## Summary
This PR consolidates base_instructions onto SessionMeta /
SessionConfiguration, so we ensure `base_instructions` is set once per
session and should be (mostly) immutable, unless:
- overridden by config on resume / fork
- sub-agent tasks, like review or collab
In a future PR, we should convert all references to `base_instructions`
to consistently used the typed struct, so it's less likely that we put
other strings there. See #9423. However, this PR is already quite
complex, so I'm deferring that to a follow-up.
## Testing
- [x] Added a resume test to assert that instructions are preserved. In
particular, `resume_switches_models_preserves_base_instructions` fails
against main.
Existing test coverage thats assert base instructions are preserved
across multiple requests in a session:
- Manual compact keeps baseline instructions:
core/tests/suite/compact.rs:199
- Auto-compact keeps baseline instructions:
core/tests/suite/compact.rs:1142
- Prompt caching reuses the same instructions across two requests:
core/tests/suite/prompt_caching.rs:150 and
core/tests/suite/prompt_caching.rs:157
- Prompt caching with explicit expected string across two requests:
core/tests/suite/prompt_caching.rs:213 and
core/tests/suite/prompt_caching.rs:222
- Resume with model switch keeps original instructions:
core/tests/suite/resume.rs:136
- Compact/resume/fork uses request 0 instructions for later expected
payloads: core/tests/suite/compact_resume_fork.rs:215
### Description
- Remove the now-unused `instructions` field from the session metadata
to simplify SessionMeta and stop propagating transient instruction text
through the rollout recorder API. This was only saving
user_instructions, and was never being read.
- Stop passing user instructions into the rollout writer at session
creation so the rollout header only contains canonical session metadata.
### Testing
- Ran `just fmt` which completed successfully.
- Ran `just fix -p codex-protocol`, `just fix -p codex-core`, `just fix
-p codex-app-server`, `just fix -p codex-tui`, and `just fix -p
codex-tui2` which completed (Clippy fixes applied) as part of
verification.
- Ran `cargo test -p codex-protocol` which passed (28 tests).
- Ran `cargo test -p codex-core` which showed failures in a small set of
tests (not caused by the protocol type change directly):
`default_client::tests::test_create_client_sets_default_headers`,
several `models_manager::manager::tests::refresh_available_models_*`,
and `shell_snapshot::tests::linux_sh_snapshot_includes_sections` (these
tests failed in this CI run).
- Ran `cargo test -p codex-app-server` which reported several failing
integration tests (including
`suite::codex_message_processor_flow::test_codex_jsonrpc_conversation_flow`,
`suite::output_schema::send_user_turn_*`, and
`suite::user_agent::get_user_agent_returns_current_codex_user_agent`).
- `cargo test -p codex-tui` and `cargo test -p codex-tui2` were
attempted but aborted due to disk space exhaustion (`No space left on
device`).
------
[Codex
Task](https://chatgpt.com/codex/tasks/task_i_696bd8ce632483228d298cf07c7eb41c)
Add support for returning threads by either `created_at` OR `updated_at`
descending. Previously core always returned threads ordered by
`created_at`.
This PR:
- updates core to be able to list threads by `updated_at` OR
`created_at` descending based on what the caller wants
- also update `thread/list` in app-server to expose this (default to
`created_at` if not specified)
All existing codepaths (app-server, TUI) still default to `created_at`,
so no behavior change is expected with this PR.
**Implementation**
To sort by `updated_at` is a bit nontrivial (whereas `created_at` is
easy due to the way we structure the folders and filenames on disk,
which are all based on `created_at`).
The most naive way to do this without introducing a cache file or sqlite
DB (which we have to implement/maintain) is to scan files in reverse
`created_at` order on disk, and look at the file's mtime (last modified
timestamp according to the filesystem) until we reach `MAX_SCAN_FILES`
(currently set to 10,000). Then, we can return the most recent N
threads.
Based on some quick and dirty benchmarking on my machine with ~1000
rollout files, calling `thread/list` with limit 50, the `updated_at`
path is slower as expected due to all the I/O:
- updated-at: average 103.10 ms
- created-at: average 41.10 ms
Those absolute numbers aren't a big deal IMO, but we can certainly
optimize this in a followup if needed by introducing more state stored
on disk.
**Caveat**
There's also a limitation in that any files older than `MAX_SCAN_FILES`
will be excluded, which means if a user continues a REALLY old thread,
it's possible to not be included. In practice that should not be too big
of an issue.
If a user makes...
- 1000 rollouts/day → threads older than 10 days won't show up
- 100 rollouts/day → ~100 days
If this becomes a problem for some reason, even more motivation to
implement an updated_at cache.
- capture the header from SSE/WS handshakes, store it per
ModelClientSession using `Oncelock`, echo it on turn-scoped requests,
and add SSE+WS integration tests for within-turn persistence +
cross-turn reset.
- keep `x-codex-turn-state` sticky within a user turn to maintain
routing continuity for retries/tool follow-ups.
The second part of breaking up PR
https://github.com/openai/codex/pull/9116
Summary:
- Add `TextElement` / `ByteRange` to protocol user inputs and user
message events with defaults.
- Thread `text_elements` through app-server v1/v2 request handling and
history rebuild.
- Preserve UI metadata only in user input/events (not `ContentItem`)
while keeping local image attachments in user events for rehydration.
Details:
- Protocol: `UserInput::Text` carries `text_elements`;
`UserMessageEvent` carries `text_elements` + `local_images`.
Serialization includes empty vectors for backward compatibility.
- app-server-protocol: v1 defines `V1TextElement` / `V1ByteRange` in
camelCase with conversions; v2 uses its own camelCase wrapper.
- app-server: v1/v2 input mapping includes `text_elements`; thread
history rebuilds include them.
- Core: user event emission preserves UI metadata while model history
stays clean; history replay round-trips the metadata.
moving `web_search` rollout serverside, so need a way to explicitly
disable search + signal eligibility from the client.
- Add `x‑oai‑web‑search‑eligible` header that signifies whether the
request can have web search.
- Only attach the `web_search` tool when the resolved `WebSearchMode` is
`Live` or `Cached`.
### What
Add `WebSearchMode` enum (disabled, cached live, defaults to cached) to
config + V2 protocol. This enum takes precedence over legacy flags:
`web_search_cached`, `web_search_request`, and `tools.web_search`.
Keep `--search` as live.
### Tests
Added tests
Instead of having a hard-coded default review model, use the current
model for running `/review` unless one is specified in the config.
Also inherit current reasoning effort
Have only the following Methods:
- `list_models`: getting current available models
- `try_list_models`: sync version no refresh for tui use
- `get_default_model`: get the default model (should be tightened to
core and received on session configuration)
- `get_model_info`: get `ModelInfo` for a specific model (should be
tightened to core but used in tests)
- `refresh_if_new_etag`: trigger refresh on different etags
Also move the cache to its own struct
The connection was being added to the list after the WebSocket response
was sent.
So the test can sometimes race and observe connections before the list
was updated.
After this change, connection and request is added to the list before
the response is sent.
Enterprises want to restrict the MCP servers their users can use.
Admins can now specify an allowlist of MCPs in `requirements.toml`. The
MCP servers are matched on both Name and Transport (local path or HTTP
URL) -- both must match to allow the MCP server. This prevents
circumventing the allowlist by renaming MCP servers in user config. (It
is still possible to replace the local path e.g. rewrite say
`/usr/local/github-mcp` with a nefarious MCP. We could allow hash
pinning in the future, but that would break updates. I also think this
represents a broader, out-of-scope problem.)
We introduce a new field to Constrained: "normalizer". In general, it is
a fn(T) -> T and applies when `Constrained<T>.set()` is called. In this
particular case, it disables MCP servers which do not match the
allowlist. An alternative solution would remove this and instead throw a
ConstraintError. That would stop Codex launching if any MCP server was
configured which didn't match. I think this is bad.
We currently reuse the enabled flag on MCP servers to disable them, but
don't propagate any information about why they are disabled. I'd like to
add that in a follow up PR, possibly by switching out enabled with an
enum.
In action:
```
# MCP server config has two MCPs. We are going to allowlist one of them.
➜ codex git:(gt/restrict-mcps) ✗ cat ~/.codex/config.toml | grep mcp_servers -A1
[mcp_servers.hello_world]
command = "hello-world-mcp"
--
[mcp_servers.docs]
command = "docs-mcp"
# Restrict the MCPs to the hello_world MCP.
➜ codex git:(gt/restrict-mcps) ✗ defaults read com.openai.codex requirements_toml_base64 | base64 -d
[mcp_server_allowlist.hello_world]
command = "hello-world-mcp"
# List the MCPs, observe hello_world is enabled and docs is disabled.
➜ codex git:(gt/restrict-mcps) ✗ just codex mcp list
cargo run --bin codex -- "$@"
Finished `dev` profile [unoptimized + debuginfo] target(s) in 0.25s
Running `target/debug/codex mcp list`
Name Command Args Env Cwd Status Auth
docs docs-mcp - - - disabled Unsupported
hello_world hello-world-mcp - - - enabled Unsupported
# Remove the restrictions.
➜ codex git:(gt/restrict-mcps) ✗ defaults delete com.openai.codex requirements_toml_base64
# Observe both MCPs are enabled.
➜ codex git:(gt/restrict-mcps) ✗ just codex mcp list
cargo run --bin codex -- "$@"
Finished `dev` profile [unoptimized + debuginfo] target(s) in 0.25s
Running `target/debug/codex mcp list`
Name Command Args Env Cwd Status Auth
docs docs-mcp - - - enabled Unsupported
hello_world hello-world-mcp - - - enabled Unsupported
# A new requirements that updates the command to one that does not match.
➜ codex git:(gt/restrict-mcps) ✗ cat ~/requirements.toml
[mcp_server_allowlist.hello_world]
command = "hello-world-mcp-v2"
# Use those requirements.
➜ codex git:(gt/restrict-mcps) ✗ defaults write com.openai.codex requirements_toml_base64 "$(base64 -i /Users/gt/requirements.toml)"
# Observe both MCPs are disabled.
➜ codex git:(gt/restrict-mcps) ✗ just codex mcp list
cargo run --bin codex -- "$@"
Finished `dev` profile [unoptimized + debuginfo] target(s) in 0.75s
Running `target/debug/codex mcp list`
Name Command Args Env Cwd Status Auth
docs docs-mcp - - - disabled Unsupported
hello_world hello-world-mcp - - - disabled Unsupported
```
- Add a single builder for developer permissions messaging that accepts
SandboxPolicy and approval policy. This builder now drives the developer
“permissions” message that’s injected at session start and any time
sandbox/approval settings change.
- Trim EnvironmentContext to only include cwd, writable roots, and
shell; removed sandbox/approval/network duplication and adjusted XML
serialization and tests accordingly.
Follow-up: adding a config value to replace the developer permissions
message for custom sandboxes.
Agent wouldn't "see" attached images and would instead try to use the
view_file tool:
<img width="1516" height="504" alt="image"
src="https://github.com/user-attachments/assets/68a705bb-f962-4fc1-9087-e932a6859b12"
/>
In this PR, we wrap image content items in XML tags with the name of
each image (now just a numbered name like `[Image #1]`), so that the
model can understand inline image references (based on name). We also
put the image content items above the user message which the model seems
to prefer (maybe it's more used to definitions being before references).
We also tweak the view_file tool description which seemed to help a bit
Results on a simple eval set of images:
Before
<img width="980" height="310" alt="image"
src="https://github.com/user-attachments/assets/ba838651-2565-4684-a12e-81a36641bf86"
/>
After
<img width="918" height="322" alt="image"
src="https://github.com/user-attachments/assets/10a81951-7ee6-415e-a27e-e7a3fd0aee6f"
/>
```json
[
{
"id": "single_describe",
"prompt": "Describe the attached image in one sentence.",
"images": ["image_a.png"]
},
{
"id": "single_color",
"prompt": "What is the dominant color in the image? Answer with a single color word.",
"images": ["image_b.png"]
},
{
"id": "orientation_check",
"prompt": "Is the image portrait or landscape? Answer in one sentence.",
"images": ["image_c.png"]
},
{
"id": "detail_request",
"prompt": "Look closely at the image and call out any small details you notice.",
"images": ["image_d.png"]
},
{
"id": "two_images_compare",
"prompt": "I attached two images. Are they the same or different? Briefly explain.",
"images": ["image_a.png", "image_b.png"]
},
{
"id": "two_images_captions",
"prompt": "Provide a short caption for each image (Image 1, Image 2).",
"images": ["image_c.png", "image_d.png"]
},
{
"id": "multi_image_rank",
"prompt": "Rank the attached images from most colorful to least colorful.",
"images": ["image_a.png", "image_b.png", "image_c.png"]
},
{
"id": "multi_image_choice",
"prompt": "Which image looks more vibrant? Answer with 'Image 1' or 'Image 2'.",
"images": ["image_b.png", "image_d.png"]
}
]
```
This PR configures Codex CLI so it can be built with
[Bazel](https://bazel.build) in addition to Cargo. The `.bazelrc`
includes configuration so that remote builds can be done using
[BuildBuddy](https://www.buildbuddy.io).
If you are familiar with Bazel, things should work as you expect, e.g.,
run `bazel test //... --keep-going` to run all the tests in the repo,
but we have also added some new aliases in the `justfile` for
convenience:
- `just bazel-test` to run tests locally
- `just bazel-remote-test` to run tests remotely (currently, the remote
build is for x86_64 Linux regardless of your host platform). Note we are
currently seeing the following test failures in the remote build, so we
still need to figure out what is happening here:
```
failures:
suite::compact::manual_compact_twice_preserves_latest_user_messages
suite::compact_resume_fork::compact_resume_after_second_compaction_preserves_history
suite::compact_resume_fork::compact_resume_and_fork_preserve_model_history_view
```
- `just build-for-release` to build release binaries for all
platforms/architectures remotely
To setup remote execution:
- [Create a buildbuddy account](https://app.buildbuddy.io/) (OpenAI
employees should also request org access at
https://openai.buildbuddy.io/join/ with their `@openai.com` email
address.)
- [Copy your API key](https://app.buildbuddy.io/docs/setup/) to
`~/.bazelrc` (add the line `build
--remote_header=x-buildbuddy-api-key=YOUR_KEY`)
- Use `--config=remote` in your `bazel` invocations (or add `common
--config=remote` to your `~/.bazelrc`, or use the `just` commands)
## CI
In terms of CI, this PR introduces `.github/workflows/bazel.yml`, which
uses Bazel to run the tests _locally_ on Mac and Linux GitHub runners
(we are working on supporting Windows, but that is not ready yet). Note
that the failures we are seeing in `just bazel-remote-test` do not occur
on these GitHub CI jobs, so everything in `.github/workflows/bazel.yml`
is green right now.
The `bazel.yml` uses extra config in `.github/workflows/ci.bazelrc` so
that macOS CI jobs build _remotely_ on Linux hosts (using the
`docker://docker.io/mbolin491/codex-bazel` Docker image declared in the
root `BUILD.bazel`) using cross-compilation to build the macOS
artifacts. Then these artifacts are downloaded locally to GitHub's macOS
runner so the tests can be executed natively. This is the relevant
config that enables this:
```
common:macos --config=remote
common:macos --strategy=remote
common:macos --strategy=TestRunner=darwin-sandbox,local
```
Because of the remote caching benefits we get from BuildBuddy, these new
CI jobs can be extremely fast! For example, consider these two jobs that
ran all the tests on Linux x86_64:
- Bazel 1m37s
https://github.com/openai/codex/actions/runs/20861063212/job/59940545209?pr=8875
- Cargo 9m20s
https://github.com/openai/codex/actions/runs/20861063192/job/59940559592?pr=8875
For now, we will continue to run both the Bazel and Cargo jobs for PRs,
but once we add support for Windows and running Clippy, we should be
able to cutover to using Bazel exclusively for PRs, which should still
speed things up considerably. We will probably continue to run the Cargo
jobs post-merge for commits that land on `main` as a sanity check.
Release builds will also continue to be done by Cargo for now.
Earlier attempt at this PR: https://github.com/openai/codex/pull/8832
Earlier attempt to add support for Buck2, now abandoned:
https://github.com/openai/codex/pull/8504
---------
Co-authored-by: David Zbarsky <dzbarsky@gmail.com>
Co-authored-by: Michael Bolin <mbolin@openai.com>
This seems to be necessary to get the Bazel builds on ARM Linux to go
green on https://github.com/openai/codex/pull/8875.
I don't feel great about timeout-whack-a-mole, but we're still learning
here...
I have seen this test flake out sometimes when running the macOS build
using Bazel in CI: https://github.com/openai/codex/pull/8875. Perhaps
Bazel runs with greater parallelism, inducing a heavier load, causing an
issue?
Historically we started with a CodexAuth that knew how to refresh it's
own tokens and then added AuthManager that did a different kind of
refresh (re-reading from disk).
I don't think it makes sense for both `CodexAuth` and `AuthManager` to
be mutable and contain behaviors.
Move all refresh logic into `AuthManager` and keep `CodexAuth` as a data
object.
Add metrics capabilities to Codex. The `README.md` is up to date.
This will not be merged with the metrics before this PR of course:
https://github.com/openai/codex/pull/8350
To support Bazelification in https://github.com/openai/codex/pull/8875,
this PR introduces a new `find_resource!` macro that we use in place of
our existing logic in tests that looks for resources relative to the
compile-time `CARGO_MANIFEST_DIR` env var.
To make this work, we plan to add the following to all `rust_library()`
and `rust_test()` Bazel rules in the project:
```
rustc_env = {
"BAZEL_PACKAGE": native.package_name(),
},
```
Our new `find_resource!` macro reads this value via
`option_env!("BAZEL_PACKAGE")` so that the Bazel package _of the code
using `find_resource!`_ is injected into the code expanded from the
macro. (If `find_resource()` were a function, then
`option_env!("BAZEL_PACKAGE")` would always be
`codex-rs/utils/cargo-bin`, which is not what we want.)
Note we only consider the `BAZEL_PACKAGE` value when the `RUNFILES_DIR`
environment variable is set at runtime, indicating that the test is
being run by Bazel. In this case, we have to concatenate the runtime
`RUNFILES_DIR` with the compile-time `BAZEL_PACKAGE` value to build the
path to the resource.
In testing this change, I discovered one funky edge case in
`codex-rs/exec-server/tests/common/lib.rs` where we have to _normalize_
(but not canonicalize!) the result from `find_resource!` because the
path contains a `common/..` component that does not exist on disk when
the test is run under Bazel, so it must be semantically normalized using
the [`path-absolutize`](https://crates.io/crates/path-absolutize) crate
before it is passed to `dotslash fetch`.
Because this new behavior may be non-obvious, this PR also updates
`AGENTS.md` to make humans/Codex aware that this API is preferred.
Adds a new feature
`enable_request_compression` that will compress using zstd requests to
the codex-backend. Currently only enabled for codex-backend so only enabled for openai providers when using chatgpt::auth even when the feature is enabled
Added a new info log line too for evaluating the compression ratio and
overhead off compressing before requesting. You can enable with
`RUST_LOG=$RUST_LOG,codex_client::transport=info`
```
2026-01-06T00:09:48.272113Z INFO codex_client::transport: Compressed request body with zstd pre_compression_bytes=28914 post_compression_bytes=11485 compression_duration_ms=0
```
We used to override truncation policy by comparing model info vs config
value in context manager. A better way to do it is to construct model
info using the config value
