Skip to content

Make deep permuter progress clearer with live score cards - #140

Merged
RetroGameDeveloper merged 1 commit into
mainfrom
feat/deep-permuter-progress
Oct 28, 2026
Merged

RetroGameDeveloper merged 1 commit into
mainfrom
feat/deep-permuter-progress

Conversation

@RetroGameDeveloper

Copy link
Copy Markdown

Summary

  • Replace the single deep-permuter status line with a compact scorecard: starting percentage and score, current best percentage and score, and verified improvement count.
  • Keep the starting measurement and improvement count across the transition from local repairs to random rewrites. Equal/worse candidates, cached misses and failed compiles do not count as improvements.
  • Put time/compile limits and Run/Stop together, with tested/skipped/time counters beneath the scores.
  • Collapse technical output by default and keep the shared progress bar visible outside it. Disable log tail following for this panel.
  • Keep percentages below 100% until the exact judge returns a match. Both ROM and imported searches use their existing shared score scale.

Validation

  • Ruff passed.
  • 49 focused deep-search and GUI tests passed, including stale starting scores, phase transitions, strict improvement counting, 99.9% near misses, exact matches, cancellation/error cleanup and the visible progress bar.
  • Rendered the real panel markup/styles/renderer with synthetic data in headless Chrome and inspected the screenshot.

No game code or data included. No change to the match judge or search budgets. The running app was not restarted because a search may be in progress.

@github-actions github-actions Bot added the enhancement New feature or request label Oct 28, 2026
@RetroGameDeveloper
RetroGameDeveloper merged commit c58a3a6 into main Oct 28, 2026
6 checks passed
@RetroGameDeveloper
RetroGameDeveloper deleted the feat/deep-permuter-progress branch October 28, 2026 23:24
RetroGameDeveloper pushed a commit that referenced this pull request Oct 30, 2026
The launcher gets a "Multiple models" choice: tick the models (an agent's own, or an API endpoint's), and one start
begins the chosen number of helpers on each. The selection is remembered, and the button says how many helpers will
start. Started from #116 (its author is away), finished here:

- one start begins at most 24 helpers in total (MAX_PER_START), refused on the server with a sentence before any
  starts: per-model starts made 50 models x 8 = 400 helpers possible from one click or one request;
- an API helper keeps the model it was started on across Start again and an app restart (picked_model): the
  endpoint's default replaced it, and its costs were recorded against the wrong model;
- a remembered selection drops models of endpoints removed since, and the label says what is actually started;
- an endpoint's model list is capped at 100.

Co-authored-by: ensapra <61149758+ensapra@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

enhancement New feature or request

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant