feat: add evaluated_only filter to GET /api/v2/refractor/cards (#174)

Closes #174

Adds `evaluated_only: bool = Query(default=True)` to `list_card_states()`.
When True (the default), cards with `last_evaluated_at IS NULL` are excluded —
these are placeholder rows created at pack-open time but never run through the
evaluator. At team scale this eliminates ~2739 zero-value rows from the
default response, making the Discord /refractor status command efficient
without any bot-side changes.

Set `evaluated_only=false` to include all rows (admin/pipeline use case).

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

app/routers_v2/refractor.py

@@ -107,6 +107,7 @@ async def list_card_states(
     tier: Optional[int] = Query(default=None, ge=0, le=4),
     season: Optional[int] = Query(default=None),
     progress: Optional[str] = Query(default=None),
+    evaluated_only: bool = Query(default=True),
     limit: int = Query(default=10, ge=1, le=100),
     offset: int = Query(default=0, ge=0),
     token: str = Depends(oauth2_scheme),
@@ -114,15 +115,18 @@ async def list_card_states(
     """List RefractorCardState rows for a team, with optional filters and pagination.
 
     Required:
-      team_id   -- filter to this team's cards; returns empty list if team has no states
+      team_id        -- filter to this team's cards; returns empty list if team has no states
 
     Optional filters:
-      card_type -- one of 'batter', 'sp', 'rp'; filters by RefractorTrack.card_type
-      tier      -- filter by current_tier (0-4)
-      season    -- filter to players who have batting or pitching season stats in that
-                   season (EXISTS subquery against batting/pitching_season_stats)
-      progress  -- 'close' = only cards within 80% of their next tier threshold;
-                   fully evolved cards are always excluded from this filter
+      card_type      -- one of 'batter', 'sp', 'rp'; filters by RefractorTrack.card_type
+      tier           -- filter by current_tier (0-4)
+      season         -- filter to players who have batting or pitching season stats in that
+                        season (EXISTS subquery against batting/pitching_season_stats)
+      progress       -- 'close' = only cards within 80% of their next tier threshold;
+                        fully evolved cards are always excluded from this filter
+      evaluated_only -- default True; when True, excludes cards where last_evaluated_at
+                        is NULL (cards created but never run through the evaluator).
+                        Set to False to include all rows, including zero-value placeholders.
 
     Pagination:
       limit     -- page size (1-100, default 10)
@@ -199,6 +203,9 @@ async def list_card_states(
             & (RefractorCardState.current_value >= next_threshold_expr * 0.8)
         )
 
+    if evaluated_only:
+        query = query.where(RefractorCardState.last_evaluated_at.is_null(False))
+
     total = query.count()
     items = []
     for state in query.offset(offset).limit(limit):

tests/test_refractor_state_api.py

@@ -956,3 +956,83 @@ def test_get_card_state_last_evaluated_at_null(setup_state_api_db, state_api_cli
         f"Expected last_evaluated_at=null for un-evaluated card, "
         f"got {data['last_evaluated_at']!r}"
     )
+
+
+# ---------------------------------------------------------------------------
+# T3-8: GET /refractor/cards?team_id=X&evaluated_only=false includes un-evaluated
+# ---------------------------------------------------------------------------
+
+
+def test_list_cards_evaluated_only_false_includes_unevaluated(
+    setup_state_api_db, state_api_client
+):
+    """GET /refractor/cards?team_id=X&evaluated_only=false returns cards with last_evaluated_at=NULL.
+
+    What: Create two RefractorCardState rows for the same team — one with
+    last_evaluated_at=None (never evaluated) and one with last_evaluated_at set
+    to a timestamp (has been evaluated).  Call the endpoint twice:
+      1. evaluated_only=true (default) — only the evaluated card appears.
+      2. evaluated_only=false          — both cards appear.
+
+    Why: The default evaluated_only=True filter uses
+    `last_evaluated_at IS NOT NULL` to exclude placeholder rows created at
+    pack-open time but never run through the evaluator.  At team scale (2753
+    rows, ~14 evaluated) this filter is critical for bot performance.
+    This test verifies the opt-out path (evaluated_only=false) exposes all rows,
+    which is needed for admin/pipeline use cases.
+    """
+    from datetime import datetime, timezone
+
+    team = _sa_make_team("SA_T38", gmid=30380)
+    track = _sa_make_track("batter")
+
+    # Card 1: never evaluated — last_evaluated_at is NULL
+    player_unevaluated = _sa_make_player("T38 Unevaluated", pos="1B")
+    card_unevaluated = _sa_make_card(player_unevaluated, team)  # noqa: F841
+    RefractorCardState.create(
+        player=player_unevaluated,
+        team=team,
+        track=track,
+        current_tier=0,
+        current_value=0.0,
+        fully_evolved=False,
+        last_evaluated_at=None,
+    )
+
+    # Card 2: has been evaluated — last_evaluated_at is a timestamp
+    player_evaluated = _sa_make_player("T38 Evaluated", pos="RF")
+    card_evaluated = _sa_make_card(player_evaluated, team)  # noqa: F841
+    RefractorCardState.create(
+        player=player_evaluated,
+        team=team,
+        track=track,
+        current_tier=1,
+        current_value=5.0,
+        fully_evolved=False,
+        last_evaluated_at=datetime(2025, 4, 1, tzinfo=timezone.utc),
+    )
+
+    # Default (evaluated_only=true) — only the evaluated card should appear
+    resp_default = state_api_client.get(
+        f"/api/v2/refractor/cards?team_id={team.id}", headers=AUTH_HEADER
+    )
+    assert resp_default.status_code == 200
+    data_default = resp_default.json()
+    assert data_default["count"] == 1, (
+        f"evaluated_only=true should return 1 card, got {data_default['count']}"
+    )
+    assert data_default["items"][0]["player_name"] == "T38 Evaluated"
+
+    # evaluated_only=false — both cards should appear
+    resp_all = state_api_client.get(
+        f"/api/v2/refractor/cards?team_id={team.id}&evaluated_only=false",
+        headers=AUTH_HEADER,
+    )
+    assert resp_all.status_code == 200
+    data_all = resp_all.json()
+    assert data_all["count"] == 2, (
+        f"evaluated_only=false should return 2 cards, got {data_all['count']}"
+    )
+    names = {item["player_name"] for item in data_all["items"]}
+    assert "T38 Unevaluated" in names
+    assert "T38 Evaluated" in names