ai generated solutions to our ai generated problems

frontend/public/data/README.md

@@ -0,0 +1,33 @@
+# Static Public Data
+
+The frontend tries these JSON snapshots before falling back to the live API:
+
+- `/data/leaderboard-teams.json`
+- `/data/leaderboard-players.json`
+- `/data/home-teams.json`
+- `/data/recent-games.json`
+- `/data/teams/{key}.json`
+- `/data/teams/{key}.games.json`
+- `/data/players/{key}.json`
+- `/data/games/{key}.json`
+- `/data/games/{key}.logs.json`
+
+`{key}` is `encodeURIComponent(value).replace(/%/g, '~')`.
+
+Snapshot files should keep the same response shape as the matching `/api/tss/*`
+endpoint. Missing files are fine: the app remembers the miss for the current
+browser session and uses the existing API route instead.
+
+Do not generate every possible entity forever. Keep `/data` bounded:
+
+- Always generate the small global snapshots: leaderboards, home teams, recent
+  games.
+- Generate team/player/game detail snapshots only for hot items, such as current
+  leaderboard entries, recent games, recently viewed pages, or pinned items.
+- Prune detail snapshots by age and access. For example: keep recent games for
+  7-30 days, keep leaderboard teams/players while they remain ranked, and delete
+  cold files that have not been refreshed recently.
+- Write snapshots atomically (`file.tmp` then rename) so readers never see a
+  partial JSON file.
+- Serve compressed responses from the web server/CDN when possible. Keep source
+  JSON minified unless humans need to inspect a fixture.