langchain-ai/langgraph
1
0
Fork 0
mirror of https://github.com/langchain-ai/langgraph.git synced 2026-07-15 12:35:43 -04:00
Code Issues 314 Packages Projects Releases 17 Wiki Activity

[PR #6057] Fix human-in-the-loop interrupts without checkpointer #4807

New Issue
Closed
opened 2026-02-20 17:50:46 -05:00 by yindo · 0 comments
yindo commented 2026-02-20 17:50:46 -05:00
Owner
Copy Link
Copy Source

Original Pull Request: https://github.com/langchain-ai/langgraph/pull/6057

State: closed
Merged: No

Fixes #6053

Problem

Human-in-the-loop interrupts using the interrupt() function were not working when graphs were compiled without a checkpointer and invoked without a thread_id. The interrupt key was missing from invoke() results, causing tool calls to proceed normally instead of triggering the review panel.

Root Cause

In the _suppress_interrupt() method in libs/langgraph/langgraph/pregel/_loop.py, interrupt information was not being included in the final output for invoke mode when no checkpointer was present.

Solution

Modified the _suppress_interrupt() method to:

  1. Properly distinguish interrupt types: Static interrupts (interrupt_before/interrupt_after) vs dynamic interrupts (interrupt() function)
  2. Emit proper stream chunks: For both interrupt types in stream mode
  3. Include interrupt data in final output: Ensure interrupt is present in invoke() results even without checkpointer
  4. Handle different output formats: Support both dict and non-dict outputs

Testing

  • Dynamic interrupts now work without checkpointer
  • No regression in existing functionality with checkpointer
  • Normal execution without interrupts unaffected
  • Stream mode continues to work correctly

Files Changed

  • libs/langgraph/langgraph/pregel/_loop.py: Core fix in _suppress_interrupt() method
  • examples/human_in_the_loop/minimal_reproduce_bug.py: Example demonstrating the original issue
  • examples/human_in_the_loop/fixed_human_loop.py: Example showing the fix

This enables human-in-the-loop functionality to work correctly even without a checkpointer, making it easier for developers to implement review workflows.

**Original Pull Request:** https://github.com/langchain-ai/langgraph/pull/6057 **State:** closed **Merged:** No --- Fixes #6053 ## Problem Human-in-the-loop interrupts using the interrupt() function were not working when graphs were compiled without a checkpointer and invoked without a thread_id. The __interrupt__ key was missing from invoke() results, causing tool calls to proceed normally instead of triggering the review panel. ## Root Cause In the _suppress_interrupt() method in libs/langgraph/langgraph/pregel/_loop.py, interrupt information was not being included in the final output for invoke mode when no checkpointer was present. ## Solution Modified the _suppress_interrupt() method to: 1. **Properly distinguish interrupt types**: Static interrupts (interrupt_before/interrupt_after) vs dynamic interrupts (interrupt() function) 2. **Emit proper stream chunks**: For both interrupt types in stream mode 3. **Include interrupt data in final output**: Ensure __interrupt__ is present in invoke() results even without checkpointer 4. **Handle different output formats**: Support both dict and non-dict outputs ## Testing - Dynamic interrupts now work without checkpointer - No regression in existing functionality with checkpointer - Normal execution without interrupts unaffected - Stream mode continues to work correctly ## Files Changed - libs/langgraph/langgraph/pregel/_loop.py: Core fix in _suppress_interrupt() method - examples/human_in_the_loop/minimal_reproduce_bug.py: Example demonstrating the original issue - examples/human_in_the_loop/fixed_human_loop.py: Example showing the fix This enables human-in-the-loop functionality to work correctly even without a checkpointer, making it easier for developers to implement review workflows.
yindo added the pull-request label 2026-02-20 17:50:46 -05:00
yindo closed this issue 2026-02-20 17:50:46 -05:00
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
agents
awaiting author revision
bug
change
chat models
checkpointer
ci
code quality
code quality
dependencies
documentation
drawing
duplicate
enhancement
good first issue
help wanted
invalid
maintainer
meta
missing information
not maintained by langchain
open-swe-auto
open-swe-max
open-swe-max-auto
pending
performance
pull-request
Mirrored from GitHub Pull Request
 question
requires design
sdk
server
stale
studio
tests
type checking
usage question
v0.5
v1.1
No Label pull-request
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
yindo
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: langchain-ai/langgraph#4807
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?