$ the-wire · showcase
Codex lifts the sample prompt limit and moves Windows sandbox onto MXC SDK V1
By RepoJournal · Filed · About OpenAI · Composed from the cited sources · methodology
Codex loosened a request-validation cap on thread/sample, rewired its Windows sandbox to the public MXC V1 SDK, and the SDKs shipped routine releases, with openai-node 7.60.0 adding Live MCP server configuration.
Remove the 4,000-byte prompt limit for `thread/sample` (#62378) openai/codex
Prompts longer than 4,000 bytes now pass thread/sample request validation; only expectedTurnId and prompt must still be nonblank, so clients no longer have to truncate or reject long prompts before sending.
Migrate the Windows MXC sandbox to the public V1 SDK (#62331) openai/codex
The Windows MXC sandbox now launches commands in a ProcessContainer via the public mxc-sdk V1 API, building the request from Codex's filesystem, network, UI, and environment settings. The SDK now rejects deny paths the host cannot enforce, and Codex's direct flatbuffers dependency is gone. This is breaking: hosts must satisfy the V1 contract instead of relying on the old path.
v7.60.0 openai/openai-node
by openai-sdks[bot]
Version 7.60.0 adds Live MCP server configuration and standalone CSAM moderation types to the API surface.
Prepare managed proxy settings once per sandbox command openai/codex
SandboxManager now prepares the proxy environment and sandbox network context together when no prepared route is supplied, so relaunch no longer overwrites settings bound to a command, including credentials prepared for its shell.
Expose custom status lines in `/experimental` openai/codex
Custom status line moves out of the TUI client's feature flag and into /experimental, with the toggle saved on the connected app server; restart the TUI and use /statusline to pick a provider. Elsewhere the SDKs were quiet: openai-python 3.55.1 documents CSAM moderation response fields and bumps pytest-asyncio, and Dependabot refreshed dev dependencies across both repos.