api: dedup ghost history rows (int vs string ep_id); normalize ep_id on write

server/app.py

@@ -125,9 +125,17 @@ HIST_DIR = Path(os.environ.get("ANIWEB_DATA", Path(__file__).parent / "data")) /
 
 def _load_hist(user: str):
     try:
-        return json.loads((HIST_DIR / f"{user}.json").read_text())
+        entries = json.loads((HIST_DIR / f"{user}.json").read_text())
     except (FileNotFoundError, json.JSONDecodeError):
         return []
+    # ponytail: dedup legacy ghost rows — the same episode was stored with ep_id
+    # as int AND string; keep the newest per (slug_id, str(ep_id))
+    best = {}
+    for e in entries:
+        k = (str(e.get("slug_id")), str(e.get("ep_id")))
+        if e.get("t", 0) >= best.get(k, {"t": 0})["t"]:
+            best[k] = e
+    return sorted(best.values(), key=lambda e: e.get("t", 0), reverse=True)
 
 
 def _save_hist(user: str, entries):
@@ -162,11 +170,18 @@ def api_history_delete(request: Request, slug_id: str, ep_id: str):
 def api_history_post(entry: dict = Body(...), request: Request = None):
     if "ep_id" not in entry:
         raise HTTPException(422, "ep_id required")
+    # normalize: ghost rows were born when ep_id arrived as str against an
+    # int row (or vice versa) — int() makes the stored type stable
+    try:
+        entry["ep_id"] = int(entry["ep_id"])
+    except (TypeError, ValueError):
+        raise HTTPException(422, "ep_id must be an integer")
     entry["t"] = time.time()
     u = request.state.user
     entries = _load_hist(u)
     for i, e in enumerate(entries):
-        if e.get("slug_id") == entry.get("slug_id") and e.get("ep_id") == entry.get("ep_id"):
+        # loose str() match: merge onto a legacy ghost row of the other type
+        if e.get("slug_id") == entry.get("slug_id") and str(e.get("ep_id")) == str(entry["ep_id"]):
             entries[i] = {**e, **entry}
             break
     else: