phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-12-13 04:22:35 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

chore: Support embedding params from metadata

Browse source
This commit is contained in:
Swapna Lekkala 2025-10-14 16:19:02 -07:00
parent ef4bc70bbe
commit a695b4fa84
2 changed files with 230 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

20
llama_stack/providers/utils/memory/openai_vector_store_mixin.py
View file

@ -353,9 +353,25 @@ class OpenAIVectorStoreMixin(ABC):
created_at = int(time.time())
extra = params.model_extra or {}
metadata = params.metadata or {}
provider_vector_db_id = extra.get("provider_vector_db_id")
embedding_model = extra.get("embedding_model")
embedding_dimension = extra.get("embedding_dimension")
# Use embedding info from metadata if available, otherwise from extra_body
if metadata.get("embedding_model"):
# If either is in metadata, use metadata as source
embedding_model = metadata.get("embedding_model")
embedding_dimension = int(metadata["embedding_dimensions"]) if metadata.get("embedding_dimensions") else 768
# Check for conflicts with extra_body
if (extra.get("embedding_model") and extra["embedding_model"] != embedding_model) or (
extra.get("embedding_dimension") and extra["embedding_dimension"] != embedding_dimension
):
raise ValueError("Embedding config inconsistent between metadata and extra_body")
else:
embedding_model = extra.get("embedding_model")
embedding_dimension = extra.get("embedding_dimension", 768)
# use provider_id set by router; fallback to provider's own ID when used directly via --stack-config
provider_id = extra.get("provider_id") or getattr(self, "__provider_id__", None)
# Derive the canonical vector_db_id (allow override, else generate)