fix(cli): handle multiple concurrent interrupts in HITL workflow #318
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Fixes RuntimeError when parallel subagents create multiple pending interrupts.
When multiple subagents call tools requiring approval simultaneously, LangGraph
needs responses for all interrupts mapped by their IDs.
Changes:
- Collect all pending interrupts in a dict keyed by interrupt_id
- Handle each interrupt (auto-approve or manual prompt)
- Build proper resume format:
* Single interrupt: pass response directly
* Multiple interrupts: pass {interrupt_id: response} dict
- Maintains backward compatibility with single interrupt handling
This enables parallel subagent execution with both auto-approve and manual
HITL approval workflows.
Fixes issue where interrupts created after resuming were not properly handled.
When the agent resumes from multiple interrupts and creates new interrupts,
the code now queries agent.get_state() to get ALL pending interrupts from
LangGraph's state, not just those from the current stream iteration.
This ensures proper formatting of the resume command:
- Single interrupt: pass response directly
- Multiple interrupts: pass {interrupt_id: response} dict
Previously, pending_interrupts was reset each loop iteration, causing the code
to miss interrupts that were still pending in the graph state.
eyurtsev
force-pushed
the
rlm/fix-interrupts
branch
from
47ddf93 to
11ab2b1
Compare
Collaborator
|
Pushed a rebase on top of master -- taking over the PR @rlancemartin |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary
Fixes
RuntimeErrorwhen parallel subagents create multiple pending interrupts in the human-in-the-loop (HITL) workflow.Problem
I was testing multiple sub-agents, which are used in the web-research skill. When spawning multiple subagents in parallel that require HITL approval (e.g., 3 research subagents all calling
web_search), LangGraph raises:The previous code only handled one interrupt at a time and passed a simple response dict. LangGraph requires responses for all pending interrupts to be mapped by their interrupt IDs when multiple are present.
Solution
Changes to
execution.py:hitl_requestto a dictpending_interruptskeyed by interrupt ID{"decisions": [...]}{interrupt_id_1: {"decisions": [...]}, interrupt_id_2: {...}}Test Plan
Tested all w/
web-research skill.--auto-approve- should auto-approve all