fix: stop gating Quickstart's OCR model dropdown on the vision heuristic

[admonstrator]

Summary

• Quickstart auto-detect classifies models by name heuristics only (classifyModelName()), and the "Suggested OCR model" dropdown/checkbox in Step 5 was gated entirely on the resulting visionModels list. A dedicated OCR model with an unfamiliar name — e.g. Mistral's mistral-ocr-latest — matches none of the hardcoded vision hints, so it classifies as ['text'] and never appeared in that list. The dropdown showed "No vision-capable models found" and the "enable OCR" checkbox was disabled, even though the model exists and works.
• detectAndClassify() also always returned ocrProvider: 'custom', never the 'mistral' OCR provider that manual setup already fully supports.
• Considered and rejected: parsing the capabilities object some OpenAI-compatible /v1/models responses include (Mistral does this; LM Studio/Ollama/plain OpenAI-compatible servers don't). That's an undocumented, vendor-specific extension, not part of the OpenAI spec, and would only special-case one provider while leaving every other naming scheme on the same guesswork.
• Fix: stop trying to classify which models are OCR-capable at all (name heuristic or vendor metadata can both miss real models) and instead offer the same non-embedding candidate list already used for the AI dropdown (textModels — every model is already exactly one of ['embedding'] / ['text'] / ['text','vision'], so "has text capability" already means "not embedding-only"). A heuristic suggestedOcrModel still pre-selects a sensible default when one exists; otherwise the user picks manually instead of hitting a dead-end dropdown.
• Added resolveOcrProviderDefault() — a plain host-string check (mirrors the existing api.mistral.ai check in setupService.getMistralUrlValidationOptions) — so the OCR provider defaults to 'mistral' only when the detected host is actually Mistral's API, 'custom' otherwise. A wrong default is never a dead end: the "OCR fallback" step is its own always-visible wizard step (not nested under the AI step's Quickstart/Manual toggle), so the user can change the provider dropdown themselves regardless.
• Updated the dropdown label in views/setup.ejs from "Suggested OCR model (vision-capable)" to "OCR model" since it's no longer heuristic-filtered.
• Also fixed the same three pre-existing ESLint issues in setup.js as fix: reset hidden aiProvider when switching to manual AI configuration #238 (missing /* global Swal */, unused catch (_error), stale no-await-in-loop disable comment).

Test plan

• tests/test-quickstart-model-classification.js: added classifyModelName('mistral-ocr-latest') === ['text'] (documents why the dropdown must not gate on the vision heuristic) and resolveOcrProviderDefault() cases (Mistral host, versioned Mistral URL, local/non-Mistral host, blank host).
• tests/test-setup-wizard-quickstart.js: added a second detection fixture (openai-compatible, api.mistral.ai, a model present only in textModels) asserting the OCR dropdown/checkbox stay enabled, manual selection flows through to the manual OCR model field, and the OCR provider defaults to mistral. Verified this assertion actually fails without the fix (reverted the source changes locally, confirmed red, restored).
• node scripts/run-tests.js --all: 42 passed, 7 skipped (server-dependent), 0 failed.
• npx eslint clean on all changed JS files.
• node scripts/regen-openapi.js produces no diff (no new response field, just a different ocrProvider value).
• As in fix: reset hidden aiProvider when switching to manual AI configuration #238/fix: resolve AI response/prompt log paths relative to process.cwd() #239, npx prettier --check still fails on setup.js/quickstartService.js/the two test files — pre-existing formatting drift confirmed present on origin/main before this change, intentionally left untouched per the same discussion with the repo maintainer.

Closes #236

---

Generated by Claude Code