Vector store inference api (#598)

# What does this PR do?
Moves all the memory providers to use the inference API and improved the
memory tests to setup the inference stack correctly and use the
embedding models


## Test Plan
torchrun $CONDA_PREFIX/bin/pytest -v -s -k "meta_reference"
--inference-model="Llama3.2-3B-Instruct"
--embedding-model="sentence-transformers/all-MiniLM-L6-v2"
llama_stack/providers/tests/inference/test_embeddings.py --env
EMBEDDING_DIMENSION=384


pytest -v -s llama_stack/providers/tests/memory/test_memory.py
--providers="inference=together,memory=weaviate"
--embedding-model="togethercomputer/m2-bert-80M-2k-retrieval" --env
EMBEDDING_DIMENSION=768 --env TOGETHER_API_KEY=<API-KEY> --env
WEAVIATE_API_KEY=foo --env WEAVIATE_CLUSTER_URL=bar
 
pytest -v -s llama_stack/providers/tests/memory/test_memory.py
--providers="inference=together,memory=chroma"
--embedding-model="togethercomputer/m2-bert-80M-2k-retrieval" --env
EMBEDDING_DIMENSION=768 --env TOGETHER_API_KEY=<API-KEY>--env
CHROMA_HOST=localhost --env CHROMA_PORT=8000

pytest -v -s llama_stack/providers/tests/memory/test_memory.py
--providers="inference=together,memory=pgvector"
--embedding-model="togethercomputer/m2-bert-80M-2k-retrieval" --env
PGVECTOR_DB=postgres --env PGVECTOR_USER=postgres --env
PGVECTOR_PASSWORD=mysecretpassword --env PGVECTOR_HOST=0.0.0.0 --env
EMBEDDING_DIMENSION=768 --env TOGETHER_API_KEY=<API-KEY>

pytest -v -s llama_stack/providers/tests/memory/test_memory.py
--providers="inference=together,memory=faiss"
--embedding-model="togethercomputer/m2-bert-80M-2k-retrieval" --env
EMBEDDING_DIMENSION=768 --env TOGETHER_API_KEY=<API-KEY>

llama_stack/providers/remote/memory/weaviate/weaviate.py

@@ -12,10 +12,11 @@ import weaviate
 import weaviate.classes as wvc
 from numpy.typing import NDArray
 from weaviate.classes.init import Auth
+from weaviate.classes.query import Filter
 
 from llama_stack.apis.memory import *  # noqa: F403
 from llama_stack.distribution.request_headers import NeedsRequestProviderData
-from llama_stack.providers.datatypes import MemoryBanksProtocolPrivate
+from llama_stack.providers.datatypes import Api, MemoryBanksProtocolPrivate
 from llama_stack.providers.utils.memory.vector_store import (
     BankWithIndex,
     EmbeddingIndex,
@@ -80,12 +81,21 @@ class WeaviateIndex(EmbeddingIndex):
 
         return QueryDocumentsResponse(chunks=chunks, scores=scores)
 
+    async def delete(self, chunk_ids: List[str]) -> None:
+        collection = self.client.collections.get(self.collection_name)
+        collection.data.delete_many(
+            where=Filter.by_property("id").contains_any(chunk_ids)
+        )
+
 
 class WeaviateMemoryAdapter(
-    Memory, NeedsRequestProviderData, MemoryBanksProtocolPrivate
+    Memory,
+    NeedsRequestProviderData,
+    MemoryBanksProtocolPrivate,
 ):
-    def __init__(self, config: WeaviateConfig) -> None:
+    def __init__(self, config: WeaviateConfig, inference_api: Api.inference) -> None:
         self.config = config
+        self.inference_api = inference_api
         self.client_cache = {}
         self.cache = {}
 
@@ -117,7 +127,7 @@ class WeaviateMemoryAdapter(
         memory_bank: MemoryBank,
     ) -> None:
         assert (
-            memory_bank.memory_bank_type == MemoryBankType.vector
+            memory_bank.memory_bank_type == MemoryBankType.vector.value
         ), f"Only vector banks are supported {memory_bank.memory_bank_type}"
 
         client = self._get_client()
@@ -135,11 +145,11 @@ class WeaviateMemoryAdapter(
                 ],
             )
 
-        index = BankWithIndex(
-            bank=memory_bank,
-            index=WeaviateIndex(client=client, collection_name=memory_bank.identifier),
+        self.cache[memory_bank.identifier] = BankWithIndex(
+            memory_bank,
+            WeaviateIndex(client=client, collection_name=memory_bank.identifier),
+            self.inference_api,
         )
-        self.cache[memory_bank.identifier] = index
 
     async def list_memory_banks(self) -> List[MemoryBank]:
         # TODO: right now the Llama Stack is the source of truth for these banks. That is
@@ -163,6 +173,7 @@ class WeaviateMemoryAdapter(
         index = BankWithIndex(
             bank=bank,
             index=WeaviateIndex(client=client, collection_name=bank_id),
+            inference_api=self.inference_api,
         )
         self.cache[bank_id] = index
         return index