MohamedAbdallah-14/prompt-to-asset

Deep test-coverage pass. No behavior changes — every source module surfaces more of its branches to the test harness so regressions surface in CI instead of user bug reports.

Highlights

• Provider registry test suite (62 tests) — covers all 17 image-generation providers through a `fetch` mock: `supportsModel`, `isAvailable`, `generate` happy path, `dryRun`, `ProviderError` on missing keys, HTTP error propagation, asynchronous polling (BFL, Leonardo, Replicate), aspect-ratio bucketing (Google Imagen 4, Ideogram), and native-format detection.
• Tool-coverage test suite (36 tests) — drives the `asset_generate_*` family through `inline_svg`, `external_prompt_only`, and `api` soft-fallback paths, plus `validateAsset`, `ingestExternal`, `trainBrandLora`, and the `removeBackground` / `vectorizeImage` / `upscaleRefine` thin wrappers.
• Pure-logic test suite (24 tests) — locks behavior for `rewriter` dialects (tag-salad, prose+quoted, prose+flags, prose), `svg-briefs` per asset type, `modelsList` / `modelsInspect` filters, `capabilities`, and `doctor`.
• Core-logic test suite (25 tests) — brand-source parsing (native JSON, DTCG tokens, AdCP, markdown, raw text), cache key stability, data-integrity assertions, and router branches.
• Pipeline extras (22 tests) — `matte`, `upscale`, `vectorize`, `tier1Alignment`, and `tier2Vlm`.
• CLI tests (46 tests) — `p2a` through `main`, `models`, `doctor`, `sprite-sheet`, `nine-slice`, `export`, `init --yes`, `pick --yes`, and `mcp` (transport validation).
• Generate-api tests (10 tests) — full `api`-mode pipeline for `logo`, `favicon`, `app-icon`, `hero`, `illustration`, `splash-screen`, and `ingestExternal`.
• `init-brand` tests (18 tests) — every project-detection branch (Next.js, Astro, Vite, Remix, Nuxt, Expo, Flutter, Xcode, Android, Electron, React Native, plain Node, unknown) and the overwrite guard.
• `doctor-fix` tests (3 tests) — dry-run, skip list, step well-formedness.

Coverage

| Metric | Before | After |
| ---------- | ------ | ------ |
| Statements | 37.25% | 87.25% |
| Branches | 38.41% | 74.83% |
| Functions | 90.52% | 95.37% |
| Tests | 186 | 477 |

The remaining gaps are Satori WASM paths in `og-render.ts` (can't load in the test sandbox), `sharp`-missing fallbacks in `pipeline/sharp.ts`, and real shell-out paths in `doctor-fix.ts` (won't exercise without installing binaries).

Changelog

See CHANGELOG.md for the full list.