phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-07-21 03:59:42 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

feat: Implement hybrid search in SQLite-vec

Browse source 
Signed-off-by: Varsha Prasad Narsing <varshaprasad96@gmail.com>
This commit is contained in:
Varsha Prasad Narsing 2025-05-29 14:32:54 -07:00
parent 941f505eb0
commit eab85a7121
13 changed files with 335 additions and 10 deletions
Download patch file Download diff file Expand all files Collapse all files

15
llama_stack/providers/utils/memory/vector_store.py
View file

@ -202,6 +202,12 @@ class EmbeddingIndex(ABC):
async def query_keyword(self, query_string: str, k: int, score_threshold: float) -> QueryChunksResponse:
raise NotImplementedError()
@abstractmethod
async def query_hybrid(
self, embedding: NDArray, query_string: str, k: int, score_threshold: float
) -> QueryChunksResponse:
raise NotImplementedError()
@abstractmethod
async def delete(self):
raise NotImplementedError()
@ -246,9 +252,14 @@ class VectorDBWithIndex:
mode = params.get("mode")
score_threshold = params.get("score_threshold", 0.0)
query_string = interleaved_content_as_str(query)
# Calculate embeddings for both vector and hybrid modes
embeddings_response = await self.inference_api.embeddings(self.vector_db.embedding_model, [query_string])
query_vector = np.array(embeddings_response.embeddings[0], dtype=np.float32)
if mode == "keyword":
return await self.index.query_keyword(query_string, k, score_threshold)
elif mode == "hybrid":
return await self.index.query_hybrid(query_vector, query_string, k, score_threshold)
else:
embeddings_response = await self.inference_api.embeddings(self.vector_db.embedding_model, [query_string])
query_vector = np.array(embeddings_response.embeddings[0], dtype=np.float32)
return await self.index.query_vector(query_vector, k, score_threshold)