$ the-wire · showcase
Codex ships a config probe, python fixes realtime and Pydantic v1
By RepoJournal · Filed · About OpenAI · Composed from the cited sources · methodology
Codex gained a way to validate its startup configuration without booting anything, while openai-python closed two bugs that were silently changing what callers get back on the wire.
fix(realtime): honor native options for legacy beta sockets openai/openai-python
Legacy client.beta.realtime.connect() inherited a 1 MiB native receive limit, so any larger event closed the socket with code 1009 unless callers had already passed max_size=None. Beta connections now default to no receive size limit while still honoring every caller value, including finite limits and zero, and the sync options type no longer advertises write_limit, which only native async conn...
fix: preserve declared aliases during Pydantic v1 validation openai/openai-python
Under Pydantic v1, a valid Responses function call carrying both "async": true and an extra "async_": "future" could have the validated boolean replaced by the extra value, and nested SDK models with other aliased fields hit the same collision. Validation now keeps the declared wire field and drops only the colliding extra, leaving unrelated extras and other declared aliases intact.
Add one-shot app-server startup configuration discovery (#55300) openai/codex
codex app-server --print-startup-config validates local and cloud-backed configuration and emits the version and configuration requirements as JSON without initializing environments, thread storage, or plugins; failures return structured JSON errors and a nonzero exit. A companion --startup-sign-out flag strips account credentials and reloads signed-out requirements.
Migrate the Windows MXC sandbox to `mxc-sdk` (#55378) openai/codex
The Windows MXC sandbox now builds against mxc-sdk instead of the separate MXC crates, with request construction adapted to the SDK's policy and optional environment fields. The native availability probe is preserved by caching process security environment creation and startup attribute initialization without launching a process.
feat(api): add 30-minute cache-write usage counts openai/openai-node
Completions and Decisions usage results gain an optional input_cache_write_30m_tokens metric that may be null or absent in older responses, alongside the existing cache-write counters. The remaining tail was mostly bookkeeping: a per-file inventory of the 47 files and 2,743 patch lines of custom edits over Castiron output, the same optional compaction trigger IDs landing in openai-python, a dup...
