src/actix/api/update_api.rs (1)

331-360: update_batch is missing API key extraction.

The update_batch handler does not extract or propagate ApiKeys, unlike upsert_points and update_vectors in this file, and unlike the gRPC update_batch handler in src/tonic/api/points_api.rs (lines 290-292) which does include API key extraction. Since batch operations can include upsert operations that may require external API keys for inference, this appears to be an oversight.

Suggested fix

 #[post("/collections/{name}/points/batch")]
+#[allow(clippy::too_many_arguments)]
 async fn update_batch(
     dispatcher: web::Data<Dispatcher>,
     collection: Path<CollectionPath>,
     operations: Json<UpdateOperations>,
     params: Query<UpdateParams>,
     service_config: web::Data<ServiceConfig>,
     ActixAccess(access): ActixAccess,
+    api_keys: ApiKeys,
     inference_token: InferenceToken,
 ) -> impl Responder {
     let operations = operations.into_inner();

     let request_hw_counter = get_request_hardware_counter(
         &dispatcher,
         collection.name.clone(),
         service_config.hardware_reporting(),
         Some(params.wait),
     );

-    let inference_params = InferenceParams::new(inference_token.clone(), params.timeout);
+    let inference_params =
+        InferenceParams::new(inference_token, params.timeout).with_ext_api_keys(api_keys);
     let timing = Instant::now();

src/tonic/api/points_api.rs (1)

668-670: Add API key extraction to query_groups handler.

The query_groups handler at lines 668-670 is missing the external API key extraction that both query and query_batch handlers include. Add:

let api_keys = extract_api_key(request.metadata());

and chain it to InferenceParams:

let inference_params = InferenceParams::new(inference_token, timeout).with_ext_api_keys(api_keys);

src/actix/api/query_api.rs (1)

109-132: Inconsistent API keys support across query endpoints.

query_points_batch does not extract or forward API keys, while query_points does. If external API keys are needed for inference during query operations, this endpoint should also support them for consistency.

The same issue applies to query_points_groups (lines 196-221).

Suggested fix for query_points_batch

 #[post("/collections/{name}/points/query/batch")]
+#[allow(clippy::too_many_arguments)]
 async fn query_points_batch(
     dispatcher: web::Data<Dispatcher>,
     collection: Path<CollectionPath>,
     request: Json<QueryRequestBatch>,
     params: Query<ReadParams>,
     service_config: web::Data<ServiceConfig>,
     ActixAccess(access): ActixAccess,
+    api_keys: ApiKeys,
     inference_token: InferenceToken,
 ) -> impl Responder {
     ...
-    let inference_params = InferenceParams::new(inference_token, params.timeout());
+    let inference_params =
+        InferenceParams::new(inference_token, params.timeout()).with_ext_api_keys(api_keys);