Skip to content

Alternate fix for conversations that stop responding due to stale execution status #2470

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
DoubleDensity wants to merge 5 commits into
base: main
Choose a base branch
from
Open

Alternate fix for conversations that stop responding due to stale execution status #2470

Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
55d51b4
Removed STUCK from the pause check inside the while loop; The STUCK →…
DoubleDensity Mar 16, 2026
ef5f3d1
Outside the loop, STUCK → RUNNING was correctly reset. But inside the…
DoubleDensity Mar 17, 2026
bfbf857
Track event count to prevent repeated stuck detection
DoubleDensity Mar 17, 2026
502df78
Merge branch 'main' into stuck_conversation_fix_alternate
DoubleDensity Mar 17, 2026
56a0d06
Merge branch 'main' into stuck_conversation_fix_alternate
xingyaoww Mar 18, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
35 changes: 24 additions & 11 deletions openhands-sdk/openhands/sdk/conversation/impl/local_conversation.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -592,17 +592,18 @@ def run(self) -> None:
self._state.execution_status = ConversationExecutionStatus.RUNNING

iteration = 0
last_event_count = len(self._state.events)
try:
while True:
logger.debug(f"Conversation run iteration {iteration}")
with self._state:
# Pause attempts to acquire the state lock
# Before value can be modified step can be taken
# Ensure step conditions are checked when lock is already acquired
if self._state.execution_status in [
ConversationExecutionStatus.PAUSED,
ConversationExecutionStatus.STUCK,
]:
if (
self._state.execution_status
== ConversationExecutionStatus.PAUSED
):
break

# Handle stop hooks on FINISHED
Expand Down Expand Up @@ -634,15 +635,25 @@ def run(self) -> None:
break

# Check for stuck patterns if enabled
# Only check if new events have been added since last check
# to avoid repeatedly triggering on the same pattern
Copy link
Collaborator

@enyst enyst Mar 21, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Do you know by any chance what was the pattern? Or which of the stuck scenarios got triggered?

if self._stuck_detector:
is_stuck = self._stuck_detector.is_stuck()
current_event_count = len(self._state.events)
if current_event_count > last_event_count:
is_stuck = self._stuck_detector.is_stuck()

if is_stuck:
logger.warning("Stuck pattern detected.")
self._state.execution_status = (
ConversationExecutionStatus.STUCK
)
continue
if is_stuck:
logger.warning("Stuck pattern detected.")
self._state.execution_status = (
ConversationExecutionStatus.STUCK
)
# Exit the loop when stuck is detected, rather than
# continuing. Continuing would cause an infinite loop
# because is_stuck() would return true again.
# The caller can then decide how to handle it.
break
# Update event count after checking
last_event_count = current_event_count

# clear the flag before calling agent.step() (user approved)
if (
Expand All @@ -657,6 +668,8 @@ def run(self) -> None:
self, on_event=self._on_event, on_token=self._on_token
)
iteration += 1
# Update event count after agent.step() adds new events
last_event_count = len(self._state.events)

# Check for non-finished terminal conditions
# Note: We intentionally do NOT check for FINISHED status here.
Expand Down
Loading