LangChain's langchain-openai 1.5.1 fixes encrypted reasoning streaming
The patch keeps OpenAI's encrypted reasoning intact during streamed responses.
Deep Dive
LangChain's open-source repo — with 144k stars — just released langchain-openai 1.5.1. This update fixes an issue to preserve streamed encrypted reasoning and adds LangSmith gateway support for CI, per the release notes.
Key Points
- Fixes streamed encrypted reasoning loss (PR #39635) for OpenAI models like o1
- Patch bumps langchain-openai from 1.5.0 to 1.5.1, backwards-compatible
- Adds LangSmith gateway support in CI (PR #39651) for easier testing
Why It Matters
Reliable streaming of encrypted reasoning is critical for agents and production LLM apps that parse token-by-token output.