…d two-pass Codex search strategy with deterministic fallback behavior,
and remove an obsolete prompt file that was no longer used.
### Changes
- Updated `workflows/issue-deduplicator.yml`:
- Added richer issue input fields (`state`, `updatedAt`, `labels`) for
model context.
- Added two candidate pools:
- `codex-existing-issues-all.json` (`--state all`)
- `codex-existing-issues-open.json` (`--state open`)
- Added body truncation during JSON preparation to reduce prompt noise.
- Added **Pass 1** Codex run over all issues.
- Added normalization/validation step for Pass 1 output:
- tolerant JSON parsing
- self-issue filtering
- deduplication
- cap to 5 results
- Added **Pass 2 fallback** Codex run over open issues only, triggered
only when Pass 1 has no usable matches.
- Added normalization/validation step for Pass 2 output (same
filtering/dedup/cap behavior).
- Added final deterministic selector:
- prefer pass 2 if it finds matches
- otherwise use pass 1
- otherwise return no matches
- Added observability logs:
- pool sizes
- per-pass parse/match status
- final pass selected and final duplicate count
- Kept public issue-comment format unchanged.
- Added comment documenting that prompt text now lives inline in
workflow.
- Deleted obsolete file:
- `/prompts/issue-deduplicator.txt`
### Behavior Impact
- Better duplicate recall when broad search fails by retrying against
active issues only.
- More deterministic/noise-resistant output handling.
- No change to workflow trigger conditions, permissions, or issue
comment structure.
Currently, if there are syntax errors detected in the starlark rules
file, the entire policy is silently ignored by the CLI. The app server
correctly emits a message that can be displayed in a GUI.
This PR changes the CLI (both the TUI and non-interactive exec) to fail
when the rules file can't be parsed. It then prints out an error message
and exits with a non-zero exit code. This is consistent with the
handling of errors in the config file.
This addresses #11603
We've had a few cases recently where someone enabled a feature flag for
a feature that's still under development or experimental. This test
should prevent this.
I don't think this policy change increases the risk, other than
potentially exposing the caller to bugs in these kernel calls, which are
unlikely.
Without this change, some tools are silently failing or making incorrect
decisions about the processor type (e.g. installing x86 binaries rather
than Apple silicon binaries).
This addresses #11210
---------
Co-authored-by: viyatb-oai <viyatb@openai.com>
Users were reporting that when they were actively editing a skill file,
they would see frequent errors (one per second) across all of their
active session until they fixed all frontmatter parse errors. This
change will reduce the chatter at the expense of a slightly longer delay
before skills are updated in the UI.
This addresses #11385
The dynamic model refresh feature (`https://api.openai.com/v1/models`
endpoint) is currently gated on a runtime check for an auth method other
than API Key. It should be gated on a check specifically for ChatGPT
Auth because some custom model providers (e.g. for local models) use no
auth mechanism. A call to `self.auth_manager.auth_mode()` will return
`None` in this case.
Addresses #11213
On some platforms, the "notify" file watcher library emits events for
file opens and reads, not just file modifications or deletes. The
previous implementation didn't take this into account.
Furthermore, the `tracing.info!` call that I previously added was
emitting a lot of logs. I had assumed incorrectly that `info` level
logging was disabled by default, but it's apparently enabled for this
crate. This is resulting in large logs (hundreds of MB) for some users.
- Defer rollout persistence for fresh threads (`InitialHistory::New`):
keep rollout events in memory and only materialize rollout file + state
DB row on first `EventMsg::UserMessage`.
- Keep precomputed rollout path available before materialization.
- Change `thread/start` to build thread response from live config
snapshot and optional precomputed path.
- Improve pre-materialization behavior in app-server/TUI: clearer
invalid-request errors for file-backed ops and a friendlier `/fork` “not
ready yet” UX.
- Update tests to match deferred semantics across
start/read/archive/unarchive/fork/resume/review flows.
- Improved resilience of user_shell test, which should be unrelated to
this change but must be affected by timing changes
For Reviewers:
* The primary change is in recorder.rs
* Most of the other changes were to fix up broken assumptions in
existing tests
Testing:
* Manually tested CLI
* Exercised app server paths by manually running IDE Extension with
rebuilt CLI binary
* Only user-visible change is that `/fork` in TUI generates visible
error if used prior to first turn
Hardens PTY Python REPL test and make MCP test startup deterministic
**Summary**
- `utils/pty/src/tests.rs`
- Added a REPL readiness handshake (`wait_for_python_repl_ready`) that
repeatedly sends a marker and waits for it in PTY output before sending
test commands.
- Updated `pty_python_repl_emits_output_and_exits` to:
- wait for readiness first,
- preserve startup output,
- append output collected through process exit.
- Reduces Windows/ConPTY flakiness from early stdin writes racing REPL
startup.
- `mcp-server/tests/suite/codex_tool.rs`
- Avoid remote model refresh during MCP test startup, reducing
timeout-prone nondeterminism.
This PR changes stdio MCP child processes to run in their own process
group
* Add guarded teardown in codex-rmcp-client: send SIGTERM to the group
first, then SIGKILL after a short grace period.
* Add terminate_process_group helper in process_group.rs.
* Add Unix regression test in process_group_cleanup.rs to verify wrapper
+ grandchild are reaped on client drop.
Addresses reported MCP process/thread storm: #10581
## Summary
Stabilize v2 review integration tests by making them hermetic with
respect to model discovery.
`app-server` review tests were intermittently timing out in CI
(especially on Windows runners) because their test config allowed remote
model refresh. During `thread/start`, the test process could issue live
`/v1/models` requests, introducing external network latency and
nondeterministic timing before review flow assertions.
This change disables remote model fetching in the review test config
helper used by these tests.
This PR removes the inaccurate "Disable in /experimental." statement now
that the "personality" feature flag is no longer experimental.
This addresses #10850
Add a centralized FileWatcher in codex-core (using notify) that watches
skill roots from the config layer stack (recursive)
Send `SkillsChanged` events when relevant file system changes are
detected
On `SkillsChanged`:
* Invalidate the skills cache immediately in ThreadManager
* Emit EventMsg::SkillsUpdateAvailable to active sessions
~~* Broadcast a new app-server notification:
SkillsListUpdatedNotification~~
This change does not inject new items into the event stream. That means
the agent will not know about new skills, so it won't be able to
implicitly invoke new skills. It also won't know about changes to
existing skills, so if it has already read the contents of a modified
skill, it will not honor the new behavior.
This change also does not detect modifications to AGENTS.md.
I plan to address these limitations in a follow-on PR modeled after
#9985. Injection of new skills and AGENTS was deemed to risky, hence the
need to split the feature into two stages. The changes in this PR were
designed to easily accommodate the second stage once we have some other
foundational changes in place.
Testing: In addition to automated tests, I did manual testing to confirm
that newly-created skills, deleted skills, and renamed skills are
reflected in the TUI skill picker menu. Also confirmed that
modifications to behaviors for explicitly-invoked skills are honored.
---------
Co-authored-by: Xin Lin <xl@openai.com>
The current bug template uses CLI-specific instructions for getting the
version.
The current feature template doesn't ask the user to provide the Codex
variant (surface) they are using.
This PR addresses these problems.
This PR addresses #10395
When a user is asked to pick the trust level of a project, the code
currently reloads the config if they select "trusted". It doesn't reload
the config in the "untrusted" case but should. This causes the sandbox
mode to be reported incorrectly in `/status` during the first run (it's
displayed as `read-only` even though it acts as though it's
`workspace-write`).
Summary
- require `CODEX_HOME` to point to an existing directory before
canonicalizing and surface clear errors otherwise
- share the same helper logic in both `core` and `rmcp-client` and add
unit tests that cover missing, non-directory, valid, and default paths
This addresses #9222
This PR adds a new `tui.notifications_method` config option that accepts
values of "auto", "osc9" and "bel". It defaults to "auto", which
attempts to auto-detect whether the terminal supports OSC 9 escape
sequences and falls back to BEL if not.
The PR also removes the inconsistent handling of notifications on
Windows when WSL was used.
Fix resume --last prompt parsing by dropping the clap conflict on the
codex resume subcommand so a positional prompt is accepted when --last
is set. This aligns interactive resume behavior with exec-mode logic and
avoids the “--last cannot be used with SESSION_ID” error.
This addresses #6717
We've recently standardized a [feature maturity
model](https://developers.openai.com/codex/feature-maturity) that we're
using in our docs and support forums to communicate expectations to
users. This PR updates the internal stage names and descriptions to
match.
This change involves a simple internal rename and updates to a few
user-visible strings. No functional change.
In a [recent PR](https://github.com/openai/codex/pull/9182), I made some
improvements to config error messages so errors didn't leave app server
clients in a dead state. This is a follow-on PR to make these error
messages more readable and actionable for both TUI and GUI users. For
example, see #9668 where the user was understandably confused about the
source of the problem and how to fix it.
The improved error message:
1. Clearly identifies the config file where the error was found (which
is more important now that we support layered configs)
2. Provides a line and column number of the error
3. Displays the line where the error occurred and underlines it
For example, if my `config.toml` includes the following:
```toml
[features]
collaboration_modes = "true"
```
Here's the current CLI error message:
```
Error loading config.toml: invalid type: string "true", expected a boolean in `features`
```
And here's the improved message:
```
Error loading config.toml:
/Users/etraut/.codex/config.toml:43:23: invalid type: string "true", expected a boolean
|
43 | collaboration_modes = "true"
| ^^^^^^
```
The bulk of the new logic is contained within a new module
`config_loader/diagnostics.rs` that is responsible for calculating the
text range for a given toml path (which is more involved than I would
have expected).
In addition, this PR adds the file name and text range to the
`ConfigWarningNotification` app server struct. This allows GUI clients
to present the user with a better error message and an optional link to
open the errant config file. This was a suggestion from @.bolinfest when
he reviewed my previous PR.
This PR adds support for chained (layered) config.toml file merging for
clients that use the app server interface. This feature already exists
for the TUI, but it does not work for GUI clients.
It does the following:
* Changes code paths for new thread, resume thread, and fork thread to
use the effective config based on the cwd.
* Updates the `config/read` API to accept an optional `cwd` parameter.
If specified, the API returns the effective config based on that cwd
path. Also optionally includes all layers including project config
files. If cwd is not specified, the API falls back on its older behavior
where it considers only the global (non-project) config files when
computing the effective config.
The changes in codex_message_processor.rs look deceptively large. They
mostly just involve moving existing blocks of code to a later point in
some functions so it can use the cwd to calculate the config.
This PR builds upon #9509 and should be reviewed and merged after that
PR.
Tested:
* Verified change with (dependent, as-yet-uncommitted) changes to IDE
Extension and confirmed correct behavior
The full fix requires additional changes in the IDE Extension code base,
but they depend on this PR.
This PR fixes a small issue with chained (layered) config.toml file
merging. The old logic didn't properly handle profiles.
In particular, if a lower-layer config overrides a profile defined in a
higher-layer config, the override did not take effect. This prevents
users from having project-specific profile overrides and contradicts the
(soon-to-be) documented behavior of config merging.
The change adds a unit test for this case. It also exposes a function
from the config crate that is needed by the app server code paths to
implement support for layered configs.
Users of Azure endpoints are reporting that when they use `/review`,
they sometimes see an error "Invalid 'input[3].id". I suspect this is
specific to the Azure implementation of the `responses` API. The Azure
team generally copies the OpenAI code for this endpoint, but they do
have minor differences and sometimes lag in rolling out bug fixes or
updates.
The error appears to be triggered because the `/review` implementation
is using a user ID with a colon in it.
Addresses #9360
We already support reading from `config.toml` through a symlink, but the
code was not properly handling updates to a symlinked config file. This
PR generalizes safe symlink-chain resolution and atomic writes into
path_utils, updating all config write paths to use the shared logic
(including set_default_oss_provider, which previously didn't use the
common path), and adds tests for symlink chains and cycles.
This resolves#6646.
Notes:
* Symlink cycles or resolution failures replace the top-level symlink
with a real file.
* Shared config write path now handles symlinks consistently across
edits, defaults, and empty-user-layer creation.
This PR was inspired by https://github.com/openai/codex/pull/9437, which
was contributed by @ryoppippi
PR #9245 made `codex resume --last` honor cwd, but I forgot to make the
same change for `codex exec resume --last`. This PR fixes the
inconsistency.
This addresses #8700
A recent change in commit ccba737d26 modified the styling of the
placeholder text (e.g. "Implement {feature}") in the input box of the
CLI, changing it from non-italic to italic. I think this was likely
unintentional. It results in a bad display appearance on some terminal
emulators, and several users have complained about it.
This change switches back to non-italic styling, restoring the older
behavior.
It addresses #9262
This PR changes `codex resume --last` to work consistently with `codex
resume`. Namely, it filters based on the cwd when selecting the last
session. It also supports the `--all` modifier as an override.
This addresses #8700
When an invalid config.toml key or value is detected, the CLI currently
just quits. This leaves the VSCE in a dead state.
This PR changes the behavior to not quit and bubble up the config error
to users to make it actionable. It also surfaces errors related to
"rules" parsing.
This allows us to surface these errors to users in the VSCE, like this:
<img width="342" height="129" alt="Screenshot 2026-01-13 at 4 29 22 PM"
src="https://github.com/user-attachments/assets/a79ffbe7-7604-400c-a304-c5165b6eebc4"
/>
<img width="346" height="244" alt="Screenshot 2026-01-13 at 4 45 06 PM"
src="https://github.com/user-attachments/assets/de874f7c-16a2-4a95-8c6d-15f10482e67b"
/>
This is a proposed fix for #8912
Information provided by Codex:
no_proxy means “don’t use any system proxy settings for this client,”
even if macOS has proxies configured in System Settings or via
environment. On macOS, reqwest’s proxy discovery can call into the
system-configuration framework; that’s the code path that was panicking
with “Attempted to create a NULL object.” By forcing a direct connection
for the OAuth discovery request, we avoid that proxy-resolution path
entirely, so the system-configuration crate never gets invoked and the
panic disappears.
Effectively:
With proxies: reqwest asks the OS for proxy config →
system-configuration gets touched → panic.
With no_proxy: reqwest skips proxy lookup → no system-configuration call
→ no panic.
So the fix doesn’t change any MCP protocol behavior; it just prevents
the OAuth discovery probe from touching the macOS proxy APIs that are
crashing in the reported environment.
This fix changes behavior for the OAuth discovery probe used in codex
mcp list/auth status detection. With no_proxy, that probe won’t use
system or env proxy settings, so:
If a server is only reachable via a proxy, the discovery call may fail
and we’ll show auth as Unsupported/NotLoggedIn incorrectly.
If the server is reachable directly (common case), behavior is
unchanged.
As an alternative, we could try to get a fix into the
[system-configuration](https://github.com/mullvad/system-configuration-rs)
library. It looks like this library is still under development but has
slow release pace.
This eliminates redundant user documentation and allows us to focus our
documentation investments.
I left tombstone files for most of the existing ".md" docs files to
avoid broken links. These now contain brief links to the developers docs
site.
This reverts commit 291b54a762.
This commit was intended to prevent the model from making code changes
during `/review`, which is sometimes does. Unfortunately, it has other
unintended side effects that cause `/review` to fail in a variety of
ways. See #8115 and #7815. We've therefore decided to revert this
change.
Normally, all tool calls within a saved session should have a response,
but there are legitimate reasons for the response to be missing. This
can occur if the user canceled the call or there was an error of some
sort during the rollout. We shouldn't panic in this case.
This is a partial fix for #7990
Codex identified this as the cause of a reported hang:
https://github.com/openai/codex/issues/7822. Apparently, the wrapping
algorithm we're using has known issues and bad worst-case behaviors when
OptimalFit is used on certain strings. It recommended switching to
FirstFit instead.
This PR changes the length validation for SKILL.md `name` and
`description` fields so they use character counts rather than byte
counts. Aligned character limits to other harnesses.
This addresses #7730.
This PR is a modified version of [a
PR](https://github.com/openai/codex/pull/7316) submitted by @yydrowz3.
* Removes a redundant `experimental_sandbox_command_assessment` flag
* Moves `mcp_oauth_credentials_store` from the `[features]` table, where
it doesn't belong
Recent model updates caused the experimental "sandbox tool assessment"
to time out most of the time leaving the user without any risk
assessment or tool summary. This change explicitly sets the reasoning
effort to medium and bumps the timeout.
This change has no effect if the user hasn't enabled the
`experimental_sandbox_command_assessment` feature flag.
It stopped working (found zero duplicates) starting three days ago when
the model was switched from `gpt-5` to `gpt-5.1`. I'm not sure why it
stopped working. This is an attempt to get it working again by using the
default model for the codex action (which is presumably
`gpt-5.1-codex-max`).
This PR adds support for a new feature flag `tui.animations`. By
default, the TUI uses animations in its welcome screen, "working"
spinners, and "shimmer" effects. This animations can interfere with
screen readers, so it's good to provide a way to disable them.
This change is inspired by [a
PR](https://github.com/openai/codex/pull/4014) contributed by @Orinks.
That PR has faltered a bit, but I think the core idea is sound. This
version incorporates feedback from @aibrahim-oai. In particular:
1. It uses a feature flag (`tui.animations`) rather than the unqualified
CLI key `no-animations`. Feature flags are the preferred way to expose
boolean switches. They are also exposed via CLI command switches.
2. It includes more complete documentation.
3. It disables a few animations that the other PR omitted.
This PR fixes the `release_event_does_not_change_selection` test so it
doesn't cause an extra `config.toml` to be emitted in the sources when
running the tests locally. Prior to this fix, I needed to delete this
file every time I ran the tests to prevent it from showing up as an
uncommitted source file.
The `generated_ts_has_no_optional_nullable_fields` test was occasionally
failing on slow CI nodes because of a timeout. This change reduces the
work done by the test. It adds some "options" for the `generate_ts`
function so it can skip work that's not needed for the test.
This PR fixes keyboard handling for the Right Alt (aka "Alt-Gr") key on
Windows. This key appears on keyboards in Central and Eastern Europe.
Codex has effectively never worked for Windows users in these regions
because the code didn't properly handle this key, which is used for
typing common symbols like `\` and `@`.
A few days ago, I merged a [community-authored
PR](https://github.com/openai/codex/pull/6720) that supplied a partial
fix for this issue. Upon closer inspect, that PR was 1) too broad (not
scoped to Windows only) and 2) incomplete (didn't fix all relevant code
paths, so paste was still broken).
This improvement is based on another [community-provided
PR](https://github.com/openai/codex/pull/3241) by @marektomas-cz. He
submitted it back in September and later closed it because it didn't
receive any attention.
This fix addresses the following bugs: #5922, #3046, #3092, #3519,
#5684, #5843.
We've received many reports of codex hanging when calling certain tools.
[Here](https://github.com/openai/codex/issues/3204) is one example. This
is likely a major cause. The problem occurs when
`consume_truncated_output` waits for `stdout` and `stderr` to be closed
once the child process terminates. This normally works fine, but it
doesn't handle the case where the child has spawned grandchild processes
that inherits `stdout` and `stderr`.
The fix was originally written by @md-oai in [this
PR](https://github.com/openai/codex/pull/1852), which has gone stale.
I've copied the original fix (which looks sound to me) and added an
integration test to prevent future regressions.
This PR addresses https://github.com/openai/codex/issues/6360. The root
problem is that the TUI was directly loading the `auth.json` file to
access the auth information. It should instead be using the AuthManager,
which records the current auth information. The `auth.json` file can be
overwritten at any time by other instances of the CLI or extension, so
its information can be out of sync with the current instance. The
`/status` command should always report the auth information associated
with the current instance.
An alternative fix for this bug was submitted by @chojs23 in [this
PR](https://github.com/openai/codex/pull/6495). That approach was only a
partial fix.
The CLA action is designed to automatically lock a PR when it is closed.
This preserves the CLA agreement statements, preventing the contributor
from deleting them after the fact. However, this action is currently
locking PRs that are closed without merging. I'd like to keep such PRs
open so the contributor can respond with additional comments. I'm
currently manually unlocking PRs that I close, but I'd like to eliminate
this manual step.
I recently fixed a bug in [this
PR](https://github.com/openai/codex/pull/6285) that prevented Ctrl+C
from dismissing the login menu in the TUI and leaving the user unauthed.
A [user pointed out](https://github.com/openai/codex/issues/6418) that
this makes Ctrl+C can no longer be used to exit the app. This PR changes
the behavior so we exit the app rather than ignoring the Ctrl+C.
This is a simplified version of [a
PR](https://github.com/openai/codex/pull/6134) supplied by a community
member.
It updates the docs to reflect a recent config deprecation.
This PR updates the AI prompt used for the workflow that adds automated
labels to incoming issues. I've been updating and refining the list of
labels as I work through the issue backlog, and the old prompt was
becoming somewhat outdated.
This PR makes an "insufficient quota" error fatal so we don't attempt to
retry it multiple times in the agent loop.
We have multiple bug reports from users about intermittent retry
behaviors, and this could explain some of them. With this change, we'll
eliminate the retries and surface a clear error message.
The PR is a nearly identical copy of [this
PR](https://github.com/openai/codex/pull/4837) contributed by
@abimaelmartell. The original PR has gone stale. Rather than wait for
the contributor to resolve merge conflicts, I wanted to get this change
in.
Some PRs are being submitted without reference to existing bug reports
or feature requests. This updates the PR template and contributing
guidelines to request that all PRs from the community contain such a
link. This provides additional context and helps prioritize, track, and
assess PRs.
We currently allow the user to dismiss the login menu via Ctrl+C. This
leaves them in a bad state where they're not auth'ed but have an input
prompt. In the extension, this isn't a problem because we don't allow
the user to dismiss the login screen.
Testing: I confirmed that Ctrl+C no longer dismisses the login menu.
This is an alternative (simpler) fix for a [community
PR](https://github.com/openai/codex/pull/3234).
I'm seeing two tests fail intermittently in CI. This PR attempts to
address (or at least mitigate) the flakiness.
* summarize_context_three_requests_and_instructions - The test snapshots
server.received_requests() immediately after observing TaskComplete.
Because the OpenAI /v1/responses call is streamed, the HTTP request can
still be draining when that event fires, so wiremock occasionally
reports only two captured requests. Fix is to wait for async activity to
complete.
* archive_conversation_moves_rollout_into_archived_directory - times out
on a slow CI run. Mitigation is to increase timeout value from 10s to
20s.
Currently, when the access token expires, we attempt to use the refresh
token to acquire a new access token. This works most of the time.
However, there are situations where the refresh token is expired,
exhausted (already used to perform a refresh), or revoked. In those
cases, the current logic treats the error as transient and attempts to
retry it repeatedly.
This PR changes the token refresh logic to differentiate between
permanent and transient errors. It also changes callers to treat the
permanent errors as fatal rather than retrying them. And it provides
better error messages to users so they understand how to address the
problem. These error messages should also help us further understand why
we're seeing examples of refresh token exhaustion.
Here is the error message in the CLI. The same text appears within the
extension.
<img width="863" height="38" alt="image"
src="https://github.com/user-attachments/assets/7ffc0d08-ebf0-4900-b9a9-265064202f4f"
/>
I also correct the spelling of "Re-connecting", which shouldn't have a
hyphen in it.
Testing: I manually tested these code paths by adding temporary code to
programmatically cause my refresh token to be exhausted (by calling the
token refresh endpoint in a tight loop more than 50 times). I then
simulated an access token expiration, which caused the token refresh
logic to be invoked. I confirmed that the updated logic properly handled
the error condition.
Note: We earlier discussed the idea of forcefully logging out the user
at the point where token refresh failed. I made several attempts to do
this, and all of them resulted in a bad UX. It's important to surface
this error to users in a way that explains the problem and tells them
that they need to log in again. We also previously discussed deleting
the auth.json file when this condition is detected. That also creates
problems because it effectively changes the auth status from logged in
to logged out, and this causes odd failures and inconsistent UX. I think
it's therefore better not to delete auth.json in this case. If the user
closes the CLI or VSCE and starts it again, we properly detect that the
access token is expired and the refresh token is "dead", and we force
the user to go through the login flow at that time.
This should address aspects of #6191, #5679, and #5505
This fixes bug #6121.
The `input_messages` field passed to the notify handler is currently
empty because the logic is incorrectly including the OutputText rather
than InputText. I've fixed that and added proper filtering to remove
messages associated with AGENTS.md and other context injected by the
harness.
Testing: I wrote a notify handler and verified that the user prompt is
correctly passed through to the handler.
Addresses issue https://github.com/openai/codex/issues/3582 where an
"archive conversation" command in the extension fails on Windows.
The problem is that the `archive_conversation` api server call is not
canonicalizing the path to the rollout path when performing its check to
verify that the rollout path is in the sessions directory. This causes
it to fail 100% of the time on Windows.
Testing: I was able to repro the error on Windows 100% prior to this
change. After the change, I'm no longer able to repro.
When I enable `experimental_sandbox_command_assessment`, I get an
incorrect deprecation warning: "experimental_sandbox_command_assessment
is deprecated. Use experimental_sandbox_command_assessment instead."
This PR fixes this error.
* Removed sandbox risk categories; feedback indicates that these are not
that useful and "less is more"
* Tweaked the assessment prompt to generate terser answers
* Fixed bug in orchestrator that prevents this feature from being
exposed in the extension
This PR addresses a current hole in the TypeScript code generation for
the API server protocol. Fields that are marked as "Optional<>" in the
Rust code are serialized such that the value is omitted when it is
deserialized — appearing as `undefined`, but the TS type indicates
(incorrectly) that it is always defined but possibly `null`. This can
lead to subtle errors that the TypeScript compiler doesn't catch. The
fix is to include the `#[ts(optional_fields = nullable)]` macro for all
protocol structs that contain one or more `Optional<>` fields.
This PR also includes a new test that validates that all TS protocol
code containing "| null" in its type is marked optional ("?") to catch
cases where `#[ts(optional_fields = nullable)]` is omitted.
Addresses https://github.com/openai/codex/issues/5773
Testing: I tested that images work (regardless of order that they are
associated with the task prompt) in both the CLI and Extension. Also
verified that conversations in CLI and extension with images can be
resumed.
This PR does the following:
1. Changes `try_refresh_token` to handle the case where the endpoint
returns a response without an `id_token`. The OpenID spec indicates that
this field is optional and clients should not assume it's present.
2. Changes the `attempt_stream_responses` to propagate token refresh
errors rather than silently ignoring them.
3. Fixes a typo in a couple of error messages (unrelated to the above,
but something I noticed in passing) - "reconnect" should be spelled
without a hyphen.
This PR does not implement the additional suggestion from @pakrym-oai
that we should sign out when receiving `refresh_token_expired` from the
refresh endpoint. Leaving this as a follow-on because I'm undecided on
whether this should be implemented in `try_refresh_token` or its
callers.
Currently, `approval_policy` is supported in profiles, but
`sandbox_mode` is not. This PR adds support for `sandbox_mode`.
Note: a fix for this was submitted in [this
PR](https://github.com/openai/codex/pull/2397), but the underlying code
has changed significantly since then.
This addresses issue #3034
This PR fixes a test that is sporadically failing in CI.
The problem is that two unit tests (the older `login_and_cancel_chatgpt`
and a recently added
`login_chatgpt_includes_forced_workspace_query_param`) exercise code
paths that start the login server. The server binds to a hard-coded
localhost port number, so attempts to start more than one server at the
same time will fail. If these two tests happen to run concurrently, one
of them will fail.
To fix this, I've added a simple mutex. We can use this same mutex for
future tests that use the same pattern.
This PR adds support for a model-based summary and risk assessment for
commands that violate the sandbox policy and require user approval. This
aids the user in evaluating whether the command should be approved.
The feature works by taking a failed command and passing it back to the
model and asking it to summarize the command, give it a risk level (low,
medium, high) and a risk category (e.g. "data deletion" or "data
exfiltration"). It uses a new conversation thread so the context in the
existing thread doesn't influence the answer. If the call to the model
fails or takes longer than 5 seconds, it falls back to the current
behavior.
For now, this is an experimental feature and is gated by a config key
`experimental_sandbox_command_assessment`.
Here is a screen shot of the approval prompt showing the risk assessment
and summary.
<img width="723" height="282" alt="image"
src="https://github.com/user-attachments/assets/4597dd7c-d5a0-4e9f-9d13-414bd082fd6b"
/>
This PR fixes a bug that results in a hang in the oauth login flow if a
user logs in, then logs out, then logs in again without first closing
the browser window.
Root cause of problem: We use a local web server for the oauth flow, and
it's implemented using the `tiny_http` rust crate. During the first
login, a socket is created between the browser and the server. The
`tiny_http` library creates worker threads that persist for as long as
this socket remains open. Currently, there's no way to close the
connection on the server side — the library provides no API to do this.
The library also filters all "Connect: close" headers, which makes it
difficult to tell the client browser to close the connection. On the
second login attempt, the browser uses the existing connection rather
than creating a new one. Since that connection is associated with a
server instance that no longer exists, it is effectively ignored.
I considered switching from `tiny_http` to a different web server
library, but that would have been a big change with significant
regression risk. This PR includes a more surgical fix that works around
the limitation of `tiny_http` and sends a "Connect: close" header on the
last "success" page of the oauth flow.
This addresses bug #4092
Testing:
* Confirmed error occurs prior to fix if logging in using API key and no
`~/.codex` directory exists
* Confirmed after fix that `~/.codex` directory is properly created and
error doesn't occur
When logging in using ChatGPT using the `codex login` command, a
successful login should write a new `auth.json` file with the ChatGPT
token information. The old code attempted to retain the API key and
merge the token information into the existing `auth.json` file. With the
new simplified login mechanism, `auth.json` should have auth information
for only ChatGPT or API Key, not both.
The `codex login --api-key <key>` code path was already doing the right
thing here, but the `codex login` command was incorrect. This PR fixes
the problem and adds test cases for both commands.
This PR addresses an edge-case bug that appears in the VS Code extension
in the following situation:
1. Log in using ChatGPT (using either the CLI or extension). This will
create an `auth.json` file.
2. Manually modify `config.toml` to specify a custom provider.
3. Start a fresh copy of the VS Code extension.
The profile menu in the VS Code extension will indicate that you are
logged in using ChatGPT even though you're not.
This is caused by the `get_auth_status` method returning an
`auth_method: 'chatgpt'` when a custom provider is configured and it
doesn't use OpenAI auth (i.e. `requires_openai_auth` is false). The
method should always return `auth_method: None` if
`requires_openai_auth` is false.
The same bug also causes the NUX (new user experience) screen to be
displayed in the VSCE in this situation.
This PR does the following:
* Adds the ability to paste or type an API key.
* Removes the `preferred_auth_method` config option. The last login
method is always persisted in auth.json, so this isn't needed.
* If OPENAI_API_KEY env variable is defined, the value is used to
prepopulate the new UI. The env variable is otherwise ignored by the
CLI.
* Adds a new MCP server entry point "login_api_key" so we can implement
this same API key behavior for the VS Code extension.
<img width="473" height="140" alt="Screenshot 2025-09-04 at 3 51 04 PM"
src="https://github.com/user-attachments/assets/c11bbd5b-8a4d-4d71-90fd-34130460f9d9"
/>
<img width="726" height="254" alt="Screenshot 2025-09-04 at 3 51 32 PM"
src="https://github.com/user-attachments/assets/6cc76b34-309a-4387-acbc-15ee5c756db9"
/>
Eric TrautandGitHub
