -08:00
  37c36024c7
• chore: review in read-only (#7593)
  jif-oai · 2025-12-04 10:01:12 -08:00
  291b54a762
• feat: update sandbox policy to allow TTY (#7580)

  **Change**: Seatbelt now allows file-ioctl on /dev/ttys[0-9]+ even
  without the sandbox extension so pre-created PTYs remain interactive
  (Python REPL, shells).
  
  **Risk**: A seatbelted process that already holds a PTY fd (including
  one it shouldn’t) could issue tty ioctls like TIOCSTI or termios changes
  on that fd. This doesn’t allow opening new PTYs or reading/writing them;
  it only broadens ioctl capability on existing fds.
  
  **Why acceptable**: We already hand the child its PTY for interactive
  use; restoring ioctls is required for isatty() and prompts to work. The
  attack requires being given or inheriting a sensitive PTY fd; by design
  we don’t hand untrusted processes other users’ PTYs (we don't hand them
  any PTYs actually), so the practical exposure is limited to the PTY
  intentionally allocated for the session.
  
  **Validation**:
  Running
  ```
  start a python interpreter and keep it running
  ```
  Followed by:
  * `calculate 1+1 using it` -> works as expected
  * `Use this Python session to run the command just fix in
  /Users/jif/code/codex/codex-rs` -> does not work as expected

  jif-oai · 2025-12-04 17:58:58 +00:00
  2b5d0b2935
• fix: release session ID when not used (#7592)
  jif-oai · 2025-12-04 17:42:16 +00:00
  36edb412b1
• chore: default warning messages to true (#7588)
  jif-oai · 2025-12-04 17:29:23 +00:00
  1b2509f05a
• Use shared check sandboxing (#7547)
  pakrym-oai · 2025-12-04 08:34:09 -08:00
  f1b7cdc3bd
• Slightly better status display for unified exec (#7563)

  Trim bash -lc

  pakrym-oai · 2025-12-04 08:32:54 -08:00
  c4e18f1b63
• Refactor execpolicy fallback evaluation (#7544)

  ## Refactor of the `execpolicy` crate
  
  To illustrate why we need this refactor, consider an agent attempting to
  run `apple | rm -rf ./`. Suppose `apple` is allowed by `execpolicy`.
  Before this PR, `execpolicy` would consider `apple` and `pear` and only
  render one rule match: `Allow`. We would skip any heuristics checks on
  `rm -rf ./` and immediately approve `apple | rm -rf ./` to run.
  
  To fix this, we now thread a `fallback` evaluation function into
  `execpolicy` that runs when no `execpolicy` rules match a given command.
  In our example, we would run `fallback` on `rm -rf ./` and prevent
  `apple | rm -rf ./` from being run without approval.

  zhao-oai · 2025-12-03 23:39:48 -08:00
  3d35cb4619
• whitelist command prefix integration in core and tui (#7033)

  this PR enables TUI to approve commands and add their prefixes to an
  allowlist:
  <img width="708" height="605" alt="Screenshot 2025-11-21 at 4 18 07 PM"
  src="https://github.com/user-attachments/assets/56a19893-4553-4770-a881-becf79eeda32"
  />
  
  note: we only show the option to whitelist the command when 
  1) command is not multi-part (e.g `git add -A && git commit -m 'hello
  world'`)
  2) command is not already matched by an existing rule

  zhao-oai · 2025-12-03 23:17:02 -08:00
  e925a380dc
• Migrate codex max (#7566)

  - make codex max the default
  - fix: we were doing some async work in sync function which caused tui
  to panic

  Ahmed Ibrahim · 2025-12-03 20:54:48 -08:00
  67e67e054f
• Migrate model family to models manager (#7565)

  This PR moves `ModelsFamily` to `openai_models`. It also propagates
  `ModelsManager` to session services and use it to drive model family. We
  also make `derive_default_model_family` private because it's a step
  towards what we want: one place that gives model configuration.
  
  This is a second step at having one source of truth for models
  information and config: `ModelsManager`.
  
  Next steps would be to remove `ModelsFamily` from config. That's massive
  because it's being used in 41 occasions mostly pre launching `codex`.
  Also, we need to make `find_family_for_model` private. It's also big
  because it's being used in 21 occasions ~ all tests.

  Ahmed Ibrahim · 2025-12-03 18:49:47 -08:00
  cee37a32b2
• Migrate tui to use models manager (#7555)

  - This PR treats the `ModelsManager` like `AuthManager` and propagate it
  into the tui, replacing the `builtin_model_presets`
  - We are also decreasing the visibility of `builtin_model_presets`
  
  based on https://github.com/openai/codex/pull/7552

  Ahmed Ibrahim · 2025-12-03 18:00:47 -08:00
  8da91d1c89
• Introduce ModelsManager and migrate app-server to use it. (#7552)
  Ahmed Ibrahim · 2025-12-03 17:17:56 -08:00
  00cc00ead8
• fix: Features should be immutable over the lifetime of a session/thread (#7540)

  I noticed that `features: Features` was defined on `struct
  SessionConfiguration`, which is commonly owned by `SessionState`, which
  is in turn owned by `Session`.
  
  Though I do not believe that `Features` should be allowed to be modified
  over the course of a session (if the feature state is not invariant, it
  makes it harder to reason about), which argues that it should live on
  `Session` rather than `SessionState` or `SessionConfiguration`.
  
  This PR moves `Features` to `Session` and updates all call sites. It
  appears the only place we were mutating `Features` was:
  
  - in tests
  - the sub-agent config for a review task:
  
  
  https://github.com/openai/codex/blob/3ef76ff29d5eed258fb6b8550e0e2b973d0dca21/codex-rs/core/src/tasks/review.rs#L86-L89
  
  Note this change also means it is no longer an `async` call to check the
  state of a feature, eliminating the possibility of a
  [TOCTTOU](https://en.wikipedia.org/wiki/Time-of-check_to_time-of-use)
  error between checking the state of a feature and acting on it:
  
  
  https://github.com/openai/codex/blob/3ef76ff29d5eed258fb6b8550e0e2b973d0dca21/codex-rs/core/src/codex.rs#L1069-L1076

  Michael Bolin · 2025-12-03 16:12:31 -08:00
  1cfc967eb8
• feat: Support listing and selecting skills via $ or /skills (#7506)

  List/Select skills with $-mention or /skills

  xl-openai · 2025-12-03 15:12:46 -08:00
  9a50a04400
• Migrate model preset (#7542)

  - Introduce `openai_models` in `/core`
  - Move `PRESETS` under it
  - Move `ModelPreset`, `ModelUpgrade`, `ReasoningEffortPreset`,
  `ReasoningEffortPreset`, and `ReasoningEffortPreset` to `protocol`
  - Introduce `Op::ListModels` and `EventMsg::AvailableModels`
  
  Next steps:
  - migrate `app-server` and `tui` to use the introduced Operation

  Ahmed Ibrahim · 2025-12-03 20:30:43 +00:00
  71504325d3
• chore: update unified exec sandboxing detection (#7541)

  No integration test for now because it would make them flaky. Tracking
  it in my todos to add some once we have a clock based system for
  integration tests

  jif-oai · 2025-12-03 20:06:47 +00:00
  8d0f023fa9
• feat: support list mcp servers in app server (#7505)

  ### Summary
  Added `mcp/servers/list` which is equivalent to `/mcp` slash command in
  CLI for response. This will be used in VSCE MCP settings to show log in
  status, available tools etc.

  Shijie Rao · 2025-12-03 09:51:46 -08:00
  4785344c9c
• seatbelt: allow openpty() (#7507)

  This allows `openpty(3)` to run in the default sandbox. Also permit
  reading `kern.argmax`, which is the maximum number of arguments to
  exec().

  Jeremy Rose · 2025-12-03 09:15:38 -08:00
  9b3251f28f
• feat: codex tool tips (#7440)

  <img width="551" height="316" alt="Screenshot 2025-12-01 at 12 22 26"
  src="https://github.com/user-attachments/assets/6ca3deff-8ef8-4f74-a8e1-e5ea13fd6740"
  />

  jif-oai · 2025-12-03 16:29:13 +00:00
  45f3250eec
• feat: retroactive image placeholder to prevent poisoning (#6774)

  If an image can't be read by the API, it will poison the entire history,
  preventing any new turn on the conversation.
  This detect such cases and replace the image by a placeholder

  jif-oai · 2025-12-03 11:35:56 +00:00
  51307eaf07
• feat: model warning in case of apply patch (#7494)
  jif-oai · 2025-12-03 09:07:31 +00:00
  42ae738f67
• fix(unified_exec): use platform default shell when unified_exec shell… (#7486)

  # Unified Exec Shell Selection on Windows
  
  ## Problem
  
  reference issue #7466
  
  The `unified_exec` handler currently deserializes model-provided tool
  calls into the `ExecCommandArgs` struct:
  
  ```rust
  #[derive(Debug, Deserialize)]
  struct ExecCommandArgs {
      cmd: String,
      #[serde(default)]
      workdir: Option<String>,
      #[serde(default = "default_shell")]
      shell: String,
      #[serde(default = "default_login")]
      login: bool,
      #[serde(default = "default_exec_yield_time_ms")]
      yield_time_ms: u64,
      #[serde(default)]
      max_output_tokens: Option<usize>,
      #[serde(default)]
      with_escalated_permissions: Option<bool>,
      #[serde(default)]
      justification: Option<String>,
  }
  ```
  
  The `shell` field uses a hard-coded default:
  
  ```rust
  fn default_shell() -> String {
      "/bin/bash".to_string()
  }
  ```
  
  When the model returns a tool call JSON that only contains `cmd` (which
  is the common case), Serde fills in `shell` with this default value.
  Later, `get_command` uses that value as if it were a model-provided
  shell path:
  
  ```rust
  fn get_command(args: &ExecCommandArgs) -> Vec<String> {
      let shell = get_shell_by_model_provided_path(&PathBuf::from(args.shell.clone()));
      shell.derive_exec_args(&args.cmd, args.login)
  }
  ```
  
  On Unix, this usually resolves to `/bin/bash` and works as expected.
  However, on Windows this behavior is problematic:
  
  - The hard-coded `"/bin/bash"` is not a valid Windows path.
  - `get_shell_by_model_provided_path` treats this as a model-specified
  shell, and tries to resolve it (e.g. via `which::which("bash")`), which
  may or may not exist and may not behave as intended.
  - In practice, this leads to commands being executed under a non-default
  or non-existent shell on Windows (for example, WSL bash), instead of the
  expected Windows PowerShell or `cmd.exe`.
  
  The core of the issue is that **"model did not specify `shell`" is
  currently interpreted as "the model explicitly requested `/bin/bash`"**,
  which is both Unix-specific and wrong on Windows.
  
  ## Proposed Solution
  
  Instead of hard-coding `"/bin/bash"` into `ExecCommandArgs`, we should
  distinguish between:
  
  1. **The model explicitly specifying a shell**, e.g.:
  
     ```json
     {
       "cmd": "echo hello",
       "shell": "pwsh"
     }
     ```
  
  In this case, we *do* want to respect the model’s choice and use
  `get_shell_by_model_provided_path`.
  
  2. **The model omitting the `shell` field entirely**, e.g.:
  
     ```json
     {
       "cmd": "echo hello"
     }
     ```
  
  In this case, we should *not* assume `/bin/bash`. Instead, we should use
  `default_user_shell()` and let the platform decide.
  
  To express this distinction, we can:
  
  1. Change `shell` to be optional in `ExecCommandArgs`:
  
     ```rust
     #[derive(Debug, Deserialize)]
     struct ExecCommandArgs {
         cmd: String,
         #[serde(default)]
         workdir: Option<String>,
         #[serde(default)]
         shell: Option<String>,
         #[serde(default = "default_login")]
         login: bool,
         #[serde(default = "default_exec_yield_time_ms")]
         yield_time_ms: u64,
         #[serde(default)]
         max_output_tokens: Option<usize>,
         #[serde(default)]
         with_escalated_permissions: Option<bool>,
         #[serde(default)]
         justification: Option<String>,
     }
     ```
  
  Here, the absence of `shell` in the JSON is represented as `shell:
  None`, rather than a hard-coded string value.

  Robby He · 2025-12-02 21:49:25 -08:00
  f3989f6092
• fix: inline function marked as dead code (#7508)

  I was debugging something else and noticed we could eliminate an
  instance of `#[allow(dead_code)]` pretty easily.

  Michael Bolin · 2025-12-03 00:50:34 +00:00
  06e7667d0e
• improve resume performance (#7303)

  Reading the tail can be costly if we have a very big rollout item. we
  can just read the file metadata

  Ahmed Ibrahim · 2025-12-02 16:39:40 -08:00
  1ef1fe67ec
• chore: make create_approval_requirement_for_command an async fn (#7501)

  I think this might help with https://github.com/openai/codex/pull/7033
  because `create_approval_requirement_for_command()` will soon need
  access to `Session.state`, which is a `tokio::sync::Mutex` that needs to
  be accessed via `async`.

  Michael Bolin · 2025-12-02 15:01:15 -08:00
  ec93b6daf3
• Trim history.jsonl when history.max_bytes is set (#6242)

  This PR honors the `history.max_bytes` configuration parameter by
  trimming `history.jsonl` whenever it grows past the configured limit.
  While appending new entries we retain the newest record, drop the oldest
  lines to stay within the byte budget, and serialize the compacted file
  back to disk under the same lock to keep writers safe.

  liam · 2025-12-02 14:01:05 -08:00
  4d4778ec1c
• persisting credits if new snapshot does not contain credit info (#7490)

  in response to incoming changes to responses headers where the header
  may sometimes not contain credits info (no longer forcing a credit
  check)

  zhao-oai · 2025-12-02 16:23:24 -05:00
  5ebdc9af1b
• fix: drop lock once it is no longer needed (#7500)

  I noticed this while doing a post-commit review of https://github.com/openai/codex/pull/7467.

  Michael Bolin · 2025-12-02 20:46:26 +00:00
  f6a7da4ac3
• Show token used when context window is unknown (#7497)

  - Show context window usage in tokens instead of percentage when the
  window length is unknown.

  Ahmed Ibrahim · 2025-12-02 11:45:50 -08:00
  127e307f89
• Use non-blocking mutex (#7467)
  Ahmed Ibrahim · 2025-12-02 10:50:46 -08:00
  21ad1c1c90
• feat: intercept apply_patch for unified_exec (#7446)
  jif-oai · 2025-12-02 17:54:02 +00:00
  72b95db12f
• feat: add warning message for the model (#7445)

  Add a warning message as a user turn to the model if the model does not
  behave as expected (here, for example, if the model opens too many
  `unified_exec` sessions)

  jif-oai · 2025-12-02 11:56:00 +00:00
  9ee855ec57
• chore: review everywhere (#7444)
  jif-oai · 2025-12-02 11:26:27 +00:00
  4b78e2ab09
• feat: experimental support for skills.md (#7412)

  This change prototypes support for Skills with the CLI. This is an
  **experimental** feature for internal testing.
  
  ---------
  
  Co-authored-by: Gav Verma <gverma@openai.com>

  Thibault Sottiaux · 2025-12-01 20:22:35 -08:00
  a8d5ad37b8
• fix(core): enable history lookup on windows (#7457)

  - Add portable history log id helper to support inode-like tracking on
  Unix and creation time on Windows
  - Refactor history metadata and lookup to share code paths and allow
  nonzero log ids across platforms
  - Add coverage for lookup stability after appends

  Steve Mostovoy · 2025-12-01 16:29:01 -08:00
  f443555728
• fix(apply_patch) tests for shell_command (#7307)

  ## Summary
  Adds test coverage for invocations of apply_patch via shell_command with
  heredoc, to validate behavior.
  
  ## Testing
  - [x] These are tests

  Dylan Hurd · 2025-12-01 15:09:22 -08:00
  5b25915d7e
• Fix recent_commits(limit=0) returning 1 commit instead of 0 (#7334)

  Fixes #7333
  
  This is a small bug fix.
  
  This PR fixes an inconsistency in `recent_commits` where `limit == 0`
  still returns 1 commit due to the use of `limit.max(1)` when
  constructing the `git log -n` argument.
  
  Expected behavior: requesting 0 commits should return an empty list.
  
  This PR:
  - returns an empty `Vec` when `limit == 0`
  - adds a test for `recent_commits(limit == 0)` that fails before the
  change and passes afterwards
  - maintains existing behavior for `limit > 0`
  
  This aligns behavior with API expectations and avoids downstream
  consumers misinterpreting the repository as having commit history when
  `limit == 0` is used to explicitly request none.
  
  Happy to adjust if the current behavior is intentional.

  Ali Towaiji · 2025-12-01 10:14:36 -08:00
  0cc3b50228
• fix: disable review rollout filtering (#7371)
  jif-oai · 2025-12-01 09:04:13 +00:00
  a421eba31f