Windows release builds in `.github/workflows/rust-release.yml` were
still using GitHub-hosted `windows-latest` and `windows-11-arm` runners.
This change aligns release builds with the faster dedicated Codex runner
pool already used in CI, and adds machine-spec logging at startup so
runner capacity (CPU/RAM/disk) is visible in build logs.
## What Changed
- Updated the `build` job to support matrix entries that provide a full
`runs_on` object:
- `runs-on: ${{ matrix.runs_on || matrix.runner }}`
- Switched Windows release matrix entries to Codex runners:
- `windows-latest` -> `windows-x64` with:
- `group: codex-runners`
- `labels: codex-windows-x64`
- `windows-11-arm` -> `windows-arm64` with:
- `group: codex-runners`
- `labels: codex-windows-arm64`
- Updated the ARM-specific zstd install condition to match the new
runner id:
- `matrix.runner == 'windows-arm64'`
- Added early platform-specific runner diagnostics steps
(Linux/macOS/Windows) that print OS, CPU, logical CPU count, total RAM,
and disk usage.
1. Move Windows Sandbox NUX to right after trust directory screen
2. Don't offer read-only as an option in Sandbox NUX.
Elevated/Legacy/Quit
3. Don't allow new untrusted directories. It's trust or quit
4. move experimental sandbox features to `[windows]
sandbox="elevated|unelevatd"`
5. Copy tweaks = elevated -> default, non-elevated -> non-admin
## Summary
- Increases `PASTE_BURST_CHAR_INTERVAL` from 8ms to 30ms on Windows to
fix multi-line paste issues in VS Code integrated terminal
- Follows existing pattern of platform-specific timing (like
`PASTE_BURST_ACTIVE_IDLE_TIMEOUT`)
## Problem
When pasting multi-line text in Codex CLI on Windows (especially VS Code
integrated terminal), only the first portion is captured before
auto-submit. The rest arrives as a separate message.
**Root cause**: VS Code's terminal emulation adds latency (~10-15ms per
character) between key events. The 8ms `PASTE_BURST_CHAR_INTERVAL`
threshold is too tight - characters arrive slower than expected, so
burst detection fails and Enter submits instead of inserting a newline.
## Solution
Use Windows-specific timing (30ms) for `PASTE_BURST_CHAR_INTERVAL`,
following the same pattern already used for
`PASTE_BURST_ACTIVE_IDLE_TIMEOUT` (60ms on Windows vs 8ms on Unix).
30ms is still fast enough to distinguish paste from typing (humans type
~200ms between keystrokes).
## Test plan
- [x] All existing paste_burst tests pass
- [ ] Test multi-line paste in VS Code integrated PowerShell on Windows
- [ ] Test multi-line paste in standalone Windows PowerShell
- [ ] Verify no regression on macOS/Linux
Fixes#2137
Co-authored-by: Josh McKinney <joshka@openai.com>
Reapply "Add app-server transport layer with websocket support" with
additional fixes from https://github.com/openai/codex/pull/11313/changes
to avoid deadlocking.
This reverts commit 47356ff83c.
## Summary
To avoid deadlocking when queues are full, we maintain separate tokio
tasks dedicated to incoming vs outgoing event handling
- split the app-server main loop into two tasks in
`run_main_with_transport`
- inbound handling (`transport_event_rx`)
- outbound handling (`outgoing_rx` + `thread_created_rx`)
- separate incoming and outgoing websocket tasks
## Validation
Integration tests, testing thoroughly e2e in codex app w/ >10 concurrent
requests
<img width="1365" height="979" alt="Screenshot 2026-02-10 at 2 54 22 PM"
src="https://github.com/user-attachments/assets/47ca2c13-f322-4e5c-bedd-25859cbdc45f"
/>
---------
Co-authored-by: jif-oai <jif@openai.com>
`codex-core` had accumulated config loading, requirements parsing,
constraint logic, and config-layer state handling in a single crate.
This change extracts that subsystem into `codex-config` to reduce
`codex-core` rebuild/test surface area and isolate future config work.
## What Changed
### Added `codex-config`
- Added new workspace crate `codex-rs/config` (`codex-config`).
- Added workspace/build wiring in:
- `codex-rs/Cargo.toml`
- `codex-rs/config/Cargo.toml`
- `codex-rs/config/BUILD.bazel`
- Updated lockfiles (`codex-rs/Cargo.lock`, `MODULE.bazel.lock`).
- Added `codex-core` -> `codex-config` dependency in
`codex-rs/core/Cargo.toml`.
### Moved config internals from `core` into `config`
Moved modules to `codex-rs/config/src/`:
- `core/src/config/constraint.rs` -> `config/src/constraint.rs`
- `core/src/config_loader/cloud_requirements.rs` ->
`config/src/cloud_requirements.rs`
- `core/src/config_loader/config_requirements.rs` ->
`config/src/config_requirements.rs`
- `core/src/config_loader/fingerprint.rs` -> `config/src/fingerprint.rs`
- `core/src/config_loader/merge.rs` -> `config/src/merge.rs`
- `core/src/config_loader/overrides.rs` -> `config/src/overrides.rs`
- `core/src/config_loader/requirements_exec_policy.rs` ->
`config/src/requirements_exec_policy.rs`
- `core/src/config_loader/state.rs` -> `config/src/state.rs`
`codex-config` now re-exports this surface from `config/src/lib.rs` at
the crate top level.
### Updated `core` to consume/re-export `codex-config`
- `core/src/config_loader/mod.rs` now imports/re-exports config-loader
types/functions from top-level `codex_config::*`.
- Local moved modules were removed from `core/src/config_loader/`.
- `core/src/config/mod.rs` now re-exports constraint types from
`codex_config`.
We're loading these from the web on every startup. This puts them in a
local file with a 1hr TTL.
We sign the downloaded requirements with a key compiled into the Codex
CLI to prevent unsophisticated tampering (determined circumvention is
outside of our threat model: after all, one could just compile Codex
without any of these checks).
If any of the following are true, we ignore the local cache and re-fetch
from Cloud:
* The signature is invalid for the payload (== requirements, sign time,
ttl, user identity)
* The identity does not match the auth'd user's identity
* The TTL has expired
* We cannot parse requirements.toml from the payload
We are removing feature-gated shared crates from the `codex-rs`
workspace. `codex-common` grouped several unrelated utilities behind
`[features]`, which made dependency boundaries harder to reason about
and worked against the ongoing effort to eliminate feature flags from
workspace crates.
Splitting these utilities into dedicated crates under `utils/` aligns
this area with existing workspace structure and keeps each dependency
explicit at the crate boundary.
## What changed
- Removed `codex-rs/common` (`codex-common`) from workspace members and
workspace dependencies.
- Added six new utility crates under `codex-rs/utils/`:
- `codex-utils-cli`
- `codex-utils-elapsed`
- `codex-utils-sandbox-summary`
- `codex-utils-approval-presets`
- `codex-utils-oss`
- `codex-utils-fuzzy-match`
- Migrated the corresponding modules out of `codex-common` into these
crates (with tests), and added matching `BUILD.bazel` targets.
- Updated direct consumers to use the new crates instead of
`codex-common`:
- `codex-rs/cli`
- `codex-rs/tui`
- `codex-rs/exec`
- `codex-rs/app-server`
- `codex-rs/mcp-server`
- `codex-rs/chatgpt`
- `codex-rs/cloud-tasks`
- Updated workspace lockfile entries to reflect the new dependency graph
and removal of `codex-common`.
Improve listing by doing:
1. List using the rollout file system
2. Upsert the result in the DB (if present)
3. Return the result of a DB listing
4. Fallback on the result of 1
+ some metrics on top of this
stage1_concurrent_claims_respect_running_cap was flaky due to SQLite
lock contention, not cap logic correctness. The claim flow used deferred
transactions (BEGIN) with read-then-write behavior, which can fail under
concurrency with SQLITE_BUSY_SNAPSHOT/database is locked when upgrading
a read transaction to a write transaction. We fixed this by using BEGIN
IMMEDIATE for stage1 and phase2 claim paths, so lock acquisition happens
up front and contenders serialize cleanly instead of failing during
upgrade. After the change, codex-state tests pass and stress reruns of
the flaky path no longer reproduced the failure.
## Why
`codex-core` was being built in multiple feature-resolved permutations
because test-only behavior was modeled as crate features. For a large
crate, those permutations increase compile cost and reduce cache reuse.
## Net Change
- Removed the `test-support` crate feature and related feature wiring so
`codex-core` no longer needs separate feature shapes for test consumers.
- Standardized cross-crate test-only access behind
`codex_core::test_support`.
- External test code now imports helpers from
`codex_core::test_support`.
- Underlying implementation hooks are kept internal (`pub(crate)`)
instead of broadly public.
## Outcome
- Fewer `codex-core` build permutations.
- Better incremental cache reuse across test targets.
- No intended production behavior change.
The debug output listed non-file-backed layers such as session flags and
MDM managed config, but it did not show their values. That made it
difficult to explain unexpected effective settings because users could
not inspect those layers on disk.
Now `/debug-config` might include output like this:
```
Config layer stack (lowest precedence first):
1. system (/etc/codex/config.toml) (enabled)
2. user (/Users/mbolin/.codex/config.toml) (enabled)
3. legacy managed_config.toml (mdm) (enabled)
MDM value:
# Production Codex configuration file.
[otel]
log_user_prompt = true
environment = "prod"
exporter = { otlp-http = {
endpoint = "https://example.com/otel",
protocol = "binary"
}}
```
Added multi-limit support end-to-end by carrying limit_name in
rate-limit snapshots and handling multiple buckets instead of only
codex.
Extended /usage client parsing to consume additional_rate_limits
Updated TUI /status and in-memory state to store/render per-limit
snapshots
Extended app-server rate-limit read response: kept rate_limits and added
rate_limits_by_name.
Adjusted usage-limit error messaging for non-default codex limit buckets
Problem:
1. turn id is constructed in-memory;
2. on resuming threads, turn_id might not be unique;
3. client cannot no the boundary of a turn from rollout files easily.
This PR does three things:
1. persist `task_started` and `task_complete` events;
1. persist `turn_id` in rollout turn events;
5. generate turn_id as unique uuids instead of incrementing it in
memory.
This helps us resolve the issue of clients wanting to have unique turn
ids for resuming a thread, and knowing the boundry of each turn in
rollout files.
example debug logs
```
2026-02-11T00:32:10.746876Z DEBUG codex_app_server_protocol::protocol::thread_history: built turn from rollout items turn_index=8 turn=Turn { id: "019c4a07-d809-74c3-bc4b-fd9618487b4b", items: [UserMessage { id: "item-24", content: [Text { text: "hi", text_elements: [] }] }, AgentMessage { id: "item-25", text: "Hi. I’m in the workspace with your current changes loaded and ready. Send the next task and I’ll execute it end-to-end." }], status: Completed, error: None }
2026-02-11T00:32:10.746888Z DEBUG codex_app_server_protocol::protocol::thread_history: built turn from rollout items turn_index=9 turn=Turn { id: "019c4a18-1004-76c0-a0fb-a77610f6a9b8", items: [UserMessage { id: "item-26", content: [Text { text: "hello", text_elements: [] }] }, AgentMessage { id: "item-27", text: "Hello. Ready for the next change in `codex-rs`; I can continue from the current in-progress diff or start a new task." }], status: Completed, error: None }
2026-02-11T00:32:10.746899Z DEBUG codex_app_server_protocol::protocol::thread_history: built turn from rollout items turn_index=10 turn=Turn { id: "019c4a19-41f0-7db0-ad78-74f1503baeb8", items: [UserMessage { id: "item-28", content: [Text { text: "hello", text_elements: [] }] }, AgentMessage { id: "item-29", text: "Hello. Send the specific change you want in `codex-rs`, and I’ll implement it and run the required checks." }], status: Completed, error: None }
```
backward compatibility:
if you try to resume an old session without task_started and
task_complete event populated, the following happens:
- If you resume and do nothing: those reconstructed historical IDs can
differ next time you resume.
- If you resume and send a new turn: the new turn gets a fresh UUID from
live submission flow and is persisted, so that new turn’s ID is stable
on later resumes.
I think this behavior is fine, because we only care about deterministic
turn id once a turn is triggered.
## Summary
This should rarely, if ever, happen in practice. But regardless, we
should never provide an empty list of `commands` to ExecPolicy. This PR
is almost entirely adding test around these cases.
## Testing
- [x] Adds a bunch of unit tests for this
