phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-12-23 10:19:47 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

feat(vector-io): configurable embedding models for all providers (v2)\n\nAdds embedding_model and embedding_dimension fields to all VectorIOConfig classes.\nRouter respects provider defaults with fallback.\nIntroduces embedding_utils helper.\nComprehensive docs & samples.\nResolves #2729

Browse source
This commit is contained in:
skamenan7 2025-07-17 11:51:40 -04:00
parent c8f274347d
commit d55dd3e9a0
24 changed files with 482 additions and 14 deletions
Download patch file Download diff file Expand all files Collapse all files

31
llama_stack/distribution/routers/vector_io.py
View file

@ -129,13 +129,32 @@ class VectorIORouter(VectorIO):
) -> VectorStoreObject:
logger.debug(f"VectorIORouter.openai_create_vector_store: name={name}, provider_id={provider_id}")
# If no embedding model is provided, use the first available one
# If no embedding model is provided, try provider defaults then fallback
if embedding_model is None:
embedding_model_info = await self._get_first_embedding_model()
if embedding_model_info is None:
raise ValueError("No embedding model provided and no embedding models available in the system")
embedding_model, embedding_dimension = embedding_model_info
logger.info(f"No embedding model specified, using first available: {embedding_model}")
# Try to get provider-specific embedding model configuration
if provider_id:
try:
provider_impl = self.routing_table.get_provider_impl(provider_id)
provider_config = getattr(provider_impl, "config", None)
if provider_config:
if hasattr(provider_config, "embedding_model") and provider_config.embedding_model:
embedding_model = provider_config.embedding_model
logger.info(f"Using provider config default embedding model: {embedding_model}")
if hasattr(provider_config, "embedding_dimension") and provider_config.embedding_dimension:
embedding_dimension = provider_config.embedding_dimension
logger.info(f"Using provider config embedding dimension: {embedding_dimension}")
except Exception as e:
logger.debug(f"Could not get provider config for {provider_id}: {e}")
# If still no embedding model, use system fallback
if embedding_model is None:
embedding_model_info = await self._get_first_embedding_model()
if embedding_model_info is None:
raise ValueError("No embedding model provided and no embedding models available in the system")
embedding_model, embedding_dimension = embedding_model_info
logger.info(f"No embedding model specified, using first available: {embedding_model}")
vector_db_id = f"vs_{uuid.uuid4()}"
registered_vector_db = await self.routing_table.register_vector_db(

17
llama_stack/providers/inline/vector_io/chroma/config.py
View file

@ -6,12 +6,25 @@
from typing import Any
from pydantic import BaseModel
from pydantic import BaseModel, Field
class ChromaVectorIOConfig(BaseModel):
db_path: str
embedding_model: str | None = Field(
default=None,
description="Optional default embedding model for this provider. If not specified, will use system default.",
)
embedding_dimension: int | None = Field(
default=None,
description="Optional embedding dimension override. Only needed for models with variable dimensions (e.g., Matryoshka embeddings). If not specified, will auto-lookup from model registry.",
)
@classmethod
def sample_run_config(cls, db_path: str = "${env.CHROMADB_PATH}", **kwargs: Any) -> dict[str, Any]:
return {"db_path": db_path}
return {
"db_path": db_path,
# Optional: Configure default embedding model for this provider
# "embedding_model": "all-MiniLM-L6-v2",
# "embedding_dimension": 384, # Only needed for variable-dimension models
}

15
llama_stack/providers/inline/vector_io/faiss/config.py
View file

@ -6,7 +6,7 @@
from typing import Any
from pydantic import BaseModel
from pydantic import BaseModel, Field
from llama_stack.providers.utils.kvstore.config import (
KVStoreConfig,
@ -18,6 +18,14 @@ from llama_stack.schema_utils import json_schema_type
@json_schema_type
class FaissVectorIOConfig(BaseModel):
kvstore: KVStoreConfig
embedding_model: str | None = Field(
default=None,
description="Optional default embedding model for this provider. If not specified, will use system default.",
)
embedding_dimension: int | None = Field(
default=None,
description="Optional embedding dimension override. Only needed for models with variable dimensions (e.g., Matryoshka embeddings). If not specified, will auto-lookup from model registry.",
)
@classmethod
def sample_run_config(cls, __distro_dir__: str, **kwargs: Any) -> dict[str, Any]:
@ -25,5 +33,8 @@ class FaissVectorIOConfig(BaseModel):
"kvstore": SqliteKVStoreConfig.sample_run_config(
__distro_dir__=__distro_dir__,
db_name="faiss_store.db",
)
),
# Optional: Configure default embedding model for this provider
# "embedding_model": "all-MiniLM-L6-v2",
# "embedding_dimension": 384, # Only needed for variable-dimension models
}

11
llama_stack/providers/inline/vector_io/milvus/config.py
View file

@ -20,6 +20,14 @@ class MilvusVectorIOConfig(BaseModel):
db_path: str
kvstore: KVStoreConfig = Field(description="Config for KV store backend (SQLite only for now)")
consistency_level: str = Field(description="The consistency level of the Milvus server", default="Strong")
embedding_model: str | None = Field(
default=None,
description="Optional default embedding model for this provider. If not specified, will use system default.",
)
embedding_dimension: int | None = Field(
default=None,
description="Optional embedding dimension override. Only needed for models with variable dimensions (e.g., Matryoshka embeddings). If not specified, will auto-lookup from model registry.",
)
@classmethod
def sample_run_config(cls, __distro_dir__: str, **kwargs: Any) -> dict[str, Any]:
@ -29,4 +37,7 @@ class MilvusVectorIOConfig(BaseModel):
__distro_dir__=__distro_dir__,
db_name="milvus_registry.db",
),
# Optional: Configure default embedding model for this provider
# "embedding_model": "all-MiniLM-L6-v2",
# "embedding_dimension": 384, # Only needed for variable-dimension models
}

11
llama_stack/providers/inline/vector_io/sqlite_vec/config.py
View file

@ -17,6 +17,14 @@ from llama_stack.providers.utils.kvstore.config import (
class SQLiteVectorIOConfig(BaseModel):
db_path: str = Field(description="Path to the SQLite database file")
kvstore: KVStoreConfig = Field(description="Config for KV store backend (SQLite only for now)")
embedding_model: str | None = Field(
default=None,
description="Optional default embedding model for this provider. If not specified, will use system default.",
)
embedding_dimension: int | None = Field(
default=None,
description="Optional embedding dimension override. Only needed for models with variable dimensions (e.g., Matryoshka embeddings). If not specified, will auto-lookup from model registry.",
)
@classmethod
def sample_run_config(cls, __distro_dir__: str) -> dict[str, Any]:
@ -26,4 +34,7 @@ class SQLiteVectorIOConfig(BaseModel):
__distro_dir__=__distro_dir__,
db_name="sqlite_vec_registry.db",
),
# Optional: Configure default embedding model for this provider
# "embedding_model": "all-MiniLM-L6-v2",
# "embedding_dimension": 384, # Only needed for variable-dimension models
}

17
llama_stack/providers/remote/vector_io/chroma/config.py
View file

@ -6,12 +6,25 @@
from typing import Any
from pydantic import BaseModel
from pydantic import BaseModel, Field
class ChromaVectorIOConfig(BaseModel):
url: str | None
embedding_model: str | None = Field(
default=None,
description="Optional default embedding model for this provider. If not specified, will use system default.",
)
embedding_dimension: int | None = Field(
default=None,
description="Optional embedding dimension override. Only needed for models with variable dimensions (e.g., Matryoshka embeddings). If not specified, will auto-lookup from model registry.",
)
@classmethod
def sample_run_config(cls, url: str = "${env.CHROMADB_URL}", **kwargs: Any) -> dict[str, Any]:
return {"url": url}
return {
"url": url,
# Optional: Configure default embedding model for this provider
# "embedding_model": "all-MiniLM-L6-v2",
# "embedding_dimension": 384, # Only needed for variable-dimension models
}

11
llama_stack/providers/remote/vector_io/milvus/config.py
View file

@ -18,6 +18,14 @@ class MilvusVectorIOConfig(BaseModel):
token: str | None = Field(description="The token of the Milvus server")
consistency_level: str = Field(description="The consistency level of the Milvus server", default="Strong")
kvstore: KVStoreConfig = Field(description="Config for KV store backend")
embedding_model: str | None = Field(
default=None,
description="Optional default embedding model for this provider. If not specified, will use system default.",
)
embedding_dimension: int | None = Field(
default=None,
description="Optional embedding dimension override. Only needed for models with variable dimensions (e.g., Matryoshka embeddings). If not specified, will auto-lookup from model registry.",
)
# This configuration allows additional fields to be passed through to the underlying Milvus client.
# See the [Milvus](https://milvus.io/docs/install-overview.md) documentation for more details about Milvus in general.
@ -32,4 +40,7 @@ class MilvusVectorIOConfig(BaseModel):
__distro_dir__=__distro_dir__,
db_name="milvus_remote_registry.db",
),
# Optional: Configure default embedding model for this provider
# "embedding_model": "all-MiniLM-L6-v2",
# "embedding_dimension": 384, # Only needed for variable-dimension models
}

13
llama_stack/providers/remote/vector_io/pgvector/config.py
View file

@ -23,13 +23,21 @@ class PGVectorVectorIOConfig(BaseModel):
user: str | None = Field(default="postgres")
password: str | None = Field(default="mysecretpassword")
kvstore: KVStoreConfig | None = Field(description="Config for KV store backend (SQLite only for now)", default=None)
embedding_model: str | None = Field(
default=None,
description="Optional default embedding model for this provider. If not specified, will use system default.",
)
embedding_dimension: int | None = Field(
default=None,
description="Optional embedding dimension override. Only needed for models with variable dimensions (e.g., Matryoshka embeddings). If not specified, will auto-lookup from model registry.",
)
@classmethod
def sample_run_config(
cls,
__distro_dir__: str,
host: str = "${env.PGVECTOR_HOST:=localhost}",
port: int = "${env.PGVECTOR_PORT:=5432}",
port: int | str = "${env.PGVECTOR_PORT:=5432}",
db: str = "${env.PGVECTOR_DB}",
user: str = "${env.PGVECTOR_USER}",
password: str = "${env.PGVECTOR_PASSWORD}",
@ -45,4 +53,7 @@ class PGVectorVectorIOConfig(BaseModel):
__distro_dir__=__distro_dir__,
db_name="pgvector_registry.db",
),
# Optional: Configure default embedding model for this provider
# "embedding_model": "all-MiniLM-L6-v2",
# "embedding_dimension": 384, # Only needed for variable-dimension models
}

13
llama_stack/providers/remote/vector_io/qdrant/config.py
View file

@ -6,7 +6,7 @@
from typing import Any
from pydantic import BaseModel
from pydantic import BaseModel, Field
from llama_stack.schema_utils import json_schema_type
@ -23,9 +23,20 @@ class QdrantVectorIOConfig(BaseModel):
prefix: str | None = None
timeout: int | None = None
host: str | None = None
embedding_model: str | None = Field(
default=None,
description="Optional default embedding model for this provider. If not specified, will use system default.",
)
embedding_dimension: int | None = Field(
default=None,
description="Optional embedding dimension override. Only needed for models with variable dimensions (e.g., Matryoshka embeddings). If not specified, will auto-lookup from model registry.",
)
@classmethod
def sample_run_config(cls, **kwargs: Any) -> dict[str, Any]:
return {
"api_key": "${env.QDRANT_API_KEY}",
# Optional: Configure default embedding model for this provider
# "embedding_model": "all-MiniLM-L6-v2",
# "embedding_dimension": 384, # Only needed for variable-dimension models
}

12
llama_stack/providers/remote/vector_io/weaviate/config.py
View file

@ -21,6 +21,15 @@ class WeaviateRequestProviderData(BaseModel):
class WeaviateVectorIOConfig(BaseModel):
embedding_model: str | None = Field(
default=None,
description="Optional default embedding model for this provider. If not specified, will use system default.",
)
embedding_dimension: int | None = Field(
default=None,
description="Optional embedding dimension override. Only needed for models with variable dimensions (e.g., Matryoshka embeddings). If not specified, will auto-lookup from model registry.",
)
@classmethod
def sample_run_config(cls, __distro_dir__: str, **kwargs: Any) -> dict[str, Any]:
return {
@ -28,4 +37,7 @@ class WeaviateVectorIOConfig(BaseModel):
__distro_dir__=__distro_dir__,
db_name="weaviate_registry.db",
),
# Optional: Configure default embedding model for this provider
# "embedding_model": "all-MiniLM-L6-v2",
# "embedding_dimension": 384, # Only needed for variable-dimension models
}

5
llama_stack/providers/utils/vector_io/embedding_utils.py Normal file
View file

@ -0,0 +1,5 @@
# Copyright (c) Meta Platforms, Inc. and affiliates.
# All rights reserved.
#
# This source code is licensed under the terms described in the LICENSE file in
# the root directory of this source tree.