LangChain and LangSmith ship security and reliability fixes

langchain-openai 1.4.2 adds handling for context-window errors, filters LangChain-generated content-block identifiers, preserves Responses API text options and redacts MCP authorisation data. LangSmith SDK 0.10.17 sanitises credentials embedded in Git remote URLs and adds replica information to reconstructed RunTrees. These are maintenance releases, not new reasoning models.

Impact: credential redaction, faithful option propagation and trace reconstruction are production concerns for tool-using agents. Teams should upgrade through a staged test, validate redaction in exported traces and confirm retry behaviour around context overflows.