Uh oh!
There was an error while loading. Please reload this page.
fix: resilience for local LLM structured output truncation - #153
Merged
neoneye merged 1 commit intoMar 7, 2026
Merged
Conversation
- select_scenario.py: add default='' to holistic_profile_of_the_plan in PlanCharacteristics. Local LLMs with constrained output budgets close nested JSON objects before reaching the last required field, causing Pydantic validation failure. Empty default lets the pipeline continue; downstream f-string interpolation renders cleanly. - run_plan_pipeline.py: safe dict access for server_iso_utc key in CreateScheduleTask. Fixes KeyError on CLI runs where the key may be absent from start_time JSON. Confirmed on Mac Mini M4 Pro 64GB, LM Studio + Qwen 3.5-35B-A3B.
Uh oh!
There was an error while loading. Please reload this page.
82deutschmark added a commit
to VoynichLabs/PlanExe2026
that referenced
this pull request
Mar 7, 2026
…tTask combined_summary and go_no_go_recommendation are the last two required fields in ExpertDetails. Local LLMs with constrained output budgets consistently omit trailing fields in nested JSON objects. Adding default='' follows the same pattern as the SelectScenarioTask fix (PR PlanExeOrg#153) and unblocks the pipeline from failing at this gate. Confirmed failure: PreProjectAssessmentTask failed 2026-03-06 20:42 EST on Mac Mini M4 Pro 64GB with Qwen 3.5-35B-A3B at num_output=8192.
82deutschmark added a commit
to VoynichLabs/PlanExe2026
that referenced
this pull request
Mar 7, 2026
feedback_root_cause is the last field in NegativeFeedbackItem. Local LLMs consistently omit trailing required fields in nested JSON list items, causing 3 validation errors for ExpertConsultation in ExpertReviewTask. Confirmed failure: ExpertReviewTask failed 2026-03-06 22:57 EST on Mac Mini M4 Pro 64GB with Qwen 3.5-35B-A3B (num_output=8192). Same truncation pattern as SelectScenarioTask (PlanExeOrg#153) and PreProjectAssessmentTask (PlanExeOrg#155, PlanExeOrg#156).
huangyingting pushed a commit
to repomesh/PlanExe
that referenced
this pull request
Jun 26, 2026
Resolves Dependabot security alerts in worker_plan/pyproject.toml by bumping to the first stable patched release of each package: - aiohttp 3.13.5 -> 3.14.1 (alerts PlanExeOrg#152,PlanExeOrg#153,PlanExeOrg#156-PlanExeOrg#164) - tornado 6.5.4 -> 6.5.7 (alerts PlanExeOrg#113,PlanExeOrg#114,PlanExeOrg#136,PlanExeOrg#155,PlanExeOrg#165,PlanExeOrg#166,PlanExeOrg#171) - python-multipart 0.0.22 -> 0.0.32 (alerts PlanExeOrg#142,PlanExeOrg#149,PlanExeOrg#167-PlanExeOrg#170) - urllib3 2.6.3 -> 2.7.0 (alerts PlanExeOrg#150,PlanExeOrg#151) - marshmallow 3.24.2 -> 3.26.2 (alert PlanExeOrg#81), staying on 3.x to avoid the breaking 4.x major transformers alert PlanExeOrg#137 is excluded: its only fix is the 5.x major line and the vulnerable Trainer class is never imported by PlanExe (handled separately). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
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 freeto join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
Problem
Two bugs found during local model QA on Mac Mini M4 Pro 64GB with Qwen 3.5-35B-A3B via LM Studio.
1.
SelectScenarioTask— missingholistic_profile_of_the_planholistic_profile_of_the_planis the last field inPlanCharacteristics. Local LLMs with constrained output budgets close the nested JSON object before generating it, causing a Pydantic validation error and pipeline failure.Fix:
default=''— pipeline continues with an empty field rather than failing. Downstreamscenarios_markdown.pyuses f-string interpolation, so empty string renders cleanly.2.
CreateScheduleTask—KeyError: 'server_iso_utc'Hard dict access fails on CLI runs where the key may be absent from
start_timeJSON.Fix: Safe
.get()with fallback key.Changes
worker_plan/worker_plan_internal/lever/select_scenario.py— 1 lineworker_plan/worker_plan_internal/plan/run_plan_pipeline.py— 1 line