LangChain AI Updates: September 27, 2026

1. LangChain Pairs TypeSafe’s Jev Decision Model With LangGraph for Cheap Routing Steps

LangChain. LangChain published a guide to building agents that hand narrow classification and routing calls to Jev, TypeSafe AI’s decision model that returns typed answers with probabilities instead of generated text, while LangGraph handles orchestration, checkpointing, and human-in-the-loop interrupts. The worked example is a litigation document review graph where Jev answers three questions per page (responsive, contains PII, possibly privileged) and escalates only to an LLM for redaction or to an attorney review interrupt. In LangChain’s runs, Jev was 5 to 6 times faster than Claude Sonnet on the classification step, and LangSmith now has dedicated trace views that show the inputs and calibrated outputs for each decision. Source