fix(wizard): preservar texto de escena y órdenes reales de cancelar/reintentar

[cursor]

Resumen ejecutivo

Qué cambia

El Wizard distingue las palabras de un cartel o diálogo de las órdenes de cancelar/reintentar. Un reintento explícito conserva su tarea original aunque el mensaje también describa la imagen que debe generar.

Para qué sirve

El hallazgo de Cursor era real: «Generate a video of a director yelling stop the video» podía cancelar otro trabajo. Su primera solución daba prioridad a cualquier mención de generación y rompía un reintento real de Flux ya cubierto por la suite completa.

Impacto para el usuario

Los textos literales permanecen intactos. Las órdenes explícitas de cancelar/reintentar, incluidas las que aparecen en otra frase, mantienen su identificador de tarea y no se convierten en una generación nueva.

Riesgo

• Bajo, clasificador acotado EN/ES.

Estado

• Listo para revisión
• CI obligatoria del HEAD actualizado aprobada

Summary

Fix task-word hijacking without giving generation unconditional priority across an entire message. Keep current development's #291 capability guards. Inspect task controls per clause; mask quoted content only in a derived classification view of a generation request. Leading task commands and explicit controls in another sentence retain the existing task precedence.

Overview

A generation clause describing a director saying stop the video remains generation. Quoted scene text can contain punctuation without creating task commands. Retry the failed image task. Create an image of a lighthouse. retains its retry_task and exact task ID. An explicit cancellation in another sentence remains a cancellation.

Detailed changes

Backend

No task/queue/retry endpoint changes.

UI and Wizard

• Use the existing generation classifiers to identify scene-description prefixes.
• Inspect cancel/retry occurrences per clause, preserving task control when it precedes generation wording.
• Remove quoted captions only from derived classifier input when the unquoted request is generative; never mutate the request or submitted prompt.
• Cover explicit task identity, quoted captions with punctuation, ordinary generation, and the existing Flux retry regression.

Data, provenance and compatibility

Existing task IDs, prompts and proposals are retained. Current how-to guards remain in place. No schema, workspace, persistence or model changes.

Files and ownership

• ui/src/features/agent/agentActions.ts
• ui/tests/agentActions.test.mjs
• Existing wizardTurnReport.test.ts remains unchanged and must pass.

Validation

• Base: bde8a3b6d01c2ccbdd71a45f83cf8f474713c8dd (development).
• Head: 8677f248b39dcd58592f8a6ab024c9fcb9609510.
• Original patch on the updated base: 2 regression groups FAIL, including the unmodified existing Flux retry test. Logs retained.
• Corrected candidate: 109 focused action/report/Labs tests PASS.
• Full local UI: 1,450 tests PASS. Clean-repo guard, i18n, ESLint, TypeScript, production build, bundle budget and code-health ratchet against exact development base PASS. No baseline or budget changes. Current-head remote CI passed.
• No actions executed, external LLM calls or new model inference during these tests.

CI and review

• Updated-head CI SUCCESS: run 34415164216. 3,022 Python passed / 9 skipped; 1,450 UI passed; 26 Chromium E2E and 5 Windows Speech export E2E passed. CI required and both Cursor automations SUCCESS. Formal Independent QA and Cursor Bugbot NEUTRAL are not approvals.
• Original run 34411499198 failed the existing Flux retry test, not an infrastructure check.
• Technical review performed here; independent Luna review unavailable in this pass.
• No remote merge or release.

Notes and limitations

This remains a bounded EN/ES intent heuristic. It does not certify arbitrary prose, all quotation syntaxes or every same-sentence compound command. Separate explicit task-control clauses retain the existing precedence instead of launching extra work. The change fixes the reported literal-scene cases and preserves the real retry fixture rather than weakening it.

[github-actions]

PR Review — Loreframe Studio

Risk: medium
Scope: 2 file(s); +104/-2; React UI

Automated review from scripts/analyze_pr.py. This is a heuristic pass (no LLM) so humans still own the merge decision.

Findings

• medium — Dangerous dynamic execution (ui/src/features/agent/agentActions.ts)
  ui/src/features/agent/agentActions.ts adds exec(). Confirm the input is trusted and sandboxed.
• low — UI changed — rebuild before merge
  Run cd ui && npm run build (CI already does this). Pinokio Update rebuilds for end users; keep ui/dist untracked.

Changed files

• modified: ui/src/features/agent/agentActions.ts, ui/tests/agentActions.test.mjs

CONTRIBUTING checklist

• python scripts/verify_clean_repo.py
• python -m compileall -q app/services app/launch.py scripts
• cd ui && npm run build if the UI changed
• No weights, CivitAI sidecars, or generated guides
• Stays local-first (no required accounts / telemetry)

Posted by the repo PR review workflow. Re-runs on each push to the PR.

[github-actions]

Code health

Quality score: 58.7/100

Higher is better. The score is a trend dashboard; the independent ratchet below remains the CI gate.

Component	Weight	Current	Change
Cyclomatic health	45%	56.2	+0.0
File concentration	25%	66.5	+0.1
Oversized-file debt	20%	42.4	+0.0
Modularity	10%	83.0	+0.0

Change vs PR base: +0.0 points.

Metric	Value
Production LOC	282,892
Production files	856
Test LOC	111,926
Functions measured	18,649
Functions complexity ≥ 15	865
Maximum complexity	674
Policy	code-health-policy-v1
HEAD	e6f1dd6de41b06ea38785ea3581f11dd96d4d8e1
Base	bde8a3b6d01c2ccbdd71a45f83cf8f474713c8dd
UI measurement	complete

Markdown, JSON catalogs and tests are out of this table. Only app/ runtime + ui/src TS/JS count.

Most complex functions

Complexity	Where
674	app/wgp.py:7221 generate_video
379	ui/src/stores/useStore.ts:4122 Async method 'startGeneration'
355	app/_launch_runtime.py:23653 _run_generation
308	app/wgp.py:12363 generate_video_tab
264	ui/src/stores/useStore.ts:8667 Async method 'loadSettingsFromOutput'
262	ui/src/components/Sidebar/SceneAnimatorPanel.tsx:496 Function 'SceneAnimatorPanel'
258	app/services/director/planners/short_film.py:3433 ShortFilmPlanner._plan_story_driven
248	app/services/director_pipeline.py:13607 _run_video_generation
245	app/services/director_pipeline.py:7712 _run_pipeline
233	ui/src/features/agent/agentActions.ts:1152 Function 'parseAction'
226	app/services/director_pipeline.py:6541 update_comic_preview
225	ui/src/features/agent/agentActions.ts:2973 Async function 'executeAgentActions'

Trend vs baseline

Metric	Δ
Production LOC	+25
Test LOC	+77
Functions ≥ 15	+0
Maximum complexity	+0

Warnings

• production LOC increased by +25
• hotspot ui/src/features/agent/agentActions.ts increased by +25 lines

Ratchet passed.

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: serverGenReqId_5d3dc609-500f-42fa-98da-adf1e493bfd4)