refactor: migrate main.py to modular routers and add project roadmap

- Migrated monolithic main.py to feature-scoped routers in app/routers/
- Added GEMINI.md for project context and AI instructional guidelines
- Updated README.md with a comprehensive modernization plan (SQL migration, robust scraping DSL, frontend modernization)
- Improved authentication with cookie support and modular JS
- Updated test suite and documentation

app/downloaders/anime_sites/AGENTS.md

@@ -0,0 +1,41 @@
+# Anime Sites Downloaders
+
+## OVERVIEW
+Handlers for French anime streaming catalogs that provide metadata and episode listings, delegating actual video extraction to video player handlers.
+
+## WHERE TO LOOK
+
+| File | Purpose |
+|------|---------|
+| `base.py` | Abstract `BaseAnimeSite` class defining the interface all anime sites implement |
+| `animesama.py` | Primary provider with dynamic domain switching, multiple video player extraction |
+| `nekosama.py` | Neko-Sama / Gupy integration (metadata-only, no direct downloads) |
+| `animeultime.py` | Anime-Ultime catalog handler |
+| `vostfree.py` | Vostfree catalog handler |
+| `frenchmanga.py` | French-Manga catalog handler |
+
+## CONVENTIONS
+
+### Interface Contract
+Each site must implement four async methods from `BaseAnimeSite`:
+- `can_handle(url: str) -> bool` — URL pattern matching
+- `search_anime(query, lang) -> list[dict]` — Returns `{title, url, cover_image}`
+- `get_episodes(anime_url, lang) -> list[dict]` — Returns `{episode_number, url, title, host}`
+- `get_anime_metadata(anime_url) -> dict` — Returns `{synopsis, genres, rating, release_year, studio, poster_image, total_episodes, status}`
+- `get_download_link(url) -> tuple[str, str]` — Returns `(video_player_url, filename)`
+
+### Key Patterns
+- **Pipe-separated URLs**: `video_url|anime_page_url|episode_title` — preserves context across extraction
+- **Language parameter**: `lang="vostfr"` or `"vf"` — controls which episodes to return
+- **Video player delegation**: Anime sites return player URLs (vidmoly, sendvid, sibnet, lpayer), not direct downloads
+- **Filename generation**: `{anime_name} - S{season} - {episode}.mp4` format
+- **HTTP headers**: Browser UA and referer required to avoid blocking
+
+### Domain Detection
+- `AnimeSamaDownloader` fetches current domain from `anime-sama.pw` dynamically
+- Uses fallback chain for video extraction: detected player → cached player → priority list
+
+### Error Handling
+- Raise `Exception` with descriptive message on failure
+- Log at appropriate level (`debug` for expected failures, `error` for unexpected)
+- Validate extracted URLs with `_test_video_url()` before returning