Stop resending TMDB base show fields on every season batch

fetch_and_cache_seasons() batches season requests to stay under TMDB's
append_to_response item cap, but every batch included the full base-show
fields (recommendations, external_ids, aggregate_credits, alternative_titles,
watch/providers) in append_to_response, not just the first. Those fields
describe the show, not a season, so they can't change between batches of the
same fetch - refetching them per batch was pure waste that scales with
season count (a 9-season show already pays for it twice; 20+ seasons would
pay 3-4x).

Only the first batch now requests the base fields and refreshes the cached
tv_data from that response; later batches request only their season-specific
append_to_response and reuse fetched_tv_data from the first batch.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

Author: daften

src/app/providers/tmdb.py

@@ -900,17 +900,30 @@ def fetch_and_cache_seasons(media_id, season_numbers, tv_data, language=None):
     result_data = {}
 
     for i in range(0, len(season_numbers), max_seasons_per_request):
+        is_first_batch = i == 0
         season_subset = season_numbers[i : i + max_seasons_per_request]
         append_text = ",".join(
             template.format(season=season)
             for season in season_subset
             for template in TV_DETAIL_SEASON_APPEND_RESPONSES
         )
 
-        params = {
-            **base_params(language),
-            "append_to_response": f"{TV_DETAIL_APPEND_RESPONSES},{append_text}",
-        }
+        # The base show fields (recommendations, external_ids,
+        # aggregate_credits, alternative_titles, watch/providers) describe
+        # the show, not a season, so they can't change between batches of
+        # the same fetch. Only the first batch needs to ask TMDB for them;
+        # later batches reuse fetched_tv_data from that first response
+        # instead of paying for it again (#512).
+        if is_first_batch:
+            params = {
+                **base_params(language),
+                "append_to_response": f"{TV_DETAIL_APPEND_RESPONSES},{append_text}",
+            }
+        else:
+            params = {
+                **base_params(language),
+                "append_to_response": append_text,
+            }
 
         try:
             response = services.api_request(
@@ -922,24 +935,26 @@ def fetch_and_cache_seasons(media_id, season_numbers, tv_data, language=None):
         except requests.exceptions.HTTPError as error:
             handle_error(error)
 
-        # Always refresh the root TV payload from the same TMDB response so
-        # season fetches can pick up updated external_ids (including tvdb_id)
-        # even when the show cache already existed before this request.
-        refreshed_tv_data = process_tv(response, media_id=media_id)
-        if not refreshed_tv_data.get("tvdb_id"):
-            discovered_tvdb_id = _discover_tmdb_tvdb_id_from_search(
-                media_id,
-                tv_data=refreshed_tv_data,
-            )
-            if discovered_tvdb_id:
-                refreshed_tv_data, _ = _apply_tvdb_id_override_to_tv_data(
+        if is_first_batch:
+            # Always refresh the root TV payload from the same TMDB response
+            # so season fetches can pick up updated external_ids (including
+            # tvdb_id) even when the show cache already existed before this
+            # request.
+            refreshed_tv_data = process_tv(response, media_id=media_id)
+            if not refreshed_tv_data.get("tvdb_id"):
+                discovered_tvdb_id = _discover_tmdb_tvdb_id_from_search(
                     media_id,
-                    refreshed_tv_data,
+                    tv_data=refreshed_tv_data,
                 )
-        tv_cache_key = _tv_cache_key(media_id, language)
-        if fetched_tv_data is None or fetched_tv_data != refreshed_tv_data:
-            fetched_tv_data = refreshed_tv_data
-            cache.set(tv_cache_key, fetched_tv_data)
+                if discovered_tvdb_id:
+                    refreshed_tv_data, _ = _apply_tvdb_id_override_to_tv_data(
+                        media_id,
+                        refreshed_tv_data,
+                    )
+            tv_cache_key = _tv_cache_key(media_id, language)
+            if fetched_tv_data is None or fetched_tv_data != refreshed_tv_data:
+                fetched_tv_data = refreshed_tv_data
+                cache.set(tv_cache_key, fetched_tv_data)
 
         # Process and cache each season
         for season_number in season_subset:

src/app/tests/providers/test_metadata.py

@@ -1030,6 +1030,25 @@ def _mock_api_request(source, _method, url, params=None):
         self.assertIn("season/8", result)
         self.assertEqual(result["season/8"]["season_number"], 8)
 
+        # The base show fields (recommendations, external_ids,
+        # aggregate_credits, alternative_titles, watch/providers) don't
+        # change between batches for the same show, so they should only be
+        # requested on the first batch and reused for the rest (#512).
+        # Compare exact comma-separated tokens (not substrings), since
+        # "watch/providers" is also a suffix of the per-season
+        # "season/N/watch/providers" append token.
+        base_fields = set(tmdb.TV_DETAIL_APPEND_RESPONSES.split(","))
+        first_batch_fields = set(append_requests[0].split(","))
+        second_batch_fields = set(append_requests[1].split(","))
+        self.assertTrue(
+            base_fields.issubset(first_batch_fields),
+            "First batch should request the base show fields.",
+        )
+        self.assertFalse(
+            base_fields & second_batch_fields,
+            "Subsequent batches should not re-request the base show fields.",
+        )
+
     @patch("app.providers.tvdb.search")
     @patch("app.providers.tmdb.services.api_request")
     @override_settings(TVDB_API_KEY="test-tvdb-key")