chore: Updating how default embedding model is set in stack

.github/workflows/integration-vector-io-tests.yml

@@ -169,9 +169,7 @@ jobs:
         run: |
           uv run --no-sync \
             pytest -sv --stack-config="files=inline::localfs,inference=inline::sentence-transformers,vector_io=${{ matrix.vector-io-provider }}" \
-            tests/integration/vector_io \
-            --embedding-model inline::sentence-transformers/nomic-ai/nomic-embed-text-v1.5 \
-            --embedding-dimension 768
+            tests/integration/vector_io
 
       - name: Check Storage and Memory Available After Tests
         if: ${{ always() }}

docs/docs/building_applications/rag.mdx

@@ -88,18 +88,19 @@ Llama Stack provides OpenAI-compatible RAG capabilities through:
 To enable automatic vector store creation without specifying embedding models, configure a default embedding model in your run.yaml like so:
 
 ```yaml
-models:
-  - model_id: nomic-ai/nomic-embed-text-v1.5
-    provider_id: inline::sentence-transformers
-    metadata:
-      embedding_dimension: 768
-      default_configured: true
+vector_stores:
+  default_provider_id: faiss
+  default_embedding_model:
+    provider_id: sentence-transformers
+    model_id: nomic-ai/nomic-embed-text-v1.5
 ```
 
 With this configuration:
-- `client.vector_stores.create()` works without requiring embedding model parameters
-- The system automatically uses the default model and its embedding dimension for any newly created vector store
-- Only one model can be marked as `default_configured: true`
+- `client.vector_stores.create()` works without requiring embedding model or provider parameters
+- The system automatically uses the default vector store provider (`faiss`) when multiple providers are available
+- The system automatically uses the default embedding model (`sentence-transformers/nomic-ai/nomic-embed-text-v1.5`) for any newly created vector store
+- The `default_provider_id` specifies which vector storage backend to use
+- The `default_embedding_model` specifies both the inference provider and model for embeddings
 
 ## Vector Store Operations
 
@@ -108,14 +109,15 @@ With this configuration:
 You can create vector stores with automatic or explicit embedding model selection:
 
 ```python
-# Automatic - uses default configured embedding model
+# Automatic - uses default configured embedding model and vector store provider
 vs = client.vector_stores.create()
 
-# Explicit - specify embedding model when you need a specific one
+# Explicit - specify embedding model and/or provider when you need specific ones
 vs = client.vector_stores.create(
     extra_body={
-        "embedding_model": "nomic-ai/nomic-embed-text-v1.5",
-        "embedding_dimension": 768
+        "provider_id": "faiss",  # Optional: specify vector store provider
+        "embedding_model": "sentence-transformers/nomic-ai/nomic-embed-text-v1.5",
+        "embedding_dimension": 768  # Optional: will be auto-detected if not provided
     }
 )
 ```

llama_stack/apis/datatypes.py

@@ -121,6 +121,7 @@ class Api(Enum, metaclass=DynamicApiMeta):
 
     models = "models"
     shields = "shields"
+    vector_dbs = "vector_dbs"  # only used for routing
     datasets = "datasets"
     scoring_functions = "scoring_functions"
     benchmarks = "benchmarks"

llama_stack/apis/vector_dbs/vector_dbs.py

@@ -4,7 +4,7 @@
 # This source code is licensed under the terms described in the LICENSE file in
 # the root directory of this source tree.
 
-from typing import Literal
+from typing import Literal, Protocol, runtime_checkable
 
 from pydantic import BaseModel
 
@@ -59,3 +59,35 @@ class ListVectorDBsResponse(BaseModel):
     """
 
     data: list[VectorDB]
+
+
+@runtime_checkable
+class VectorDBs(Protocol):
+    """Internal protocol for vector_dbs routing - no public API endpoints."""
+
+    async def list_vector_dbs(self) -> ListVectorDBsResponse:
+        """Internal method to list vector databases."""
+        ...
+
+    async def get_vector_db(
+        self,
+        vector_db_id: str,
+    ) -> VectorDB:
+        """Internal method to get a vector database by ID."""
+        ...
+
+    async def register_vector_db(
+        self,
+        vector_db_id: str,
+        embedding_model: str,
+        embedding_dimension: int | None = 384,
+        provider_id: str | None = None,
+        vector_db_name: str | None = None,
+        provider_vector_db_id: str | None = None,
+    ) -> VectorDB:
+        """Internal method to register a vector database."""
+        ...
+
+    async def unregister_vector_db(self, vector_db_id: str) -> None:
+        """Internal method to unregister a vector database."""
+        ...

llama_stack/core/datatypes.py

@@ -354,6 +354,26 @@ class AuthenticationRequiredError(Exception):
     pass
 
 
+class QualifiedModel(BaseModel):
+    """A qualified model identifier, consisting of a provider ID and a model ID."""
+
+    provider_id: str
+    model_id: str
+
+
+class VectorStoresConfig(BaseModel):
+    """Configuration for vector stores in the stack."""
+
+    default_provider_id: str | None = Field(
+        default=None,
+        description="ID of the vector_io provider to use as default when multiple providers are available and none is specified.",
+    )
+    default_embedding_model: QualifiedModel | None = Field(
+        default=None,
+        description="Default embedding model configuration for vector stores.",
+    )
+
+
 class QuotaPeriod(StrEnum):
     DAY = "day"
 
@@ -499,6 +519,11 @@ class StackRunConfig(BaseModel):
         description="Path to directory containing external API implementations. The APIs code and dependencies must be installed on the system.",
     )
 
+    vector_stores: VectorStoresConfig | None = Field(
+        default=None,
+        description="Configuration for vector stores, including default embedding model",
+    )
+
     @field_validator("external_providers_dir")
     @classmethod
     def validate_external_providers_dir(cls, v):

llama_stack/core/distribution.py

@@ -63,6 +63,10 @@ def builtin_automatically_routed_apis() -> list[AutoRoutedApiInfo]:
             routing_table_api=Api.tool_groups,
             router_api=Api.tool_runtime,
         ),
+        AutoRoutedApiInfo(
+            routing_table_api=Api.vector_dbs,
+            router_api=Api.vector_io,
+        ),
     ]
 
 

llama_stack/core/resolver.py

@@ -29,6 +29,7 @@
 from llama_stack.apis.shields import Shields
 from llama_stack.apis.telemetry import Telemetry
 from llama_stack.apis.tools import ToolGroups, ToolRuntime
+from llama_stack.apis.vector_dbs import VectorDBs
 from llama_stack.apis.vector_io import VectorIO
 from llama_stack.apis.version import LLAMA_STACK_API_V1ALPHA
 from llama_stack.core.client import get_client_impl
@@ -81,6 +82,7 @@ def api_protocol_map(external_apis: dict[Api, ExternalApiSpec] | None = None) ->
         Api.inspect: Inspect,
         Api.batches: Batches,
         Api.vector_io: VectorIO,
+        Api.vector_dbs: VectorDBs,
         Api.models: Models,
         Api.safety: Safety,
         Api.shields: Shields,

llama_stack/core/routers/__init__.py

@@ -29,6 +29,7 @@ async def get_routing_table_impl(
     from ..routing_tables.scoring_functions import ScoringFunctionsRoutingTable
     from ..routing_tables.shields import ShieldsRoutingTable
     from ..routing_tables.toolgroups import ToolGroupsRoutingTable
+    from ..routing_tables.vector_dbs import VectorDBsRoutingTable
 
     api_to_tables = {
         "models": ModelsRoutingTable,
@@ -37,6 +38,7 @@ async def get_routing_table_impl(
         "scoring_functions": ScoringFunctionsRoutingTable,
         "benchmarks": BenchmarksRoutingTable,
         "tool_groups": ToolGroupsRoutingTable,
+        "vector_dbs": VectorDBsRoutingTable,
     }
 
     if api.value not in api_to_tables:
@@ -91,6 +93,9 @@ async def get_auto_router_impl(
         await inference_store.initialize()
         api_to_dep_impl["store"] = inference_store
 
+    elif api == Api.vector_io:
+        api_to_dep_impl["vector_stores_config"] = run_config.vector_stores
+
     impl = api_to_routers[api.value](routing_table, **api_to_dep_impl)
     await impl.initialize()
     return impl

llama_stack/core/routers/vector_io.py

@@ -31,6 +31,7 @@
     VectorStoreObject,
     VectorStoreSearchResponsePage,
 )
+from llama_stack.core.datatypes import VectorStoresConfig
 from llama_stack.log import get_logger
 from llama_stack.providers.datatypes import HealthResponse, HealthStatus, RoutingTable
 
@@ -43,9 +44,11 @@ class VectorIORouter(VectorIO):
     def __init__(
         self,
         routing_table: RoutingTable,
+        vector_stores_config: VectorStoresConfig | None = None,
     ) -> None:
         logger.debug("Initializing VectorIORouter")
         self.routing_table = routing_table
+        self.vector_stores_config = vector_stores_config
 
     async def initialize(self) -> None:
         logger.debug("VectorIORouter.initialize")
@@ -122,6 +125,17 @@ async def openai_create_vector_store(
         embedding_dimension = extra.get("embedding_dimension")
         provider_id = extra.get("provider_id")
 
+        # Use default embedding model if not specified
+        if (
+            embedding_model is None
+            and self.vector_stores_config
+            and self.vector_stores_config.default_embedding_model is not None
+        ):
+            # Construct the full model ID with provider prefix
+            embedding_provider_id = self.vector_stores_config.default_embedding_model.provider_id
+            model_id = self.vector_stores_config.default_embedding_model.model_id
+            embedding_model = f"{embedding_provider_id}/{model_id}"
+
         if embedding_model is not None and embedding_dimension is None:
             embedding_dimension = await self._get_embedding_model_dimension(embedding_model)
 
@@ -132,11 +146,24 @@ async def openai_create_vector_store(
                 raise ValueError("No vector_io providers available")
             if num_providers > 1:
                 available_providers = list(self.routing_table.impls_by_provider_id.keys())
-                raise ValueError(
-                    f"Multiple vector_io providers available. Please specify provider_id in extra_body. "
-                    f"Available providers: {available_providers}"
-                )
-            provider_id = list(self.routing_table.impls_by_provider_id.keys())[0]
+                # Use default configured provider
+                if self.vector_stores_config and self.vector_stores_config.default_provider_id:
+                    default_provider = self.vector_stores_config.default_provider_id
+                    if default_provider in available_providers:
+                        provider_id = default_provider
+                        logger.debug(f"Using configured default vector store provider: {provider_id}")
+                    else:
+                        raise ValueError(
+                            f"Configured default vector store provider '{default_provider}' not found. "
+                            f"Available providers: {available_providers}"
+                        )
+                else:
+                    raise ValueError(
+                        f"Multiple vector_io providers available. Please specify provider_id in extra_body. "
+                        f"Available providers: {available_providers}"
+                    )
+            else:
+                provider_id = list(self.routing_table.impls_by_provider_id.keys())[0]
 
         vector_db_id = f"vs_{uuid.uuid4()}"
         registered_vector_db = await self.routing_table.register_vector_db(
@@ -243,8 +270,7 @@ async def openai_delete_vector_store(
         vector_store_id: str,
     ) -> VectorStoreDeleteResponse:
         logger.debug(f"VectorIORouter.openai_delete_vector_store: {vector_store_id}")
-        provider = await self.routing_table.get_provider_impl(vector_store_id)
-        return await provider.openai_delete_vector_store(vector_store_id)
+        return await self.routing_table.openai_delete_vector_store(vector_store_id)
 
     async def openai_search_vector_store(
         self,

llama_stack/core/routing_tables/common.py

@@ -134,12 +134,15 @@ async def get_provider_impl(self, routing_key: str, provider_id: str | None = No
         from .scoring_functions import ScoringFunctionsRoutingTable
         from .shields import ShieldsRoutingTable
         from .toolgroups import ToolGroupsRoutingTable
+        from .vector_dbs import VectorDBsRoutingTable
 
         def apiname_object():
             if isinstance(self, ModelsRoutingTable):
                 return ("Inference", "model")
             elif isinstance(self, ShieldsRoutingTable):
                 return ("Safety", "shield")
+            elif isinstance(self, VectorDBsRoutingTable):
+                return ("VectorIO", "vector_db")
             elif isinstance(self, DatasetsRoutingTable):
                 return ("DatasetIO", "dataset")
             elif isinstance(self, ScoringFunctionsRoutingTable):