Fix content recommender SQL parameter binding

apps/ml-service/app/recommender/content_based.py

@@ -37,10 +37,10 @@ async def get_content_recommendations(
 
     # Build exclusion clause
     exclude_clause = ""
-    params: list = [preference_vector.tolist(), top_n * 2]
+    params: list = [preference_vector.tolist()]
     if exclude_ids:
         exclude_clause = "AND r.id != ALL(%s)"
-        params.insert(1, exclude_ids)
+        params.append(exclude_ids)
 
     # Query recipes by cosine similarity to preference vector
     query = f"""
@@ -55,9 +55,8 @@ async def get_content_recommendations(
         LIMIT %s
     """
 
-    # Re-add vector for the ORDER BY clause
-    params.append(preference_vector.tolist())
-    params.append(top_n * 2)
+    # Re-add vector for the ORDER BY clause.
+    params.extend([preference_vector.tolist(), top_n * 2])
 
     result = await conn.execute(query, tuple(params))
     rows = await result.fetchall()
@@ -102,10 +101,10 @@ async def _get_nutrition_similarity(
 ) -> list[dict]:
     """Get recipes similar by nutrition vector."""
     exclude_clause = ""
-    params: list = [preference_vector.tolist(), top_n]
+    params: list = [preference_vector.tolist()]
     if exclude_ids:
         exclude_clause = "AND r.id != ALL(%s)"
-        params.insert(1, exclude_ids)
+        params.append(exclude_ids)
 
     query = f"""
         SELECT
@@ -118,8 +117,7 @@ async def _get_nutrition_similarity(
         ORDER BY r.nutrition_vector <=> %s::vector ASC
         LIMIT %s
     """
-    params.append(preference_vector.tolist())
-    params.append(top_n)
+    params.extend([preference_vector.tolist(), top_n])
 
     result = await conn.execute(query, tuple(params))
     return await result.fetchall()

apps/ml-service/tests/test_recommend.py

@@ -8,6 +8,41 @@
 
 from app.main import app
 from app.models.schemas import RecommendResponse, RecipeScore
+from app.recommender.content_based import get_content_recommendations
+
+
+class _FakeResult:
+    def __init__(self, *, one=None, many=None):
+        self._one = one
+        self._many = many or []
+
+    async def fetchone(self):
+        return self._one
+
+    async def fetchall(self):
+        return self._many
+
+
+class _ContentRecommendationConn:
+    def __init__(self, expected_calls):
+        self.expected_calls = expected_calls
+        self.calls = []
+
+    async def execute(self, query, params):
+        self.calls.append((query, params))
+
+        if "FROM user_taste_profiles" in query:
+            return _FakeResult(
+                one={
+                    "preference_vector": [0.1, 0.2, 0.3],
+                    "interaction_count": 8,
+                }
+            )
+
+        expected_params, rows = self.expected_calls.pop(0)
+        assert params == expected_params
+        assert query.count("%s") == len(params)
+        return _FakeResult(many=rows)
 
 
 @asynccontextmanager
@@ -118,3 +153,68 @@ async def test_recommend_validates_top_n(client: AsyncClient):
 async def test_recommend_validates_missing_user_id(client: AsyncClient):
     response = await client.post("/recommend", json={"top_n": 10})
     assert response.status_code == 422
+
+
+@pytest.mark.asyncio
+async def test_content_recommendations_binds_vector_params_in_sql_order():
+    vector = [0.1, 0.2, 0.3]
+    conn = _ContentRecommendationConn(
+        expected_calls=[
+            (
+                (vector, vector, 6),
+                [
+                    {
+                        "recipe_id": "ingredient-match",
+                        "title": "Ingredient Match",
+                        "similarity_score": 0.9,
+                    }
+                ],
+            ),
+            (
+                (vector, vector, 3),
+                [
+                    {
+                        "recipe_id": "nutrition-match",
+                        "title": "Nutrition Match",
+                        "similarity_score": 0.8,
+                    }
+                ],
+            ),
+        ]
+    )
+
+    recommendations = await get_content_recommendations(conn, "user-001", top_n=3)
+
+    assert [rec["recipe_id"] for rec in recommendations] == [
+        "ingredient-match",
+        "nutrition-match",
+    ]
+    assert conn.expected_calls == []
+
+
+@pytest.mark.asyncio
+async def test_content_recommendations_binds_exclusions_before_order_vector():
+    vector = [0.1, 0.2, 0.3]
+    exclude_ids = ["recipe-to-skip"]
+    conn = _ContentRecommendationConn(
+        expected_calls=[
+            (
+                (vector, exclude_ids, vector, 4),
+                [],
+            ),
+            (
+                (vector, exclude_ids, vector, 2),
+                [],
+            ),
+        ]
+    )
+
+    recommendations = await get_content_recommendations(
+        conn,
+        "user-001",
+        top_n=2,
+        exclude_ids=exclude_ids,
+    )
+
+    assert recommendations == []
+    assert conn.expected_calls == []