phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-07-18 10:52:28 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

feat: Enable ingestion of custom embeddings

Browse source 
Signed-off-by: Francisco Javier Arceo <farceo@redhat.com>
This commit is contained in:
Francisco Javier Arceo 2025-05-29 20:58:41 -04:00 • committed by Francisco Arceo
parent 2603f10f95
commit 73456878e5
8 changed files with 224 additions and 15 deletions
Download patch file Download diff file Expand all files Collapse all files

2
llama_stack/providers/inline/tool_runtime/rag/memory.py
View file

@ -146,7 +146,7 @@ class MemoryToolRuntimeImpl(ToolGroupsProtocolPrivate, ToolRuntime, RAGToolRunti
]
for i, chunk in enumerate(chunks):
metadata = chunk.metadata
tokens += metadata["token_count"]
tokens += metadata.get("token_count", 0)
tokens += metadata.get("metadata_token_count", 0)
if tokens > query_config.max_tokens_in_context:

13
llama_stack/providers/utils/memory/vector_store.py
View file

@ -199,11 +199,16 @@ class VectorDBWithIndex:
self,
chunks: list[Chunk],
) -> None:
embeddings_response = await self.inference_api.embeddings(
self.vector_db.embedding_model, [x.content for x in chunks]
)
embeddings = np.array(embeddings_response.embeddings)
chunks_to_embed = [c for c in chunks if c.embedding is None]
if chunks_to_embed:
resp = await self.inference_api.embeddings(
self.vector_db.embedding_model,
[c.content for c in chunks_to_embed],
)
for c, embedding in zip(chunks_to_embed, resp.embeddings, strict=False):
c.embedding = embedding
embeddings = np.array([c.embedding for c in chunks], dtype=np.float32)
await self.index.add_chunks(chunks, embeddings)
async def query_chunks(