LangChain: Reasoning Fixes and Decisions API Beta
Fireworks shipped a fix to preserve reasoning across streaming and tool calls in 1.7.1, while Anthropic's Haiku 5.5 proposal extends the same pattern. Core also saw a cluster of crash-prevention fixes and a beta wrapper for OpenAI's Decisions API.
Duration: PT2M26S
Episode overview
This episode is a short developer briefing from LangChain.
It explains recent repository work in plain language.
- Show: LangChain
- Published: 2026-10-08T13:02:55Z
- Audio duration: PT2M26S
Transcript excerpt
This excerpt keeps the crawler page concise. Listen to the episode or use the RSS feed for the full update.
It's October 8th, 2026, and here's what's moving in LangChain.
The lead pattern is reasoning becoming durable context, not disposable output. Merged PR 41093 fixes Chat Fireworks dropping streamed reasoning and omitting it from follow-up tool requests, while adding reasoning token counts to usage metadata. That fix shipped to users in the 1.7.1 patch release, PR 41137, with a…
The same theme shows up in open proposal PR 41140 for Claude Haiku 5.5, which keeps streamed thinking blocks intact for replay, including redacted thinking, and preserves forced tool choices for bound tools and structured output. For developers, the impact is reliability in agentic loops: models that reason before…
A second theme is core hardening against crashes and silent data loss. PR 41148 stops merge message runs from raising on consecutive chat or function messages with different roles or names, keeping them separate instead. PR 41147 fixes a Key Error in convert to Open AI messages for Bedrock Converse blocks missing a…
Finally, new surface area: PR 41125 proposes a beta wrapper for OpenAI's Decisions API with LangChain message support, LangSmith tracing, and prebuilt middleware, though without a shared…
W…
Nearby episodes from LangChain
- Weekly Recap - Profiles, Decisions, and Message Reliability
- Safer Agent Sandboxing and Measurement Fixes
- Mid-Conversation Profiles and Splitter Fixes
- Decisions API Beta and Agent Reliability Fixes
- Streaming Fidelity and Core Edge Cases
- Sharper Agent Guardrails and Credential Hardening
- Credential Leak Fixes and Required OpenAI Model
- Weekly Recap - OpenAI Defaults Retire, Agent Safety Hardens