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 #6633] fix(langgraph): generate unique interrupt IDs for parallel interrupts #5173

New Issue
Closed
opened 2026-02-20 17:51:20 -05:00 by yindo · 0 comments
yindo commented 2026-02-20 17:51:20 -05:00
Owner
Copy Link
Copy Source

Original Pull Request: https://github.com/langchain-ai/langgraph/pull/6633

State: closed
Merged: No

Summary

Fixes #6626 - interrupt() calls in parallel tools generate identical IDs, making multi-interrupt resume impossible.

Problem

When multiple tools ran in parallel within the same ToolNode (or a single node called interrupt() multiple times), all interrupts received identical IDs. This happened because the interrupt ID was computed as hash(namespace) without considering the interrupt index.

Example of the bug:

# Three parallel tool calls, each with interrupt()
Interrupts collected: 3
  0: id=1b09852344aeb4aee6776a24d09c3f9e, tool=create_circle
  1: id=1b09852344aeb4aee6776a24d09c3f9e, tool=create_square  # Same ID!
  2: id=1b09852344aeb4aee6776a24d09c3f9e, tool=create_triangle # Same ID!

# Resume dict collapses due to key collision:
resume_data = {intr.id: value for intr in interrupts}  # Only 1 entry!

Solution

Changed the interrupt ID format to include the interrupt index:

Before After
a61a86efc1202990625a73c5591eee22 a61a86efc1202990625a73c5591eee22:0
(same for all interrupts in namespace) (unique per interrupt: :0, :1, :2)

Changes

  1. langgraph/types.py

    • Interrupt.from_ns() now accepts idx parameter
    • ID format: {hash(namespace)}:{idx}
    • interrupt() function passes idx from scratchpad counter

  2. langgraph/pregel/_algo.py

    • _scratchpad() updated to match resume values with new ID format
    • Maintains backward compatibility with old-format IDs
    • Iterates through {namespace_hash}:{idx} entries in resume_map

  3. langgraph/pregel/_utils.py

    • is_xxh3_128_hexdigest() regex updated to match both formats:
      • Legacy: [0-9a-f]{32}
      • New: [0-9a-f]{32}:\d+

Backward Compatibility

  • Old checkpoints with legacy interrupt IDs continue to work
  • The _scratchpad() function checks for both old and new formats
  • Serialization roundtrip tests verify legacy interrupt deserialization

Test plan

Added comprehensive tests:

  • test_from_ns_unique_ids_for_different_indices - Unit test verifying unique IDs
  • test_multiple_interrupts_in_node_have_unique_ids - Sequential interrupts in single node
  • test_parallel_tool_interrupts_have_unique_ids - Integration test with parallel Send() dispatches
  • Existing test_parallel_interrupts and test_parallel_interrupts_double continue to pass
  • Serialization roundtrip tests for legacy interrupt format
# Run all interrupt-related tests
TEST="tests/test_pregel.py::test_parallel_interrupts tests/test_pregel.py::test_parallel_interrupts_double tests/test_interruption.py tests/test_interrupt_migration.py -v" make test
# Result: 31 passed

Related Issues

  • This fix addresses #6626
  • Related to #6624 (ToolNode doesn't collect all interrupts from parallel execution) - that's a separate issue about interrupt collection, not ID uniqueness
**Original Pull Request:** https://github.com/langchain-ai/langgraph/pull/6633 **State:** closed **Merged:** No --- ## Summary Fixes #6626 - `interrupt()` calls in parallel tools generate identical IDs, making multi-interrupt resume impossible. ### Problem When multiple tools ran in parallel within the same `ToolNode` (or a single node called `interrupt()` multiple times), all interrupts received **identical IDs**. This happened because the interrupt ID was computed as `hash(namespace)` without considering the interrupt index. **Example of the bug:** ```python # Three parallel tool calls, each with interrupt() Interrupts collected: 3 0: id=1b09852344aeb4aee6776a24d09c3f9e, tool=create_circle 1: id=1b09852344aeb4aee6776a24d09c3f9e, tool=create_square # Same ID! 2: id=1b09852344aeb4aee6776a24d09c3f9e, tool=create_triangle # Same ID! # Resume dict collapses due to key collision: resume_data = {intr.id: value for intr in interrupts} # Only 1 entry! ``` ### Solution Changed the interrupt ID format to include the interrupt index: | Before | After | |--------|-------| | `a61a86efc1202990625a73c5591eee22` | `a61a86efc1202990625a73c5591eee22:0` | | (same for all interrupts in namespace) | (unique per interrupt: `:0`, `:1`, `:2`) | ### Changes 1. **`langgraph/types.py`** - `Interrupt.from_ns()` now accepts `idx` parameter - ID format: `{hash(namespace)}:{idx}` - `interrupt()` function passes `idx` from scratchpad counter 2. **`langgraph/pregel/_algo.py`** - `_scratchpad()` updated to match resume values with new ID format - Maintains backward compatibility with old-format IDs - Iterates through `{namespace_hash}:{idx}` entries in resume_map 3. **`langgraph/pregel/_utils.py`** - `is_xxh3_128_hexdigest()` regex updated to match both formats: - Legacy: `[0-9a-f]{32}` - New: `[0-9a-f]{32}:\d+` ### Backward Compatibility - Old checkpoints with legacy interrupt IDs continue to work - The `_scratchpad()` function checks for both old and new formats - Serialization roundtrip tests verify legacy interrupt deserialization ## Test plan Added comprehensive tests: - [x] `test_from_ns_unique_ids_for_different_indices` - Unit test verifying unique IDs - [x] `test_multiple_interrupts_in_node_have_unique_ids` - Sequential interrupts in single node - [x] `test_parallel_tool_interrupts_have_unique_ids` - Integration test with parallel `Send()` dispatches - [x] Existing `test_parallel_interrupts` and `test_parallel_interrupts_double` continue to pass - [x] Serialization roundtrip tests for legacy interrupt format ```bash # Run all interrupt-related tests TEST="tests/test_pregel.py::test_parallel_interrupts tests/test_pregel.py::test_parallel_interrupts_double tests/test_interruption.py tests/test_interrupt_migration.py -v" make test # Result: 31 passed ``` ## Related Issues - This fix addresses #6626 - Related to #6624 (ToolNode doesn't collect all interrupts from parallel execution) - that's a separate issue about interrupt collection, not ID uniqueness
yindo added the pull-request label 2026-02-20 17:51:20 -05:00
yindo closed this issue 2026-02-20 17:51:20 -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#5173
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?