LangChain has released version 1.3.17, moving forward from 1.3.16 with a targeted bug fix and routine maintenance updates. This release is primarily aimed at users who rely on human-in-the-loop workflows.
The main change addresses a bug in how custom rejection reasons are framed during human-in-the-loop interactions. Previously, when a human reviewer rejected an action with a custom explanation, that reason could be lost or misrepresented in the chain’s output. With this fix, custom rejection reasons are now preserved and passed through correctly, so downstream logic and users can see exactly why a step was rejected. This matters for self-hosted deployments where audit trails and clear human feedback are essential.
Alongside the fix, the release includes routine dependency bumps for minor and patch versions. These updates help keep the library compatible with its ecosystem, address small compatibility issues, and reduce the risk of known vulnerabilities in underlying packages. They do not introduce breaking changes and are safe to apply as part of a regular maintenance cycle.
