phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-10-08 13:00:52 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

fix(vector-io): unify score calculation to use cosine and normalize to [0,1]

Browse source
This commit is contained in:
ChristianZaccaria 2025-09-04 13:03:59 +01:00
parent 9618adba89
commit a0e0c7030b
9 changed files with 166 additions and 42 deletions
Download patch file Download diff file Expand all files Collapse all files

14
llama_stack/providers/remote/vector_io/qdrant/qdrant.py
View file

@ -36,7 +36,7 @@ from llama_stack.providers.utils.memory.vector_store import (
from .config import QdrantVectorIOConfig as RemoteQdrantVectorIOConfig
log = get_logger(name=__name__, category="vector_io::qdrant")
logger = get_logger(name=__name__, category="vector_io::qdrant")
CHUNK_ID_KEY = "_chunk_id"
# KV store prefixes for vector databases
@ -98,10 +98,13 @@ class QdrantIndex(EmbeddingIndex):
points_selector=models.PointIdsList(points=chunk_ids),
)
except Exception as e:
log.error(f"Error deleting chunks from Qdrant collection {self.collection_name}: {e}")
logger.error(f"Error deleting chunks from Qdrant collection {self.collection_name}: {e}")
raise
async def query_vector(self, embedding: NDArray, k: int, score_threshold: float) -> QueryChunksResponse:
logger.info(
f"QDRANT VECTOR SEARCH CALLED: embedding_shape={embedding.shape}, k={k}, threshold={score_threshold}"
)
results = (
await self.client.query_points(
collection_name=self.collection_name,
@ -120,12 +123,15 @@ class QdrantIndex(EmbeddingIndex):
try:
chunk = Chunk(**point.payload["chunk_content"])
except Exception:
log.exception("Failed to parse chunk")
logger.exception("Failed to parse chunk")
continue
chunks.append(chunk)
scores.append(point.score)
# Cosine similarity range [-1,1] -> normalized to [0,1]
scores.append((point.score + 1.0) / 2.0)
logger.info(f"Computed score {point.score} for chunk id {point.id}")
logger.info(f"QDRANT VECTOR SEARCH RESULTS: Found {len(chunks)} chunks with scores {scores}")
return QueryChunksResponse(chunks=chunks, scores=scores)
async def query_keyword(