LangChain: Per-Call Approvals and Leaner Streaming
LangChain 1.4.6 adds opt-in per-call human-in-the-loop interrupts, while streaming work targets quadratic tool-call growth and dropped Responses answers. Small correctness fixes round out evaluation feedback, XML parsing, and model profiles.
Duration: PT2M21S
Episode overview
This episode is a short developer briefing from LangChain.
It explains recent repository work in plain language.
- Show: LangChain
- Published: 2026-10-15T13:02:52Z
- Audio duration: PT2M21S
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 15th, 2026, and this is your LangChain briefing.
The big pattern today is safer, cheaper agent runs — tighter human approvals, and leaner, more reliable streaming.
First, human approvals get granular. Pull request 41062, now released in LangChain 1.4.6, adds an opt-in per-call mode to the human-in-the-loop middleware. The default batched mode still pauses once per model turn with all gated calls in one interrupt matched by position. That forced UIs to guess what was being…
Second, streaming tool calls gets fixed on two fronts. One open proposal, pull request 41246, would teach Python stream consumers to read tool call deltas instead of full repeated argument strings. Today each delta repeats everything so far, so memory and wire bytes grow with the square of the arguments — a…
Finally, three quiet correctness fixes. Pull request 41251 stops evaluation feedback config from being silently dropped before reaching LangSmith. Pull request 41255 limits the X-M-L parser's encoding check to real declarations so normal elements with an encoding attribute parse correctly. And pull request 41249…
What's next: test per-call approvals behind the opt-in flag, and watch for the…
Nearby episodes from LangChain
- Streaming Fixes, Retry Misalignment, and Python 3.10 Removal
- Cheaper Streaming and Safer Agent Recovery
- Agent Steering, Shell Reliability, and Content Fixes
- Mid-Conversation Releases and Flexible Summarization
- Tool Message Fixes and Mid-Conversation Releases
- Weekly Recap - Profiles, Decisions, and Message Reliability
- Safer Agent Sandboxing and Measurement Fixes
- Mid-Conversation Profiles and Splitter Fixes