Keep only published artifacts in db/

.gitattributes

@@ -1,2 +1,3 @@
 # Auto detect text files and perform LF normalization
 * text=auto
+db/static/content/** -text

.github/workflows/pr-checks.yml

@@ -10,30 +10,6 @@ concurrency:
   cancel-in-progress: true
 
 jobs:
-  python-tests:
-    name: Python tests (db/)
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@v7
-
-      - name: Set up Python
-        uses: actions/setup-python@v7
-        with:
-          python-version: '3.11'
-          cache: 'pip'
-          cache-dependency-path: db/requirements.txt
-
-      - name: Install runtime + test deps
-        run: pip install -r db/requirements.txt pytest
-
-      - name: Build MiNERVA test fixture
-        working-directory: db
-        run: python tests/fixtures/build_minerva_fixture.py
-
-      - name: Run pytest
-        working-directory: db
-        run: python -m pytest tests -v
-
   flutter-analyze:
     name: Flutter analyze
     runs-on: ubuntu-latest

README.md

@@ -121,19 +121,9 @@ The built APK will be located at `build/app/outputs/flutter-apk/app-release.apk`
 
 Release builds are filtered to **arm64-v8a** to keep the APK size sane after bundling the BitTorrent runtime. Every modern Android phone and retro handheld is arm64. To produce a 4-ABI build, edit `abiFilters` in `android/app/build.gradle.kts`.
 
-### Rebuild the catalog database (optional)
+### The catalog database
 
-The app pulls a pre-built SQLite catalog from this repo. CI rebuilds it weekly. If you want to run the build locally:
-
-```bash
-cd db
-pip install -r requirements.txt
-python workflow.py                  # full rebuild
-python workflow.py --use-cached     # reuse cached HTTP responses
-python workflow.py --skip-minerva   # skip the ~1.7 GB MiNERVA mirror download
-```
-
-Output is `db/romdb.db`. Adding a new source = drop a folder under `db/sources/<id>/` with `source.yml` + `scraper.py` and add the platform routes in `db/platforms.yml`.
+The app pulls a pre-built SQLite catalog from this repo, updated weekly. `db/CATALOG.md` documents the artifact format. The catalog is provided for use in romgi; forks and derivative tools are not supported.
 
 ## Technical Details
 

db/.gitignore

@@ -1,34 +1,10 @@
-# Cached HTTP responses
+# Local pipeline leftovers, not part of the published artifacts
 cache/
-
-# Downloaded metadata files
 data/
-
-# Generated static files (PS3/PSV)
-static/
-
-# Uncompressed database (only commit the .gz)
+secrets/
 *.db
-!romdb.db.gz
-
-# Temp files
 *.db-journal
-*_temp.db
-*_old.db
-
-# Python
 __pycache__/
-*.py[cod]
-*$py.class
-.Python
-*.so
-.env
-venv/
-ENV/
-
-# Config files (use examples instead)
-config.json
-
-# Credentials (never commit)
 *_creds.json
-secrets/
+scrapers/
+static/

db/CATALOG.md

@@ -0,0 +1,45 @@
+# Catalog format
+
+The app downloads these artifacts from this directory on `main`:
+
+| File | Purpose |
+|---|---|
+| `version.json` | Catalog metadata, checked before download |
+| `romdb.db.gz` | Gzipped SQLite database |
+| `static/content/ps3/raps/*.rap` | PS3 license files, referenced by catalog URLs |
+| `static/content/psv/zrifs/*.txt` | PSVita zRIF keys, referenced by catalog URLs |
+
+## version.json
+
+```json
+{
+  "version": "20260904",
+  "generated_at": "2026-09-04T15:00:00Z",
+  "schema_version": 4,
+  "min_app_version": "1.2.0",
+  "size": 55424693,
+  "uncompressed_size": 345706496,
+  "entries": 241126,
+  "links": 400298,
+  "platforms": 50,
+  "sources": 4,
+  "retroachievements": 9576
+}
+```
+
+`schema_version` bumps trigger a wipe-and-redownload in the app. `min_app_version` gates catalogs that need newer app features.
+
+## Database schema (v4)
+
+- `platforms(id, brand, name)`
+- `entries(slug PK, rom_id, search_key, title, platform, boxart_url, ra_game_id, ra_num_achievements)`
+- `entries_fts` — FTS4 over `search_key`, `tokenize=unicode61`
+- `regions(id, name)` / `regions_entries(entry, region)`
+- `sources(id PK, name, homepage, kind, auth_required, priority, manifest_json)`
+- `source_health(source_id PK, status, last_checked, reason, entry_count, link_count)`
+- `user_sources(id, name, kind, config_json, created_at)` — app-owned, empty in published catalogs
+- `torrents(infohash PK, source_id, name, magnet, torrent_blob, total_size, piece_length, file_count, trackers_json, added_at)`
+- `links(entry, name, type, format, url, filename, host, size, size_str, source_url, source_id, requires_auth, torrent_infohash, torrent_file_index, torrent_file_path)`
+- `entry_groups(id PK, kind, title, platform, member_count, metadata_json)` / `entry_group_members(group_id, entry, member_index, member_label)` — multi-disc grouping
+
+Indexes: `idx_entries_platform` on `entries(platform)`, `idx_entry_groups_kind` on `entry_groups(kind)`.