## Overview
Adds LM Studio OSS support. Closes#1883
### Changes
This PR enhances the behavior of `--oss` flag to support LM Studio as a
provider. Additionally, it introduces a new flag`--local-provider` which
can take in `lmstudio` or `ollama` as values if the user wants to
explicitly choose which one to use.
If no provider is specified `codex --oss` will auto-select the provider
based on whichever is running.
#### Additional enhancements
The default can be set using `oss-provider` in config like:
```
oss_provider = "lmstudio"
```
For non-interactive users, they will need to either provide the provider
as an arg or have it in their `config.toml`
### Notes
For best performance, [set the default context
length](https://lmstudio.ai/docs/app/advanced/per-model) for gpt-oss to
the maximum your machine can support
---------
Co-authored-by: Matt Clayton <matt@lmstudio.ai>
Co-authored-by: Eric Traut <etraut@openai.com>
We're running into quite a bit of drag maintaining this test, since
every time we add fields to an EventMsg that happened to be dumped into
the `binary-size-log.jsonl` fixture, this test starts to fail. The fix
is usually to either manually update the `binary-size-log.jsonl` fixture
file, or update the `upgrade_event_payload_for_tests` function to map
the data in that file into something workable.
Eason says it's fine to delete this test, so let's just delete it
## Summary
- Introduces the `codex-execpolicy2` crate.
- This PR covers only the prefix-rule subset of the planned execpolicy
v2 language; a richer language will follow.
## Policy
- Policy language centers on `prefix_rule(pattern=[...], decision?,
match?, not_match?)`, where `pattern` is an ordered list of tokens; any
element may be a list to denote alternatives. `decision` defaults to
`allow`; valid values are `allow`, `prompt`, and `forbidden`. `match` /
`not_match` hold example commands that are tokenized and validated at
load time (think of these as unit tests).
## Policy shapes
- Prefix rules use Starlark syntax:
```starlark
prefix_rule(
pattern = ["cmd", ["alt1", "alt2"]], # ordered tokens; list entries denote alternatives
decision = "prompt", # allow | prompt | forbidden; defaults to allow
match = [["cmd", "alt1"]], # examples that must match this rule (enforced at compile time)
not_match = [["cmd", "oops"]], # examples that must not match this rule (enforced at compile time)
)
```
## Response shapes
- Match:
```json
{
"match": {
"decision": "allow|prompt|forbidden",
"matchedRules": [
{
"prefixRuleMatch": {
"matchedPrefix": ["<token>", "..."],
"decision": "allow|prompt|forbidden"
}
}
]
}
}
```
- No match:
```json
"noMatch"
```
- `matchedRules` lists every rule whose prefix matched the command;
`matchedPrefix` is the exact prefix that matched.
- The effective `decision` is the strictest severity across all matches
(`forbidden` > `prompt` > `allow`).
---------
Co-authored-by: Michael Bolin <mbolin@openai.com>
## Summary
Builds on FreeBSD and OpenBSD were failing due to globally enabled
Linux-specific keyring features and hardening code paths not gated by
OS. This PR scopes keyring native backends to the
appropriate targets, disables default features at the workspace root,
and adds a BSD-specific hardening function. Linux/macOS/Windows behavior
remains unchanged, while FreeBSD/OpenBSD
now build and run with a supported backend.
## Key Changes
- Keyring features:
- Disable keyring default features at the workspace root to avoid
pulling Linux backends on non-Linux.
- Move native backend features into target-specific sections in the
affected crates:
- Linux: linux-native-async-persistent
- macOS: apple-native
- Windows: windows-native
- FreeBSD/OpenBSD: sync-secret-service
- Process hardening:
- Add pre_main_hardening_bsd() for FreeBSD/OpenBSD, applying:
- Set RLIMIT_CORE to 0
- Clear LD_* environment variables
- Simplify process-hardening Cargo deps to unconditional libc (avoid
conflicting OS fragments).
- No changes to CODEX_SANDBOX_* behavior.
## Rationale
- Previously, enabling keyring native backends globally pulled
Linux-only features on BSD, causing build errors.
- Hardening logic was tailored for Linux/macOS; BSD builds lacked a
gated path with equivalent safeguards.
- Target-scoped features and BSD hardening make the crates portable
across these OSes without affecting existing behavior elsewhere.
## Impact by Platform
- Linux: No functional change; backends now selected via target cfg.
- macOS: No functional change; explicit apple-native mapping.
- Windows: No functional change; explicit windows-native mapping.
- FreeBSD/OpenBSD: Builds succeed using sync-secret-service; BSD
hardening applied during startup.
## Testing
- Verified compilation across affected crates with target-specific
features.
- Smoke-checked that Linux/macOS/Windows feature sets remain identical
functionally after scoping.
- On BSD, confirmed keyring resolves to sync-secret-service and
hardening compiles.
## Risks / Compatibility
- Minimal risk: only feature scoping and OS-gated additions.
- No public API changes in the crates; runtime behavior on non-BSD
platforms is preserved.
- On BSD, the new hardening clears LD_*; this is consistent with
security posture on other Unix platforms.
## Reviewer Notes
- Pay attention to target-specific sections for keyring in the affected
Cargo.toml files.
- Confirm pre_main_hardening_bsd() mirrors the safe subset of
Linux/macOS hardening without introducing Linux-only calls.
- Confirm no references to CODEX_SANDBOX_ENV_VAR or
CODEX_SANDBOX_NETWORK_DISABLED_ENV_VAR were added/modified.
## Checklist
- Disable keyring default features at workspace root.
- Target-specific keyring features mapped per OS
(Linux/macOS/Windows/BSD).
- Add BSD hardening (RLIMIT_CORE=0, clear LD_*).
- Simplify process-hardening dependencies to unconditional libc.
- No changes to sandbox env var code.
- Formatting and linting: just fmt + just fix -p for changed crates.
- Project tests pass for changed crates; broader suite unchanged.
---------
Co-authored-by: celia-oai <celia@openai.com>
## Summary
Fixes streaming issue where Claude models return only 1-4 characters
instead of full responses when used through certain API
providers/proxies.
## Environment
- **OS**: Windows
- **Models affected**: Claude models (e.g., claude-haiku-4-5-20251001)
- **API Provider**: AAAI API proxy (https://api.aaai.vip/v1)
- **Working models**: GLM, Google models work correctly
## Problem
When using Claude models in both TUI and exec modes, only 1-4 characters
are displayed despite the backend receiving the full response. Debug
logs revealed that some API providers send SSE chunks with an empty
string finish_reason during active streaming, rather than null or
omitting the field entirely.
The current code treats any non-null finish_reason as a termination
signal, causing the stream to exit prematurely after the first chunk.
The problematic chunks contain finish_reason with an empty string
instead of null.
## Solution
Fix empty finish_reason handling in chat_completions.rs by adding a
check to only process non-empty finish_reason values. This ensures empty
strings are ignored and streaming continues normally.
## Testing
- Tested on Windows with Claude Haiku model via AAAI API proxy
- Full responses now received and displayed correctly in both TUI and
exec modes
- Other models (GLM, Google) continue to work as expected
- No regression in existing functionality
## Impact
- Improves compatibility with API providers that send empty
finish_reason during streaming
- Enables Claude models to work correctly in Windows environment
- No breaking changes to existing functionality
## Related Issues
This fix resolves the issue where Claude models appeared to return
incomplete responses. The root cause was identified as a compatibility
issue in parsing SSE responses from certain API providers/proxies,
rather than a model-specific problem. This change improves overall
robustness when working with various API endpoints.
---------
Co-authored-by: Eric Traut <etraut@openai.com>
### Summary
- Treat AltGr chords (Ctrl+Alt) as literal character input in the Codex
TUI textarea so Windows terminals that report
backslash and other characters via AltGr insert correctly.
- Add regression test altgr_ctrl_alt_char_inserts_literal to ensure
Ctrl+Alt char events append the character and
advance the cursor.
### Motivation
On US/UK keyboard layouts, backslash is produced by a plain key, so
Ctrl+Alt handling is never exercised and the
bug isn’t visible. On many non‑US layouts (e.g., German), backslash and
other symbols require AltGr, which terminals
report as Ctrl+Alt+<char>. Our textarea previously filtered these chords
like navigation bindings, so AltGr input was
dropped on affected layouts. This change treats AltGr chords as literal
input so backslash and similar symbols work on
Windows terminals.
This fixes multiple reported Issues where the \ symbol got cut off.
Like:
C:\Users\Admin
became
C:UsersAdmin
Co-authored-by: Eric Traut <etraut@openai.com>
## What?
Fixes MCP server initialization failures on Windows when using
script-based tools like `npx`, `pnpm`, and `yarn` that rely on
`.cmd`/`.bat` files rather than `.exe` binaries.
Fixes#2945
## Why?
Windows users encounter "program not found" errors when configuring MCP
servers with commands like `npx` in their `~/.codex/config.toml`. This
happens because:
- Tools like `npx` are batch scripts (`npx.cmd`) on Windows, not
executable binaries
- Rust's `std::process::Command` bypasses the shell and cannot execute
these scripts directly
- The Windows shell normally handles this by checking `PATHEXT` for
executable extensions
Without this fix, Windows users must specify full paths or add `.cmd`
extensions manually, which breaks cross-platform compatibility.
## How?
Added platform-specific program resolution using the `which` crate to
find the correct executable path:
- **Windows**: Resolves programs through PATH/PATHEXT to find
`.cmd`/`.bat` scripts
- **Unix**: Returns the program unchanged (no-op, as Unix handles
scripts natively)
### Changes
- Added `which = "6"` dependency to `mcp-client/Cargo.toml`
- Implemented `program_resolver` module in `mcp_client.rs` with
platform-specific resolution
- Added comprehensive tests for both Windows and Unix behavior
### Testing
Added platform-specific tests to verify:
- Unix systems execute scripts without extensions
- Windows fails without proper extensions
- Windows succeeds with explicit extensions
- Cross-platform resolution enables successful execution
**Tested on:**
- Windows 11 (NT 10.0.26100.0 x64)
- PowerShell 5.1 & 7+, CMD, Git Bash
- MCP servers: playwright, context7, supabase
- WSL (verified no regression)
**Local checks passed:**
```bash
cargo test && cargo clippy --tests && cargo fmt -- --config imports_granularity=Item
```
### Results
**Before:**
```
🖐 MCP client for `playwright` failed to start: program not found
```
**After:**
```
🖐 MCP client for `playwright` failed to start: request timed out
```
Windows users can now use simple commands like `npx` in their config
without specifying full paths or extensions. The timeout issue is a
separate concern that will be addressed in a follow-up PR.
---------
Co-authored-by: Eric Traut <etraut@openai.com>
This PR does the following:
- Add compact prefix to the summary
- Change the compaction prompt
- Allow multiple compaction for long running tasks
- Filter out summary messages on the following compaction
Considerations:
- Filtering out the summary message isn't the most clean
- Theoretically, we can end up in infinite compaction loop if the user
messages > compaction limit . However, that's not possible in today's
code because we have hard cap on user messages.
- We need to address having multiple user messages because it confuses
the model.
Testing:
- Making sure that after compact we always end up with one user message
(task) and one summary, even on multiple compaction.
Fixes#4940Fixes#4892
When selecting "No, ask me to approve edits and commands" during
onboarding, the code wasn't applying the correct approval policy,
causing Codex to block all write operations instead of requesting
approval.
This PR fixes the issue by persisting the "DontTrust" decision in
config.toml as `trust_level = "untrusted"` and handling it in the
sandbox and approval policy logic, so Codex correctly asks for approval
before making changes.
## Before (bug)
<img width="709" height="500" alt="bef"
src="https://github.com/user-attachments/assets/5aced26d-d810-4754-879a-89d9e4e0073b"
/>
## After (fixed)
<img width="713" height="359" alt="aft"
src="https://github.com/user-attachments/assets/9887bbcb-a9a5-4e54-8e76-9125a782226b"
/>
---------
Co-authored-by: Eric Traut <etraut@openai.com>
For better caching performance all output items should be rendered in
the order they were produced before all new input items (for example,
all function_call before all function_call_output).
For app-server development it's been helpful to be able to trigger some
test flows end-to-end and print the JSON-RPC messages sent between
client and server.
## Summary
- default the `tui.notifications` setting to enabled so desktop
notifications work out of the box
- update configuration tests and documentation to reflect the new
default
## Testing
- `cargo test -p codex-core` *(fails:
`exec::tests::kill_child_process_group_kills_grandchildren_on_timeout`
is flaky in this sandbox because the spawned grandchild process stays
alive)*
- `cargo test -p codex-core
exec::tests::kill_child_process_group_kills_grandchildren_on_timeout`
*(fails: same sandbox limitation as above)*
------
[Codex
Task](https://chatgpt.com/codex/tasks/task_i_69166f811144832c9e8aaf8ee2642373)
## Summary
- add `TestCodex::submit_turn_with_policies` and extend the response
helpers with reusable tool-call utilities
- update the grep_files, read_file, list_dir, shell_serialization, and
tools suites to rely on the shared helpers instead of local copies
- make the list_dir helper return `anyhow::Result` so clippy no longer
warns about `expect`
## Testing
- `just fix -p codex-core`
- `cargo test -p codex-core --test all
suite::grep_files::grep_files_tool_collects_matches`
- `cargo test -p codex-core
suite::grep_files::grep_files_tool_collects_matches -- --ignored`
(filter requests ignored tests so nothing runs, but the build stays
clean)
------
[Codex
Task](https://chatgpt.com/codex/tasks/task_i_69112d53abac83219813cab4d7cb6446)
## Summary
- replace the bespoke network check in the chat completion payload and
SSE tests with the existing `skip_if_no_network!` helper so they follow
the same gating convention as the rest of the suite
## Testing
- `just fmt`
------
[Codex
Task](https://chatgpt.com/codex/tasks/task_i_69112d4cb9f08321ba773e8ccf39778e)
A partner is consuming our generated JSON schema bundle for app-server
and identified a few issues:
- not all polymorphic / one-of types have a type descriminator
- `"$ref": "#/definitions/v2/SandboxPolicy"` is missing
- "Option<>" is an invalid schema name, and also unnecessary
This PR:
- adds the type descriminator to the various types that are missing it
except for `SessionSource` and `SubAgentSource` because they are
serialized to disk (adding this would break backwards compat for
resume), and they should not be necessary to consume for an integration
with app-server.
- removes the special handling in `export.rs` of various types like
SandboxPolicy, which turned out to be unnecessary and incorrect
- filters out `Option<>` which was auto-generated for request params
that don't need a body
For context, we currently pull in wayyy more types than we need through
the `EventMsg` god object which we are **not** planning to expose in API
v2 (this is how I suspect `SessionSource` and `SubAgentSource` are being
pulled in). But until we have all the necessary v2 notifications in
place that will allow us to remove `EventMsg`, we will keep exporting it
for now.
core event to app server event mapping:
1. `codex/event/reasoning_content_delta` ->
`item/reasoning/summaryTextDelta`.
2. `codex/event/reasoning_raw_content_delta` ->
`item/reasoning/textDelta`
3. `codex/event/agent_message_content_delta` →
`item/agentMessage/delta`.
4. `codex/event/agent_reasoning_section_break` ->
`item/reasoning/summaryPartAdded`.
Also added a change in core to pass down content index, summary index
and item id from events.
Tested with the `git checkout owen/app_server_test_client && cargo run
-p codex-app-server-test-client -- send-message-v2 "hello"` and verified
that new events are emitted correctly.
## Summary
Consolidates our apply_patch tests into one suite, and ensures each test
case tests the various ways the harness supports apply_patch:
1. Freeform custom tool call
2. JSON function tool
3. Simple shell call
4. Heredoc shell call
There are a few test cases that are specific to a particular variant,
I've left those alone.
## Testing
- [x] This adds a significant number of tests
## Summary
Adds support for specifying additional directories in the TypeScript SDK
through a new `additionalDirectories` option in `ThreadOptions`.
## Changes
- Added `additionalDirectories` parameter to `ThreadOptions` interface
- Updated `CodexExec` to accept and pass through additional directories
via the `--config` flag for `sandbox_workspace_write.writable_roots`
- Added comprehensive test coverage for the new functionality
## Test plan
- Added test case that verifies `additionalDirectories` is correctly
passed as repeated flags
- Existing tests continue to pass
---------
Co-authored-by: Claude <noreply@anthropic.com>
This fixes command display for unified exec. All `cd`s and `ls`es are
now parsed.
<img width="452" height="237" alt="image"
src="https://github.com/user-attachments/assets/ce92d81f-f74c-485a-9b34-1eaa29290ec6"
/>
Deletes a ton of tests that were doing nothing from shell.rs.
---------
Co-authored-by: Pavel Krymets <pavel@krymets.com>
Previously the popup measured rows using the full content width while
the renderer drew them with 2 columns of padding, so at certain widths
the layout allocated too little vertical space and hid the third option.
Now both desired_height and render call a shared helper that subtracts
the padding before measuring, so the height we reserve always matches
what we draw and the menu doesn't drops entries.
https://github.com/user-attachments/assets/59058fd9-1e34-4325-b5fe-fc888dfcb6bc
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.
