(CODUX-81) Add provider thinking catalogs

[numnx]

🧩 Task: Add provider thinking catalogs

Automated task PR opened by Code UX for T02.

📋 Summary

Sprint: Sprint 184 · Fixes Sprint Page, Jora, Cli Thinking · Task: T02

Fix sprint-page task/live navigation across project and sprint scope, expose provider-specific CLI thinking selections, and make Jira in-work imports honor the configured transition setting.

🤖 Model & Provider

Provider	Codex CLI
Model	gpt-5.5

⏱️ Timing

Started	2026-07-07 13:54:21 UTC
Finished	2026-07-07 14:16:45 UTC
Duration	22m 25s

📊 Token Usage

Metric	Value
Input tokens	464,604
Cached input tokens	23,865,472
Output tokens	43,581
Total tokens	24,373,657
Invocations	1
Tool calls	254
Active time	21m 28s

Included usage estimate (subscription/local login): $15.56

✅ QA Review

⏳ QA review runs after this PR is opened. Check the task card in the dashboard for the outcome once it lands.

📝 Full Task Prompt

## Objective
Extend provider settings so each CLI provider exposes its real supported thinking or reasoning selections in settings, route overrides, validation, and command execution.

## Scope
- Edit `src/contracts/app-types.ts`
- Edit `src/repositories/settings-defaults.ts`
- Edit `src/domain/settings/settings-schema.ts`
- Edit `src/domain/settings/settings-sanitizers/ai-provider-sanitizer.ts`
- Edit `src/infrastructure/providers/cli/provider-command-specs.ts`
- Edit `src/services/cli-workflow-utils.ts` if prompt-level thinking hints need provider-specific wording
- Edit `dashboard/src/v2/lib/settings-view-models.ts`
- Edit `dashboard/src/v2/components/settings/panels/SettingsModelsPanel.tsx`
- Update tests in `tests/backend/domain/settings/settings-schema.test.ts`, `tests/backend/domain/settings/settings-sanitizers/ai-provider-sanitizer.test.ts`, `tests/backend/infrastructure/providers/cli/provider-command-specs.test.ts`, `tests/dashboard/lib/settings-view-models.test.ts`, and `tests/dashboard/settings-page.test.tsx`
- Update provider/settings docs in `docs/settings/provider-routing.md` and, if user-visible settings behavior is described there, `docs/dashboard/dashboard-guide.md`

## Implementation Requirements
1. Introduce a provider-keyed thinking option catalog, with labels and persisted values for `gemini`, `codex`, `claude-code`, `qwen-code`, `opencode`, and `antigravity`; keep `jules` unsupported unless the hosted provider contract proves otherwise.
2. Base the catalog on official/current provider docs or installed CLI help, using these starting references where applicable: OpenAI reasoning docs `https://developers.openai.com/api/docs/guides/reasoning`, Gemini CLI/config and Gemini thinking docs `https://geminicli.com/docs/reference/configuration/` and `https://ai.google.dev/gemini-api/docs/thinking`, Claude extended-thinking docs `https://platform.claude.com/docs/en/build-with-claude/extended-thinking`, Qwen Code reasoning config `https://qwenlm.github.io/qwen-code-docs/en/users/configuration/model-providers/`, OpenCode model variants `https://opencode.ai/docs/models/`, and Antigravity model docs `https://antigravity.google/docs/models`.
3. Update `ThinkingMode` and provider settings validation so provider base settings and route overrides accept only modes supported by that provider, while legacy saved values `SMALL`, `MEDIUM`, and `HIGH` are migrated or aliased to provider-appropriate defaults instead of breaking existing databases.
4. Update `createDefaultProviderSettings`, `createProjectProviderDraft`, and settings cloning/view-model helpers so newly added provider instances default to a valid provider-specific thinking value.
5. Update `SettingsModelsPanel` so the base provider and route override selects use the provider-specific option list, display user-friendly labels, hide the control for unsupported providers, and never render an option invalid for the provider instance.
6. Update CLI command construction so providers with first-class headless flags/config receive the selected thinking value through those mechanisms; use provider-specific prompt wording only for providers whose CLI does not expose a reliable headless argument.

## Constraints
- Do not remove support for existing persisted `SMALL`, `MEDIUM`, or `HIGH` values without a compatibility path.
- Do not expose thinking controls for Jules unless the implementation proves the hosted provider accepts them.
- Do not add new dashboard dependencies or broad settings-page redesigns.
- Keep route override inheritance behavior unchanged except for the provider-specific option set.
- Keep strict TypeScript typing; avoid `any` for the new catalog and option helpers.

## Verification
- Run `pnpm exec vitest run tests/backend/domain/settings/settings-schema.test.ts tests/backend/domain/settings/settings-sanitizers/ai-provider-sanitizer.test.ts tests/backend/infrastructure/providers/cli/provider-command-specs.test.ts tests/dashboard/lib/settings-view-models.test.ts tests/dashboard/settings-page.test.tsx`.
- Run `pnpm run lint`.
- Success means each non-Jules CLI provider shows only its supported thinking selections, invalid provider/mode combinations fail validation or sanitize safely, existing legacy settings still load, and command specs apply provider-specific thinking where supported.

🌿 Branch Info

Base: feature/CODUX-184-fixes-sprint-page-jora-cli-thinking
Head: task/feature-codux-184-fixes-sprint-page-jora-cli-thi-t02-codex-mrapn43k

---

🤖 Generated by Code UX