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"
/>
The API schema export is currently broken:
```
> cargo run -p codex-app-server-protocol --bin export -- --out DIR
Error: this type cannot be exported
```
This PR fixes the error message so we get more info:
```
> cargo run -p codex-app-server-protocol --bin export -- --out DIR
Error: failed to export client responses: dependency core::option::Option<codex_protocol::account::Account> cannot be exported
```
And fixes the root cause which is the `account/read` response.
## Summary
- wrap the default reqwest::Client inside a new
CodexHttpClient/CodexRequestBuilder pair and log the HTTP method, URL,
and status for each request
- update the auth/model/provider plumbing to use the new builder helpers
so headers and bearer auth continue to be applied consistently
- add the shared `http` dependency that backs the header conversion
helpers
## Testing
- `CODEX_SANDBOX=seatbelt CODEX_SANDBOX_NETWORK_DISABLED=1 cargo test -p
codex-core`
- `CODEX_SANDBOX=seatbelt CODEX_SANDBOX_NETWORK_DISABLED=1 cargo test -p
codex-chatgpt`
- `CODEX_SANDBOX=seatbelt CODEX_SANDBOX_NETWORK_DISABLED=1 cargo test -p
codex-tui`
------
https://chatgpt.com/codex/tasks/task_i_68fa5038c17483208b1148661c5873be
1. I have seen too many reports of people hitting startup timeout errors
and thinking Codex is broken. Hopefully this will help people
self-serve. We may also want to consider raising the timeout to ~15s.
2. Make it more clear what PAT is (personal access token) in the GitHub
error
<img width="2378" height="674" alt="CleanShot 2025-10-23 at 22 05 06"
src="https://github.com/user-attachments/assets/d148ce1d-ade3-4511-84a4-c164aefdb5c5"
/>
I want to centralize input processing and management to
`ConversationHistory`. This would need `ConversationHistory` to have
access to `token_info` (i.e. preventing adding a big input to the
history). Besides, it makes more sense to have it on
`ConversationHistory` than `state`.
Walk the sessions tree instead of using file_search so gitignored
CODEX_HOME directories can resume sessions. Add a regression test that
covers a .gitignore'd sessions directory.
Fixes#5247Fixes#5412
---------
Co-authored-by: Owen Lin <owen@openai.com>
Currently we collect all all turn items in a vector, then we add it to
the history on success. This result in losing those items on errors
including aborting `ctrl+c`.
This PR:
- Adds the ability for the tool call to handle cancellation
- bubble the turn items up to where we are recording this info
Admittedly, this logic is an ad-hoc logic that doesn't handle a lot of
error edge cases. The right thing to do is recording to the history on
the spot as `items`/`tool calls output` come. However, this isn't
possible because of having different `task_kind` that has different
`conversation_histories`. The `try_run_turn` has no idea what thread are
we using. We cannot also pass an `arc` to the `conversation_histories`
because it's a private element of `state`.
That's said, `abort` is the most common case and we should cover it
until we remove `task kind`
This shows the aggregated (stdout + stderr) buffer regardless of exit
code.
Many commands output useful / relevant info on stdout when returning a
non-zero exit code, or the same on stderr when returning an exit code of
0. Often, useful info is present on both stdout AND stderr. Also, the
model sees both. So it is confusing to see commands listed as "(no
output)" that in fact do have output, just on the stream that doesn't
match the exit status, or to see some sort of trivial output like "Tests
failed" but lacking any information about the actual failure.
As such, always display the aggregated output in the display. Transcript
mode remains unchanged as it was already displaying the text that the
model sees, which seems correct for transcript mode.
Michael Bolin