langchain-ai/langchainjs langchain@1.5.18 on GitHub

#11877 2f4d891 Thanks @byhow! - feat(agents): add opt-in interruptMode: "per_call" to humanInTheLoopMiddleware. Each gated tool call raises its own interrupt (type: "tool_approval", with a typed responseSchema built from the tool's allowed decisions), answered by interrupt ID. respond is supported. Edits are pinned to the called tool and checked against the input side of its Zod v4 schema, so its transforms run once, when the tool does. An invalid answer throws without being saved. Also adds Interrupt.response_schema. In both modes, interruptOn: { tool: true } now allows every decision, respond included, as in Python; batched mode is otherwise unchanged.