mirror of https://github.com/openai/codex.git synced 2026-04-26 23:55:25 +00:00

Files

Shaqayeq 35d03a0508 2026-03-07 Stabilize experimental python SDK foundation

Description:
- pin datamodel-code-generator to an exact version and invoke it through the active interpreter
- make generate_types own the maintained generated surfaces and regenerate committed artifacts
- make sdk/python tests hermetic and regeneration checks idempotent

2026-03-08 00:47:14 -08:00

2.3 KiB

Raw Blame History

FAQ

Thread vs turn

A Thread is conversation state.
A Turn is one model execution inside that thread.
Multi-turn chat means multiple turns on the same Thread.

`run()` vs `stream()`

Turn.run() is the easiest path. It consumes events until completion and returns TurnResult.
Turn.stream() yields raw notifications (Notification) so you can react event-by-event.

Choose run() for most apps. Choose stream() for progress UIs, custom timeout logic, or custom parsing.

Sync vs async clients

Codex is the minimal sync SDK and best default.
AsyncAppServerClient wraps the sync transport with asyncio.to_thread(...) for async-friendly call sites.

If your app is not already async, stay with Codex.

`thread(...)` vs `thread_resume(...)`

codex.thread(thread_id) only binds a local helper to an existing thread ID.
codex.thread_resume(thread_id, ...) performs a thread/resume RPC and can apply overrides (model, instructions, sandbox, etc.).

Use thread(...) for simple continuation. Use thread_resume(...) when you need explicit resume semantics or override fields.

Why does constructor fail?

Codex() is eager: it starts transport and calls initialize in __init__.

Common causes:

bundled runtime binary missing for your OS/arch under src/codex_app_server/bin/*
local auth/session is missing
incompatible/old app-server

Maintainers can refresh bundled binaries with:

cd sdk/python
python scripts/update_sdk_artifacts.py --channel stable --bundle-all-platforms

Why does a turn "hang"?

A turn is complete only when turn/completed arrives for that turn ID.

run() waits for this automatically.
With stream(), make sure you keep consuming notifications until completion.

How do I retry safely?

Use retry_on_overload(...) for transient overload failures (ServerBusyError).

Do not blindly retry all errors. For InvalidParamsError or MethodNotFoundError, fix inputs/version compatibility instead.

Common pitfalls

Starting a new thread for every prompt when you wanted continuity.
Forgetting to close() (or not using with Codex() as codex:).
Ignoring TurnResult.status and TurnResult.error.
Mixing SDK input classes with raw dicts incorrectly in minimal API paths.

2.3 KiB Raw Blame History

FAQ