microsoft/semantic-kernel · error · ServiceResponseException

Search failed: {e}

Error message

Search failed: {e}

What it means

Raised as a ServiceResponseException in MilvusMemoryStore.get_nearest_matches when .load(), .index(), or .search() throws any Exception. The search involves loading the collection, reading the index metric_type, and executing the ANN search; failures at any step are wrapped with the original error chained.

Source

Thrown at python/semantic_kernel/connectors/memory_stores/milvus/milvus_memory_store.py:429

            raise ServiceResourceNotFoundError(f"Collection {collection_name} does not exist, cannot search.")
        # Search requests takes a list of requests.
        if len(embedding.shape) == 1:
            embedding = expand_dims(embedding, axis=0)

        try:
            self.collections[collection_name].load()
            metric = self.collections[collection_name].index(index_name=SEARCH_FIELD_EMBEDDING).params["metric_type"]
            # Try with passed in metric
            results = self.collections[collection_name].search(
                data=embedding,
                anns_field=SEARCH_FIELD_EMBEDDING,
                limit=limit,
                output_fields=OUTPUT_FIELDS_W_EMBEDDING if with_embeddings else OUTPUT_FIELDS_WO_EMBEDDING,
                param={"metric_type": metric},
            )[0]
        except Exception as e:
            logger.debug(f"Search failed: {e}")
            raise ServiceResponseException(f"Search failed: {e}") from e
        return [
            (milvus_dict_to_memoryrecord(result.fields), result.distance)
            for result in results
            if result.distance >= min_relevance_score
        ]

    async def get_nearest_match(
        self,
        collection_name: str,
        embedding: ndarray,
        min_relevance_score: float = 0.0,
        with_embedding: bool = False,
    ) -> tuple[MemoryRecord, float] | None:
        """Find the nearest match for an embedding.

        Args:
            collection_name (str): The collection to search.
            embedding (ndarray): The embedding to search for.

View on GitHub (pinned to c028a0c7dc)

Solutions

  1. Inspect the interpolated {e} to pinpoint load vs index vs search failure.
  2. Confirm the embedding dimension matches the collection's schema.
  3. Ensure a vector index exists on SEARCH_FIELD_EMBEDDING before searching.
  4. Retry on transient server/network errors with backoff; scale query nodes if OOM recurs.

Example fix

// before
matches = await store.get_nearest_matches('docs', embedding, limit=5)  # ServiceResponseException: Search failed: ...
// after
try:
    matches = await store.get_nearest_matches('docs', embedding, limit=5)
except ServiceResponseException as e:
    logging.error('Milvus search failed: %s', e)
    matches = []
Defensive patterns

Strategy: try-catch

Validate before calling

import numpy as np

def embedding_matches_dim(emb: np.ndarray, expected_dim: int) -> bool:
    return emb is not None and emb.size == expected_dim

if not embedding_matches_dim(embedding, expected_dim=1536):
    raise ValueError('Embedding dimension mismatch')

Try / catch

from semantic_kernel.exceptions import ServiceResponseException

try:
    matches = await store.get_nearest_matches('docs', embedding, limit=5)
except ServiceResponseException as e:
    logging.error('Milvus search failed: %s', e)
    matches = []

Prevention

When it happens

Trigger: Search fails due to: embedding dimension mismatch with the index, missing vector index on SEARCH_FIELD_EMBEDDING, memory pressure during load(), invalid search parameters, or network/server errors.

Common situations: Querying an embedding whose dimension differs from the collection schema. Collection has no vector index built yet. Milvus query node out of memory. Connection drop during search. min_relevance_score filtering post-search.

Related errors

AI-assisted analysis of microsoft/semantic-kernel@c028a0c7dc (2026-08-13). Data as JSON: /api/errors/4db07969bb6449fb. Report an issue: GitHub.