phoenix-oss/llama-stack
1
0
Fork 0
forked from phoenix-oss/llama-stack-mirror
Code Pull requests Releases Packages 2 Activity Actions

chore: move embedding deps to RAG tool where they are needed (#1210)

Browse source 
`EMBEDDING_DEPS` were wrongly associated with `vector_io` providers.
They are needed by
https://github.com/meta-llama/llama-stack/blob/main/llama_stack/providers/utils/memory/vector_store.py#L142
and related code and is used by the RAG tool and as such should only be
needed by the `inline::rag-runtime` provider.
This commit is contained in:
Ashwin Bharambe 2025-02-21 11:33:41 -08:00 • committed by GitHub
parent 11697f85c5
commit 992f865b2e
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
34 changed files with 85 additions and 132 deletions
Download patch file Download diff file Expand all files Collapse all files

47
llama_stack/providers/registry/vector_io.py
View file

@ -14,33 +14,13 @@ from llama_stack.providers.datatypes import (
remote_provider_spec,
)
EMBEDDING_DEPS = [
"blobfile",
"chardet",
"pypdf",
"tqdm",
"numpy",
"scikit-learn",
"scipy",
"nltk",
"sentencepiece",
"transformers",
# this happens to work because special dependencies are always installed last
# so if there was a regular torch installed first, this would be ignored
# we need a better way to do this to identify potential conflicts, etc.
# for now, this lets us significantly reduce the size of the container which
# does not have any "local" inference code (and hence does not need GPU-enabled torch)
"torch torchvision --index-url https://download.pytorch.org/whl/cpu",
"sentence-transformers --no-deps",
]
def available_providers() -> List[ProviderSpec]:
return [
InlineProviderSpec(
api=Api.vector_io,
provider_type="inline::meta-reference",
pip_packages=EMBEDDING_DEPS + ["faiss-cpu"],
pip_packages=["faiss-cpu"],
module="llama_stack.providers.inline.vector_io.faiss",
config_class="llama_stack.providers.inline.vector_io.faiss.FaissVectorIOConfig",
deprecation_warning="Please use the `inline::faiss` provider instead.",
@ -49,24 +29,33 @@ def available_providers() -> List[ProviderSpec]:
InlineProviderSpec(
api=Api.vector_io,
provider_type="inline::faiss",
pip_packages=EMBEDDING_DEPS + ["faiss-cpu"],
pip_packages=["faiss-cpu"],
module="llama_stack.providers.inline.vector_io.faiss",
config_class="llama_stack.providers.inline.vector_io.faiss.FaissVectorIOConfig",
api_dependencies=[Api.inference],
),
InlineProviderSpec(
api=Api.vector_io,
provider_type="inline::sqlite_vec",
pip_packages=EMBEDDING_DEPS + ["sqlite-vec"],
provider_type="inline::sqlite-vec",
pip_packages=["sqlite-vec"],
module="llama_stack.providers.inline.vector_io.sqlite_vec",
config_class="llama_stack.providers.inline.vector_io.sqlite_vec.SQLiteVectorIOConfig",
api_dependencies=[Api.inference],
),
InlineProviderSpec(
api=Api.vector_io,
provider_type="inline::sqlite_vec",
pip_packages=["sqlite-vec"],
module="llama_stack.providers.inline.vector_io.sqlite_vec",
config_class="llama_stack.providers.inline.vector_io.sqlite_vec.SQLiteVectorIOConfig",
deprecation_warning="Please use the `inline::sqlite-vec` provider (notice the hyphen instead of underscore) instead.",
api_dependencies=[Api.inference],
),
remote_provider_spec(
Api.vector_io,
AdapterSpec(
adapter_type="chromadb",
pip_packages=EMBEDDING_DEPS + ["chromadb-client"],
pip_packages=["chromadb-client"],
module="llama_stack.providers.remote.vector_io.chroma",
config_class="llama_stack.providers.remote.vector_io.chroma.ChromaVectorIOConfig",
),
@ -75,7 +64,7 @@ def available_providers() -> List[ProviderSpec]:
InlineProviderSpec(
api=Api.vector_io,
provider_type="inline::chromadb",
pip_packages=EMBEDDING_DEPS + ["chromadb"],
pip_packages=["chromadb"],
module="llama_stack.providers.inline.vector_io.chroma",
config_class="llama_stack.providers.inline.vector_io.chroma.ChromaVectorIOConfig",
api_dependencies=[Api.inference],
@ -84,7 +73,7 @@ def available_providers() -> List[ProviderSpec]:
Api.vector_io,
AdapterSpec(
adapter_type="pgvector",
pip_packages=EMBEDDING_DEPS + ["psycopg2-binary"],
pip_packages=["psycopg2-binary"],
module="llama_stack.providers.remote.vector_io.pgvector",
config_class="llama_stack.providers.remote.vector_io.pgvector.PGVectorVectorIOConfig",
),
@ -94,7 +83,7 @@ def available_providers() -> List[ProviderSpec]:
Api.vector_io,
AdapterSpec(
adapter_type="weaviate",
pip_packages=EMBEDDING_DEPS + ["weaviate-client"],
pip_packages=["weaviate-client"],
module="llama_stack.providers.remote.vector_io.weaviate",
config_class="llama_stack.providers.remote.vector_io.weaviate.WeaviateVectorIOConfig",
provider_data_validator="llama_stack.providers.remote.vector_io.weaviate.WeaviateRequestProviderData",
@ -115,7 +104,7 @@ def available_providers() -> List[ProviderSpec]:
Api.vector_io,
AdapterSpec(
adapter_type="qdrant",
pip_packages=EMBEDDING_DEPS + ["qdrant-client"],
pip_packages=["qdrant-client"],
module="llama_stack.providers.remote.vector_io.qdrant",
config_class="llama_stack.providers.remote.vector_io.qdrant.QdrantVectorIOConfig",
),