Imported from Nav1212/LeagueAnalytics (
AGENTS.md). Install upstream withnpx skills add Nav1212/LeagueAnalytics. Copyright stays with the author.
Champion.GG repository guidance
Read codex.md for the schema documentation rules. The maintained architecture overview is the wiki.
For changes to ingestion, storage, publication, API queries, frontend data loading, source selection, or deployment, read and apply championgg-architecture. Read only the wiki pages relevant to the task.
Preserve these established boundaries unless the user explicitly changes them:
- Python alone ingests upstream JSON, transforms private Bronze/Silver data, and publishes Gold through the validated DuckDB rebuild.
- TypeScript reads only published Gold snapshots through a read-only connection. Private database access, legacy-table fallbacks, and migrations do not belong in the API. The browser loads JSON from the local API; CDN images remain allowed.
- Publish fresh Gold-only files and atomically switch the manifest. Retain the last valid snapshot on failure and keep each request on one snapshot.
- Keep Riot/demo sources separate through keys, joins, nested results, and rate denominators. Static/search data stays source-independent.
- Only Python receives private storage and Riot credentials. The API's Gold container mount stays read-only.
Use the development guide for change locations and checks. New endpoints need real HTTP integration coverage. Preserve existing private data and unrelated worktree changes. When an authorized change alters a boundary, update the wiki, skill, and relevant checks with the implementation.
