fix: Display sequential interrupts in multi-turn conversations

[hardiknahata]

Fixes an issue where only the first interrupt in a multi-turn conversation would display in the UI. Subsequent interrupts were processed correctly by the backend but failed to render in the interface.

Problem

The interrupt rendering logic only displayed interrupts when attached to the last message or when no AI/tool messages existed. When users responded to an interrupt, new messages were added to the thread, causing the previous message to no longer be "last", which prevented subsequent interrupts from displaying.

Solution

Thread-level interrupt rendering (index.tsx)

• Added dedicated interrupt component rendering after all messages
• Renders when: interrupt exists, AI/tool messages present, and not loading
• Uses unique key "interrupt-msg-after-messages" to prevent conflicts

Interrupt component updates (ai.tsx)

• Added isInterruptOnly prop to identify interrupt-only component renders
• Updated Interrupt rendering conditions to include isInterruptOnly
• When message === undefined, component renders only the interrupt content

This preserves existing behavior for message-attached interrupts while enabling sequential interrupts to render independently.

Testing

Verified with a LangGraph agent triggering multiple sequential interrupts:

• ✅ First interrupt displays correctly
• ✅ User responds to first interrupt
• ✅ Second interrupt displays in UI (previously broken)
• ✅ Multi-turn interrupt conversations work seamlessly

Enables Agent Chat to support graphs requiring multiple human-in-the-loop interactions, matching LangGraph Studio behavior.

[vercel]

@hardiknahata is attempting to deploy a commit to the LangChain Team on Vercel.

A member of the Team first needs to authorize it.