fixes

llama_stack/core/routers/vector_io.py

@@ -135,6 +135,8 @@ class VectorIORouter(VectorIO):
         logger.debug(f"VectorIORouter.openai_create_vector_store: name={params.name}, provider_id={provider_id}")
 
         # If no embedding model is provided, use the first available one
+        # TODO: this branch will soon be deleted so you _must_ provide the embedding_model when 
+        # creating a vector store
         if embedding_model is None:
             embedding_model_info = await self._get_first_embedding_model()
             if embedding_model_info is None:
@@ -153,7 +155,14 @@ class VectorIORouter(VectorIO):
         )
         provider = await self.routing_table.get_provider_impl(registered_vector_db.identifier)
 
-        # Pass params as-is to provider - it will extract what it needs from model_extra
+        # Update model_extra with registered values so provider uses the already-registered vector_db
+        if params.model_extra is None:
+            params.model_extra = {}
+        params.model_extra["provider_vector_db_id"] = registered_vector_db.provider_resource_id
+        params.model_extra["provider_id"] = registered_vector_db.provider_id
+        params.model_extra["embedding_model"] = embedding_model
+        params.model_extra["embedding_dimension"] = embedding_dimension
+
         return await provider.openai_create_vector_store(params)
 
     async def openai_list_vector_stores(

llama_stack/providers/utils/memory/openai_vector_store_mixin.py

@@ -10,8 +10,9 @@ import mimetypes
 import time
 import uuid
 from abc import ABC, abstractmethod
-from typing import Any
+from typing import Annotated, Any
 
+from fastapi import Body
 from pydantic import TypeAdapter
 
 from llama_stack.apis.common.errors import VectorStoreNotFoundError
@@ -342,7 +343,7 @@ class OpenAIVectorStoreMixin(ABC):
 
     async def openai_create_vector_store(
         self,
-        params: OpenAICreateVectorStoreRequestWithExtraBody,
+        params: Annotated[OpenAICreateVectorStoreRequestWithExtraBody, Body(...)],
     ) -> VectorStoreObject:
         """Creates a vector store."""
         created_at = int(time.time())
@@ -978,7 +979,7 @@ class OpenAIVectorStoreMixin(ABC):
     async def openai_create_vector_store_file_batch(
         self,
         vector_store_id: str,
-        params: OpenAICreateVectorStoreFileBatchRequestWithExtraBody,
+        params: Annotated[OpenAICreateVectorStoreFileBatchRequestWithExtraBody, Body(...)],
     ) -> VectorStoreFileBatchObject:
         """Create a vector store file batch."""
         if vector_store_id not in self.openai_vector_stores:

llama_stack/providers/utils/memory/vector_store.py

@@ -21,6 +21,7 @@ from llama_stack.apis.common.content_types import (
     URL,
     InterleavedContent,
 )
+from llama_stack.apis.inference import OpenAIEmbeddingsRequestWithExtraBody
 from llama_stack.apis.tools import RAGDocument
 from llama_stack.apis.vector_dbs import VectorDB
 from llama_stack.apis.vector_io import Chunk, ChunkMetadata, QueryChunksResponse
@@ -274,10 +275,11 @@ class VectorDBWithIndex:
                 _validate_embedding(c.embedding, i, self.vector_db.embedding_dimension)
 
         if chunks_to_embed:
-            resp = await self.inference_api.openai_embeddings(
-                self.vector_db.embedding_model,
-                [c.content for c in chunks_to_embed],
+            params = OpenAIEmbeddingsRequestWithExtraBody(
+                model=self.vector_db.embedding_model,
+                input=[c.content for c in chunks_to_embed],
             )
+            resp = await self.inference_api.openai_embeddings(params)
             for c, data in zip(chunks_to_embed, resp.data, strict=False):
                 c.embedding = data.embedding
 
@@ -316,7 +318,11 @@ class VectorDBWithIndex:
         if mode == "keyword":
             return await self.index.query_keyword(query_string, k, score_threshold)
 
-        embeddings_response = await self.inference_api.openai_embeddings(self.vector_db.embedding_model, [query_string])
+        params = OpenAIEmbeddingsRequestWithExtraBody(
+            model=self.vector_db.embedding_model,
+            input=[query_string],
+        )
+        embeddings_response = await self.inference_api.openai_embeddings(params)
         query_vector = np.array(embeddings_response.data[0].embedding, dtype=np.float32)
         if mode == "hybrid":
             return await self.index.query_hybrid(