Repository navigation
feat(reliability): wire FailurePatternDetector into workflow failure handling (#10628) - #10631
Merged
Merged
Conversation
…handling (#10628) The FailurePatternDetector was fully built but had zero production callers on either side — nothing wrote patterns and nothing read them. Wire the full loop: - WorkflowRunner._handle_workflow_execution_failure now calls a guarded helper _record_failure_pattern(plan, error) that: - detect_pattern(causal_chain, error_type) — READ: is this a known recurring failure? (causal_chain = "workflow:{strategy}:{error_type}") - learn_pattern(...) — WRITE: record/increment the pattern in Redis. - annotates the returned failure dict with `known_failure_pattern` {pattern_id, occurrences, resolution_success_rate} when it has recurred, so the learned data is actually consumed by callers/logs. - Fire-and-forget: any detector/Redis error is swallowed so failure handling is never disrupted. Tests: tests/enhanced_orchestration/test_failure_pattern_wiring.py — recurring failure flagged, first-time returns no annotation, detector error never breaks the failure path (3 passing). Part of #10602 / umbrella #10603. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Contributor
✅ SSOT Configuration Compliance: Passing🎉 No hardcoded values detected that have SSOT config equivalents! |
… (review #10631) Address code review: - Record the pattern once at the top level (_depth == 0) instead of on every fallback recursion, so occurrence_count isn't inflated by fallback depth. Extracted the GOAP-replan/fallback logic into _attempt_failure_recovery; the known_failure_pattern annotation now rides whatever failure result is ultimately returned (covers GOAP-replan-failure + fallback-exhausted paths). - Correct the docstring: the detector call is awaited inline (guarded), not fire-and-forget; note the detector's sync Redis client is tracked separately. Tests: +3 (records once at top level + annotates failure; no record/annotate on recursion; success recovery not annotated) — 6 total, passing.
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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.
Thinking Path
The codebase audit (umbrella #10603, #10602) found
FailurePatternDetectorfully built with a Redis-backed singleton (get_pattern_detector()) but zero production callers on either side — nothing wrote patterns and nothing read them, so the learned-failure machinery was dead. Re-verified on the current base before implementing. This wires the full loop (write + read consumer) rather than a one-sided wire that would leave orphan data.What Changed
WorkflowRunner._handle_workflow_execution_failure(enhanced_orchestration/workflow_runner.py) now calls a new guarded helper_record_failure_pattern(plan, error)that:detect_pattern(causal_chain, error_type)to check if this failure mode (causal_chain = "workflow:{strategy}:{error_type}") has recurred.learn_pattern(...)to record/increment the pattern in Redis.known_failure_pattern: {pattern_id, occurrences, resolution_success_rate}when the pattern has recurred, so the learned data is actually surfaced to callers/logs.logger.debug) so workflow failure handling is never disrupted.No code deleted; no
_v2; detector reused as-is.Verification
tests/enhanced_orchestration/test_failure_pattern_wiring.py— 3 passing:learn_patterncalled +known_failure_patternannotation returnedflake8 --max-line-length=120clean;black/isortclean;py_compileclean.Part of #10602 / umbrella #10603. Closes #10628.
Model Used
Claude Opus 4.8 (1M context)