Skip to content

v0.3.0: warning-level quality gates + run_comparison, bug fixes - #1

Merged
MasoudMiM merged 1 commit into
mainfrom
develop
Sep 7, 2026
Merged

v0.3.0: warning-level quality gates + run_comparison, bug fixes#1
MasoudMiM merged 1 commit into
mainfrom
develop

Conversation

@MasoudMiM

Copy link
Copy Markdown
Owner

Bug fixes:

  • simulate() options render Modelica booleans (true/false) instead of Python literals
  • summarize_for_llm() suppression count now counts only actually-suppressed records
  • check_model() no longer duplicates diagnostics from the return text
  • diagnostic parser accepts omc single-position locations (no end range)

New features:

  • Warning-level quality gates: warning_gate_complaints() matches warnings (un/over-specified initial conditions, unit inconsistencies, over/under-determined systems); AgentLoop(warning_gate=...) reports stage 'quality' attempts and feeds complaints into the fix prompt; run_ladder threads the gate through the benchmark (opt-in)
  • run_comparison(): multi-run variance + cross-model comparison with per-model/per-task pass rates, attempts mean/stdev, transcripts under transcripts//rep/, aggregated in comparison.json

Tests: 101 unit + 4 integration (was 88 + 4). Version bumped to 0.3.0.

Bug fixes:
- simulate() options render Modelica booleans (true/false) instead of Python literals
- summarize_for_llm() suppression count now counts only actually-suppressed records
- check_model() no longer duplicates diagnostics from the return text
- diagnostic parser accepts omc single-position locations (no end range)

New features:
- Warning-level quality gates: warning_gate_complaints() matches warnings
  (un/over-specified initial conditions, unit inconsistencies,
  over/under-determined systems); AgentLoop(warning_gate=...) reports
  stage 'quality' attempts and feeds complaints into the fix prompt;
  run_ladder threads the gate through the benchmark (opt-in)
- run_comparison(): multi-run variance + cross-model comparison with
  per-model/per-task pass rates, attempts mean/stdev, transcripts under
  transcripts/<model>/rep<k>/, aggregated in comparison.json

Tests: 101 unit + 4 integration (was 88 + 4). Version bumped to 0.3.0.
@MasoudMiM
MasoudMiM merged commit 1be4cde into main Sep 7, 2026
5 checks passed
@MasoudMiM
MasoudMiM deleted the develop branch September 7, 2026 01:14
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant