impls -> inline, adapters -> remote (#381)

llama_stack/providers/remote/memory/__init__.py

@@ -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.

llama_stack/providers/remote/memory/chroma/__init__.py

@@ -0,0 +1,15 @@
+# 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.
+
+from llama_stack.distribution.datatypes import RemoteProviderConfig
+
+
+async def get_adapter_impl(config: RemoteProviderConfig, _deps):
+    from .chroma import ChromaMemoryAdapter
+
+    impl = ChromaMemoryAdapter(config.url)
+    await impl.initialize()
+    return impl

llama_stack/providers/remote/memory/chroma/chroma.py

@@ -0,0 +1,159 @@
+# 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.
+
+import json
+from typing import List
+from urllib.parse import urlparse
+
+import chromadb
+from numpy.typing import NDArray
+
+from pydantic import parse_obj_as
+
+from llama_stack.apis.memory import *  # noqa: F403
+
+from llama_stack.providers.datatypes import MemoryBanksProtocolPrivate
+from llama_stack.providers.utils.memory.vector_store import (
+    BankWithIndex,
+    EmbeddingIndex,
+)
+
+
+class ChromaIndex(EmbeddingIndex):
+    def __init__(self, client: chromadb.AsyncHttpClient, collection):
+        self.client = client
+        self.collection = collection
+
+    async def add_chunks(self, chunks: List[Chunk], embeddings: NDArray):
+        assert len(chunks) == len(
+            embeddings
+        ), f"Chunk length {len(chunks)} does not match embedding length {len(embeddings)}"
+
+        await self.collection.add(
+            documents=[chunk.json() for chunk in chunks],
+            embeddings=embeddings,
+            ids=[f"{c.document_id}:chunk-{i}" for i, c in enumerate(chunks)],
+        )
+
+    async def query(
+        self, embedding: NDArray, k: int, score_threshold: float
+    ) -> QueryDocumentsResponse:
+        results = await self.collection.query(
+            query_embeddings=[embedding.tolist()],
+            n_results=k,
+            include=["documents", "distances"],
+        )
+        distances = results["distances"][0]
+        documents = results["documents"][0]
+
+        chunks = []
+        scores = []
+        for dist, doc in zip(distances, documents):
+            try:
+                doc = json.loads(doc)
+                chunk = Chunk(**doc)
+            except Exception:
+                import traceback
+
+                traceback.print_exc()
+                print(f"Failed to parse document: {doc}")
+                continue
+
+            chunks.append(chunk)
+            scores.append(1.0 / float(dist))
+
+        return QueryDocumentsResponse(chunks=chunks, scores=scores)
+
+
+class ChromaMemoryAdapter(Memory, MemoryBanksProtocolPrivate):
+    def __init__(self, url: str) -> None:
+        print(f"Initializing ChromaMemoryAdapter with url: {url}")
+        url = url.rstrip("/")
+        parsed = urlparse(url)
+
+        if parsed.path and parsed.path != "/":
+            raise ValueError("URL should not contain a path")
+
+        self.host = parsed.hostname
+        self.port = parsed.port
+
+        self.client = None
+        self.cache = {}
+
+    async def initialize(self) -> None:
+        try:
+            print(f"Connecting to Chroma server at: {self.host}:{self.port}")
+            self.client = await chromadb.AsyncHttpClient(host=self.host, port=self.port)
+        except Exception as e:
+            import traceback
+
+            traceback.print_exc()
+            raise RuntimeError("Could not connect to Chroma server") from e
+
+    async def shutdown(self) -> None:
+        pass
+
+    async def register_memory_bank(
+        self,
+        memory_bank: MemoryBankDef,
+    ) -> None:
+        assert (
+            memory_bank.type == MemoryBankType.vector.value
+        ), f"Only vector banks are supported {memory_bank.type}"
+
+        collection = await self.client.get_or_create_collection(
+            name=memory_bank.identifier,
+            metadata={"bank": memory_bank.json()},
+        )
+        bank_index = BankWithIndex(
+            bank=memory_bank, index=ChromaIndex(self.client, collection)
+        )
+        self.cache[memory_bank.identifier] = bank_index
+
+    async def list_memory_banks(self) -> List[MemoryBankDef]:
+        collections = await self.client.list_collections()
+        for collection in collections:
+            try:
+                data = json.loads(collection.metadata["bank"])
+                bank = parse_obj_as(MemoryBankDef, data)
+            except Exception:
+                import traceback
+
+                traceback.print_exc()
+                print(f"Failed to parse bank: {collection.metadata}")
+                continue
+
+            index = BankWithIndex(
+                bank=bank,
+                index=ChromaIndex(self.client, collection),
+            )
+            self.cache[bank.identifier] = index
+
+        return [i.bank for i in self.cache.values()]
+
+    async def insert_documents(
+        self,
+        bank_id: str,
+        documents: List[MemoryBankDocument],
+        ttl_seconds: Optional[int] = None,
+    ) -> None:
+        index = self.cache.get(bank_id, None)
+        if not index:
+            raise ValueError(f"Bank {bank_id} not found")
+
+        await index.insert_documents(documents)
+
+    async def query_documents(
+        self,
+        bank_id: str,
+        query: InterleavedTextMedia,
+        params: Optional[Dict[str, Any]] = None,
+    ) -> QueryDocumentsResponse:
+        index = self.cache.get(bank_id, None)
+        if not index:
+            raise ValueError(f"Bank {bank_id} not found")
+
+        return await index.query_documents(query, params)

llama_stack/providers/remote/memory/pgvector/__init__.py

@@ -0,0 +1,15 @@
+# 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.
+
+from .config import PGVectorConfig
+
+
+async def get_adapter_impl(config: PGVectorConfig, _deps):
+    from .pgvector import PGVectorMemoryAdapter
+
+    impl = PGVectorMemoryAdapter(config)
+    await impl.initialize()
+    return impl

llama_stack/providers/remote/memory/pgvector/config.py

@@ -0,0 +1,17 @@
+# 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.
+
+from llama_models.schema_utils import json_schema_type
+from pydantic import BaseModel, Field
+
+
+@json_schema_type
+class PGVectorConfig(BaseModel):
+    host: str = Field(default="localhost")
+    port: int = Field(default=5432)
+    db: str = Field(default="postgres")
+    user: str = Field(default="postgres")
+    password: str = Field(default="mysecretpassword")

llama_stack/providers/remote/memory/pgvector/pgvector.py

@@ -0,0 +1,212 @@
+# 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.
+
+from typing import List, Tuple
+
+import psycopg2
+from numpy.typing import NDArray
+from psycopg2 import sql
+from psycopg2.extras import execute_values, Json
+
+from pydantic import BaseModel, parse_obj_as
+
+from llama_stack.apis.memory import *  # noqa: F403
+
+from llama_stack.providers.datatypes import MemoryBanksProtocolPrivate
+from llama_stack.providers.utils.memory.vector_store import (
+    ALL_MINILM_L6_V2_DIMENSION,
+    BankWithIndex,
+    EmbeddingIndex,
+)
+
+from .config import PGVectorConfig
+
+
+def check_extension_version(cur):
+    cur.execute("SELECT extversion FROM pg_extension WHERE extname = 'vector'")
+    result = cur.fetchone()
+    return result[0] if result else None
+
+
+def upsert_models(cur, keys_models: List[Tuple[str, BaseModel]]):
+    query = sql.SQL(
+        """
+        INSERT INTO metadata_store (key, data)
+        VALUES %s
+        ON CONFLICT (key) DO UPDATE
+        SET data = EXCLUDED.data
+    """
+    )
+
+    values = [(key, Json(model.dict())) for key, model in keys_models]
+    execute_values(cur, query, values, template="(%s, %s)")
+
+
+def load_models(cur, cls):
+    cur.execute("SELECT key, data FROM metadata_store")
+    rows = cur.fetchall()
+    return [parse_obj_as(cls, row["data"]) for row in rows]
+
+
+class PGVectorIndex(EmbeddingIndex):
+    def __init__(self, bank: MemoryBankDef, dimension: int, cursor):
+        self.cursor = cursor
+        self.table_name = f"vector_store_{bank.identifier}"
+
+        self.cursor.execute(
+            f"""
+            CREATE TABLE IF NOT EXISTS {self.table_name} (
+                id TEXT PRIMARY KEY,
+                document JSONB,
+                embedding vector({dimension})
+            )
+        """
+        )
+
+    async def add_chunks(self, chunks: List[Chunk], embeddings: NDArray):
+        assert len(chunks) == len(
+            embeddings
+        ), f"Chunk length {len(chunks)} does not match embedding length {len(embeddings)}"
+
+        values = []
+        for i, chunk in enumerate(chunks):
+            values.append(
+                (
+                    f"{chunk.document_id}:chunk-{i}",
+                    Json(chunk.dict()),
+                    embeddings[i].tolist(),
+                )
+            )
+
+        query = sql.SQL(
+            f"""
+        INSERT INTO {self.table_name} (id, document, embedding)
+        VALUES %s
+        ON CONFLICT (id) DO UPDATE SET embedding = EXCLUDED.embedding, document = EXCLUDED.document
+    """
+        )
+        execute_values(self.cursor, query, values, template="(%s, %s, %s::vector)")
+
+    async def query(
+        self, embedding: NDArray, k: int, score_threshold: float
+    ) -> QueryDocumentsResponse:
+        self.cursor.execute(
+            f"""
+        SELECT document, embedding <-> %s::vector AS distance
+        FROM {self.table_name}
+        ORDER BY distance
+        LIMIT %s
+    """,
+            (embedding.tolist(), k),
+        )
+        results = self.cursor.fetchall()
+
+        chunks = []
+        scores = []
+        for doc, dist in results:
+            chunks.append(Chunk(**doc))
+            scores.append(1.0 / float(dist))
+
+        return QueryDocumentsResponse(chunks=chunks, scores=scores)
+
+
+class PGVectorMemoryAdapter(Memory, MemoryBanksProtocolPrivate):
+    def __init__(self, config: PGVectorConfig) -> None:
+        self.config = config
+        self.cursor = None
+        self.conn = None
+        self.cache = {}
+
+    async def initialize(self) -> None:
+        try:
+            self.conn = psycopg2.connect(
+                host=self.config.host,
+                port=self.config.port,
+                database=self.config.db,
+                user=self.config.user,
+                password=self.config.password,
+            )
+            self.conn.autocommit = True
+            self.cursor = self.conn.cursor(cursor_factory=psycopg2.extras.DictCursor)
+
+            version = check_extension_version(self.cursor)
+            if version:
+                print(f"Vector extension version: {version}")
+            else:
+                raise RuntimeError("Vector extension is not installed.")
+
+            self.cursor.execute(
+                """
+                CREATE TABLE IF NOT EXISTS metadata_store (
+                    key TEXT PRIMARY KEY,
+                    data JSONB
+                )
+            """
+            )
+        except Exception as e:
+            import traceback
+
+            traceback.print_exc()
+            raise RuntimeError("Could not connect to PGVector database server") from e
+
+    async def shutdown(self) -> None:
+        pass
+
+    async def register_memory_bank(
+        self,
+        memory_bank: MemoryBankDef,
+    ) -> None:
+        assert (
+            memory_bank.type == MemoryBankType.vector.value
+        ), f"Only vector banks are supported {memory_bank.type}"
+
+        upsert_models(
+            self.cursor,
+            [
+                (memory_bank.identifier, memory_bank),
+            ],
+        )
+
+        index = BankWithIndex(
+            bank=memory_bank,
+            index=PGVectorIndex(memory_bank, ALL_MINILM_L6_V2_DIMENSION, self.cursor),
+        )
+        self.cache[memory_bank.identifier] = index
+
+    async def list_memory_banks(self) -> List[MemoryBankDef]:
+        banks = load_models(self.cursor, MemoryBankDef)
+        for bank in banks:
+            if bank.identifier not in self.cache:
+                index = BankWithIndex(
+                    bank=bank,
+                    index=PGVectorIndex(bank, ALL_MINILM_L6_V2_DIMENSION, self.cursor),
+                )
+                self.cache[bank.identifier] = index
+        return banks
+
+    async def insert_documents(
+        self,
+        bank_id: str,
+        documents: List[MemoryBankDocument],
+        ttl_seconds: Optional[int] = None,
+    ) -> None:
+        index = self.cache.get(bank_id, None)
+        if not index:
+            raise ValueError(f"Bank {bank_id} not found")
+
+        await index.insert_documents(documents)
+
+    async def query_documents(
+        self,
+        bank_id: str,
+        query: InterleavedTextMedia,
+        params: Optional[Dict[str, Any]] = None,
+    ) -> QueryDocumentsResponse:
+        index = self.cache.get(bank_id, None)
+        if not index:
+            raise ValueError(f"Bank {bank_id} not found")
+
+        return await index.query_documents(query, params)

llama_stack/providers/remote/memory/qdrant/__init__.py

@@ -0,0 +1,15 @@
+# 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.
+
+from .config import QdrantConfig
+
+
+async def get_adapter_impl(config: QdrantConfig, _deps):
+    from .qdrant import QdrantVectorMemoryAdapter
+
+    impl = QdrantVectorMemoryAdapter(config)
+    await impl.initialize()
+    return impl

llama_stack/providers/remote/memory/qdrant/config.py

@@ -0,0 +1,25 @@
+# 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.
+
+from typing import Optional
+
+from llama_models.schema_utils import json_schema_type
+from pydantic import BaseModel
+
+
+@json_schema_type
+class QdrantConfig(BaseModel):
+    location: Optional[str] = None
+    url: Optional[str] = None
+    port: Optional[int] = 6333
+    grpc_port: int = 6334
+    prefer_grpc: bool = False
+    https: Optional[bool] = None
+    api_key: Optional[str] = None
+    prefix: Optional[str] = None
+    timeout: Optional[int] = None
+    host: Optional[str] = None
+    path: Optional[str] = None

llama_stack/providers/remote/memory/qdrant/qdrant.py

@@ -0,0 +1,170 @@
+# 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.
+
+import traceback
+import uuid
+from typing import Any, Dict, List
+
+from numpy.typing import NDArray
+from qdrant_client import AsyncQdrantClient, models
+from qdrant_client.models import PointStruct
+
+from llama_stack.providers.datatypes import MemoryBanksProtocolPrivate
+
+from llama_stack.apis.memory import *  # noqa: F403
+
+from llama_stack.providers.remote.memory.qdrant.config import QdrantConfig
+from llama_stack.providers.utils.memory.vector_store import (
+    BankWithIndex,
+    EmbeddingIndex,
+)
+
+CHUNK_ID_KEY = "_chunk_id"
+
+
+def convert_id(_id: str) -> str:
+    """
+    Converts any string into a UUID string based on a seed.
+
+    Qdrant accepts UUID strings and unsigned integers as point ID.
+    We use a seed to convert each string into a UUID string deterministically.
+    This allows us to overwrite the same point with the original ID.
+    """
+    return str(uuid.uuid5(uuid.NAMESPACE_DNS, _id))
+
+
+class QdrantIndex(EmbeddingIndex):
+    def __init__(self, client: AsyncQdrantClient, collection_name: str):
+        self.client = client
+        self.collection_name = collection_name
+
+    async def add_chunks(self, chunks: List[Chunk], embeddings: NDArray):
+        assert len(chunks) == len(
+            embeddings
+        ), f"Chunk length {len(chunks)} does not match embedding length {len(embeddings)}"
+
+        if not await self.client.collection_exists(self.collection_name):
+            await self.client.create_collection(
+                self.collection_name,
+                vectors_config=models.VectorParams(
+                    size=len(embeddings[0]), distance=models.Distance.COSINE
+                ),
+            )
+
+        points = []
+        for i, (chunk, embedding) in enumerate(zip(chunks, embeddings)):
+            chunk_id = f"{chunk.document_id}:chunk-{i}"
+            points.append(
+                PointStruct(
+                    id=convert_id(chunk_id),
+                    vector=embedding,
+                    payload={"chunk_content": chunk.model_dump()}
+                    | {CHUNK_ID_KEY: chunk_id},
+                )
+            )
+
+        await self.client.upsert(collection_name=self.collection_name, points=points)
+
+    async def query(
+        self, embedding: NDArray, k: int, score_threshold: float
+    ) -> QueryDocumentsResponse:
+        results = (
+            await self.client.query_points(
+                collection_name=self.collection_name,
+                query=embedding.tolist(),
+                limit=k,
+                with_payload=True,
+                score_threshold=score_threshold,
+            )
+        ).points
+
+        chunks, scores = [], []
+        for point in results:
+            assert isinstance(point, models.ScoredPoint)
+            assert point.payload is not None
+
+            try:
+                chunk = Chunk(**point.payload["chunk_content"])
+            except Exception:
+                traceback.print_exc()
+                continue
+
+            chunks.append(chunk)
+            scores.append(point.score)
+
+        return QueryDocumentsResponse(chunks=chunks, scores=scores)
+
+
+class QdrantVectorMemoryAdapter(Memory, MemoryBanksProtocolPrivate):
+    def __init__(self, config: QdrantConfig) -> None:
+        self.config = config
+        self.client = AsyncQdrantClient(**self.config.model_dump(exclude_none=True))
+        self.cache = {}
+
+    async def initialize(self) -> None:
+        pass
+
+    async def shutdown(self) -> None:
+        self.client.close()
+
+    async def register_memory_bank(
+        self,
+        memory_bank: MemoryBankDef,
+    ) -> None:
+        assert (
+            memory_bank.type == MemoryBankType.vector.value
+        ), f"Only vector banks are supported {memory_bank.type}"
+
+        index = BankWithIndex(
+            bank=memory_bank,
+            index=QdrantIndex(self.client, memory_bank.identifier),
+        )
+
+        self.cache[memory_bank.identifier] = index
+
+    async def list_memory_banks(self) -> List[MemoryBankDef]:
+        # Qdrant doesn't have collection level metadata to store the bank properties
+        # So we only return from the cache value
+        return [i.bank for i in self.cache.values()]
+
+    async def _get_and_cache_bank_index(self, bank_id: str) -> Optional[BankWithIndex]:
+        if bank_id in self.cache:
+            return self.cache[bank_id]
+
+        bank = await self.memory_bank_store.get_memory_bank(bank_id)
+        if not bank:
+            raise ValueError(f"Bank {bank_id} not found")
+
+        index = BankWithIndex(
+            bank=bank,
+            index=QdrantIndex(client=self.client, collection_name=bank_id),
+        )
+        self.cache[bank_id] = index
+        return index
+
+    async def insert_documents(
+        self,
+        bank_id: str,
+        documents: List[MemoryBankDocument],
+        ttl_seconds: Optional[int] = None,
+    ) -> None:
+        index = await self._get_and_cache_bank_index(bank_id)
+        if not index:
+            raise ValueError(f"Bank {bank_id} not found")
+
+        await index.insert_documents(documents)
+
+    async def query_documents(
+        self,
+        bank_id: str,
+        query: InterleavedTextMedia,
+        params: Optional[Dict[str, Any]] = None,
+    ) -> QueryDocumentsResponse:
+        index = await self._get_and_cache_bank_index(bank_id)
+        if not index:
+            raise ValueError(f"Bank {bank_id} not found")
+
+        return await index.query_documents(query, params)

llama_stack/providers/remote/memory/sample/__init__.py

@@ -0,0 +1,17 @@
+# 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.
+
+from typing import Any
+
+from .config import SampleConfig
+
+
+async def get_adapter_impl(config: SampleConfig, _deps) -> Any:
+    from .sample import SampleMemoryImpl
+
+    impl = SampleMemoryImpl(config)
+    await impl.initialize()
+    return impl

llama_stack/providers/remote/memory/sample/config.py

@@ -0,0 +1,12 @@
+# 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.
+
+from pydantic import BaseModel
+
+
+class SampleConfig(BaseModel):
+    host: str = "localhost"
+    port: int = 9999

llama_stack/providers/remote/memory/sample/sample.py

@@ -0,0 +1,23 @@
+# 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.
+
+from .config import SampleConfig
+
+
+from llama_stack.apis.memory import *  # noqa: F403
+
+
+class SampleMemoryImpl(Memory):
+    def __init__(self, config: SampleConfig):
+        self.config = config
+
+    async def register_memory_bank(self, memory_bank: MemoryBankDef) -> None:
+        # these are the memory banks the Llama Stack will use to route requests to this provider
+        # perform validation here if necessary
+        pass
+
+    async def initialize(self):
+        pass

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

@@ -0,0 +1,15 @@
+# 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.
+
+from .config import WeaviateConfig, WeaviateRequestProviderData  # noqa: F401
+
+
+async def get_adapter_impl(config: WeaviateConfig, _deps):
+    from .weaviate import WeaviateMemoryAdapter
+
+    impl = WeaviateMemoryAdapter(config)
+    await impl.initialize()
+    return impl

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

@@ -0,0 +1,16 @@
+# 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.
+
+from pydantic import BaseModel
+
+
+class WeaviateRequestProviderData(BaseModel):
+    weaviate_api_key: str
+    weaviate_cluster_url: str
+
+
+class WeaviateConfig(BaseModel):
+    pass

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

@@ -0,0 +1,192 @@
+# 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.
+import json
+
+from typing import Any, Dict, List, Optional
+
+import weaviate
+import weaviate.classes as wvc
+from numpy.typing import NDArray
+from weaviate.classes.init import Auth
+
+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.utils.memory.vector_store import (
+    BankWithIndex,
+    EmbeddingIndex,
+)
+
+from .config import WeaviateConfig, WeaviateRequestProviderData
+
+
+class WeaviateIndex(EmbeddingIndex):
+    def __init__(self, client: weaviate.Client, collection_name: str):
+        self.client = client
+        self.collection_name = collection_name
+
+    async def add_chunks(self, chunks: List[Chunk], embeddings: NDArray):
+        assert len(chunks) == len(
+            embeddings
+        ), f"Chunk length {len(chunks)} does not match embedding length {len(embeddings)}"
+
+        data_objects = []
+        for i, chunk in enumerate(chunks):
+            data_objects.append(
+                wvc.data.DataObject(
+                    properties={
+                        "chunk_content": chunk.json(),
+                    },
+                    vector=embeddings[i].tolist(),
+                )
+            )
+
+        # Inserting chunks into a prespecified Weaviate collection
+        collection = self.client.collections.get(self.collection_name)
+
+        # TODO: make this async friendly
+        collection.data.insert_many(data_objects)
+
+    async def query(
+        self, embedding: NDArray, k: int, score_threshold: float
+    ) -> QueryDocumentsResponse:
+        collection = self.client.collections.get(self.collection_name)
+
+        results = collection.query.near_vector(
+            near_vector=embedding.tolist(),
+            limit=k,
+            return_metadata=wvc.query.MetadataQuery(distance=True),
+        )
+
+        chunks = []
+        scores = []
+        for doc in results.objects:
+            chunk_json = doc.properties["chunk_content"]
+            try:
+                chunk_dict = json.loads(chunk_json)
+                chunk = Chunk(**chunk_dict)
+            except Exception:
+                import traceback
+
+                traceback.print_exc()
+                print(f"Failed to parse document: {chunk_json}")
+                continue
+
+            chunks.append(chunk)
+            scores.append(1.0 / doc.metadata.distance)
+
+        return QueryDocumentsResponse(chunks=chunks, scores=scores)
+
+
+class WeaviateMemoryAdapter(
+    Memory, NeedsRequestProviderData, MemoryBanksProtocolPrivate
+):
+    def __init__(self, config: WeaviateConfig) -> None:
+        self.config = config
+        self.client_cache = {}
+        self.cache = {}
+
+    def _get_client(self) -> weaviate.Client:
+        provider_data = self.get_request_provider_data()
+        assert provider_data is not None, "Request provider data must be set"
+        assert isinstance(provider_data, WeaviateRequestProviderData)
+
+        key = f"{provider_data.weaviate_cluster_url}::{provider_data.weaviate_api_key}"
+        if key in self.client_cache:
+            return self.client_cache[key]
+
+        client = weaviate.connect_to_weaviate_cloud(
+            cluster_url=provider_data.weaviate_cluster_url,
+            auth_credentials=Auth.api_key(provider_data.weaviate_api_key),
+        )
+        self.client_cache[key] = client
+        return client
+
+    async def initialize(self) -> None:
+        pass
+
+    async def shutdown(self) -> None:
+        for client in self.client_cache.values():
+            client.close()
+
+    async def register_memory_bank(
+        self,
+        memory_bank: MemoryBankDef,
+    ) -> None:
+        assert (
+            memory_bank.type == MemoryBankType.vector.value
+        ), f"Only vector banks are supported {memory_bank.type}"
+
+        client = self._get_client()
+
+        # Create collection if it doesn't exist
+        if not client.collections.exists(memory_bank.identifier):
+            client.collections.create(
+                name=memory_bank.identifier,
+                vectorizer_config=wvc.config.Configure.Vectorizer.none(),
+                properties=[
+                    wvc.config.Property(
+                        name="chunk_content",
+                        data_type=wvc.config.DataType.TEXT,
+                    ),
+                ],
+            )
+
+        index = BankWithIndex(
+            bank=memory_bank,
+            index=WeaviateIndex(client=client, collection_name=memory_bank.identifier),
+        )
+        self.cache[memory_bank.identifier] = index
+
+    async def list_memory_banks(self) -> List[MemoryBankDef]:
+        # TODO: right now the Llama Stack is the source of truth for these banks. That is
+        # not ideal. It should be Weaviate which is the source of truth. Unfortunately,
+        # list() happens at Stack startup when the Weaviate client (credentials) is not
+        # yet available. We need to figure out a way to make this work.
+        return [i.bank for i in self.cache.values()]
+
+    async def _get_and_cache_bank_index(self, bank_id: str) -> Optional[BankWithIndex]:
+        if bank_id in self.cache:
+            return self.cache[bank_id]
+
+        bank = await self.memory_bank_store.get_memory_bank(bank_id)
+        if not bank:
+            raise ValueError(f"Bank {bank_id} not found")
+
+        client = self._get_client()
+        if not client.collections.exists(bank_id):
+            raise ValueError(f"Collection with name `{bank_id}` not found")
+
+        index = BankWithIndex(
+            bank=bank,
+            index=WeaviateIndex(client=client, collection_name=bank_id),
+        )
+        self.cache[bank_id] = index
+        return index
+
+    async def insert_documents(
+        self,
+        bank_id: str,
+        documents: List[MemoryBankDocument],
+        ttl_seconds: Optional[int] = None,
+    ) -> None:
+        index = await self._get_and_cache_bank_index(bank_id)
+        if not index:
+            raise ValueError(f"Bank {bank_id} not found")
+
+        await index.insert_documents(documents)
+
+    async def query_documents(
+        self,
+        bank_id: str,
+        query: InterleavedTextMedia,
+        params: Optional[Dict[str, Any]] = None,
+    ) -> QueryDocumentsResponse:
+        index = await self._get_and_cache_bank_index(bank_id)
+        if not index:
+            raise ValueError(f"Bank {bank_id} not found")
+
+        return await index.query_documents(query, params)