docs: add watch-order workflow

README.md

@@ -0,0 +1,35 @@
+# mal
+
+## watch-order dataset
+
+relations are loaded from a local dataset file instead of live scraping chiaki at request time.
+
+### runtime
+
+- env var: `WATCH_ORDER_FILE`
+- default path: `./data/watch_order.json`
+
+### regenerate dataset
+
+1. refresh seed ids from local db
+
+```sh
+python3 - <<'PY'
+import sqlite3, json, pathlib
+conn = sqlite3.connect('mal.db')
+cur = conn.cursor()
+cur.execute('select id from anime order by id')
+ids=[r[0] for r in cur.fetchall()]
+path=pathlib.Path('tmp/watch_order_seed_ids.json')
+path.write_text(json.dumps({'ids': ids}, indent=2), encoding='utf-8')
+print(f'wrote {len(ids)} ids to {path}')
+PY
+```
+
+2. generate dataset json
+
+```sh
+go run ./cmd/watchorder -seed "tmp/watch_order_seed_ids.json" -out "data/watch_order.json"
+```
+
+3. restart the server/container to load the updated file