phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-12-18 17:28:40 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

user inference api to generate embeddings in vector store

Browse source
This commit is contained in:
Dinesh Yeduguru 2024-12-09 12:49:35 -08:00
parent 96accc1216
commit 5bbeb985ca
12 changed files with 134 additions and 96 deletions
Download patch file Download diff file Expand all files Collapse all files

7
llama_stack/providers/remote/memory/chroma/__init__.py
View file

@ -4,12 +4,15 @@
# This source code is licensed under the terms described in the LICENSE file in
# the root directory of this source tree.
from typing import Dict
from llama_stack.distribution.datatypes import RemoteProviderConfig
from llama_stack.providers.datatypes import Api, ProviderSpec
async def get_adapter_impl(config: RemoteProviderConfig, _deps):
async def get_adapter_impl(config: RemoteProviderConfig, deps: Dict[Api, ProviderSpec]):
from .chroma import ChromaMemoryAdapter
impl = ChromaMemoryAdapter(config.url)
impl = ChromaMemoryAdapter(config.url, deps[Api.inference])
await impl.initialize()
return impl

20
llama_stack/providers/remote/memory/chroma/chroma.py
View file

@ -20,6 +20,7 @@ from llama_stack.providers.datatypes import MemoryBanksProtocolPrivate
from llama_stack.providers.utils.memory.vector_store import (
BankWithIndex,
EmbeddingIndex,
InferenceEmbeddingMixin,
)
log = logging.getLogger(__name__)
@ -71,8 +72,8 @@ class ChromaIndex(EmbeddingIndex):
await self.client.delete_collection(self.collection.name)
class ChromaMemoryAdapter(Memory, MemoryBanksProtocolPrivate):
def __init__(self, url: str) -> None:
class ChromaMemoryAdapter(InferenceEmbeddingMixin, Memory, MemoryBanksProtocolPrivate):
def __init__(self, url: str, inference_api: Api.inference) -> None:
log.info(f"Initializing ChromaMemoryAdapter with url: {url}")
url = url.rstrip("/")
parsed = urlparse(url)
@ -82,6 +83,7 @@ class ChromaMemoryAdapter(Memory, MemoryBanksProtocolPrivate):
self.host = parsed.hostname
self.port = parsed.port
self.inference_api = inference_api
self.client = None
self.cache = {}
@ -109,10 +111,9 @@ class ChromaMemoryAdapter(Memory, MemoryBanksProtocolPrivate):
name=memory_bank.identifier,
metadata={"bank": memory_bank.model_dump_json()},
)
bank_index = BankWithIndex(
bank=memory_bank, index=ChromaIndex(self.client, collection)
self.cache[memory_bank.identifier] = self._create_bank_with_index(
memory_bank, ChromaIndex(self.client, collection)
)
self.cache[memory_bank.identifier] = bank_index
async def list_memory_banks(self) -> List[MemoryBank]:
collections = await self.client.list_collections()
@ -124,11 +125,10 @@ class ChromaMemoryAdapter(Memory, MemoryBanksProtocolPrivate):
log.exception(f"Failed to parse bank: {collection.metadata}")
continue
index = BankWithIndex(
bank=bank,
index=ChromaIndex(self.client, collection),
self.cache[bank.identifier] = self._create_bank_with_index(
bank,
ChromaIndex(self.client, collection),
)
self.cache[bank.identifier] = index
return [i.bank for i in self.cache.values()]
@ -166,6 +166,6 @@ class ChromaMemoryAdapter(Memory, MemoryBanksProtocolPrivate):
collection = await self.client.get_collection(bank_id)
if not collection:
raise ValueError(f"Bank {bank_id} not found in Chroma")
index = BankWithIndex(bank=bank, index=ChromaIndex(self.client, collection))
index = self._create_bank_with_index(bank, ChromaIndex(self.client, collection))
self.cache[bank_id] = index
return index

8
llama_stack/providers/remote/memory/pgvector/__init__.py
View file

@ -4,12 +4,16 @@
# This source code is licensed under the terms described in the LICENSE file in
# the root directory of this source tree.
from typing import Dict
from llama_stack.providers.datatypes import Api, ProviderSpec
from .config import PGVectorConfig
async def get_adapter_impl(config: PGVectorConfig, _deps):
async def get_adapter_impl(config: PGVectorConfig, deps: Dict[Api, ProviderSpec]):
from .pgvector import PGVectorMemoryAdapter
impl = PGVectorMemoryAdapter(config)
impl = PGVectorMemoryAdapter(config, deps[Api.inference])
await impl.initialize()
return impl

49
llama_stack/providers/remote/memory/pgvector/pgvector.py
View file

@ -16,11 +16,12 @@ 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.datatypes import Api, MemoryBanksProtocolPrivate
from llama_stack.providers.utils.memory.vector_store import (
ALL_MINILM_L6_V2_DIMENSION,
BankWithIndex,
EmbeddingIndex,
InferenceEmbeddingMixin,
)
from .config import PGVectorConfig
@ -119,9 +120,12 @@ class PGVectorIndex(EmbeddingIndex):
self.cursor.execute(f"DROP TABLE IF EXISTS {self.table_name}")
class PGVectorMemoryAdapter(Memory, MemoryBanksProtocolPrivate):
def __init__(self, config: PGVectorConfig) -> None:
class PGVectorMemoryAdapter(
InferenceEmbeddingMixin, Memory, MemoryBanksProtocolPrivate
):
def __init__(self, config: PGVectorConfig, inference_api: Api.inference) -> None:
self.config = config
self.inference_api = inference_api
self.cursor = None
self.conn = None
self.cache = {}
@ -160,27 +164,17 @@ class PGVectorMemoryAdapter(Memory, MemoryBanksProtocolPrivate):
async def shutdown(self) -> None:
pass
async def register_memory_bank(
self,
memory_bank: MemoryBank,
) -> None:
async def register_memory_bank(self, memory_bank: MemoryBank) -> None:
assert (
memory_bank.memory_bank_type == MemoryBankType.vector.value
), f"Only vector banks are supported {memory_bank.memory_bank_type}"
upsert_models(
self.cursor,
[
(memory_bank.identifier, memory_bank),
],
upsert_models(self.cursor, [(memory_bank.identifier, memory_bank)])
index = PGVectorIndex(memory_bank, memory_bank.embedding_dimension, self.cursor)
self.cache[memory_bank.identifier] = self._create_bank_with_index(
memory_bank, index
)
index = BankWithIndex(
bank=memory_bank,
index=PGVectorIndex(memory_bank, ALL_MINILM_L6_V2_DIMENSION, self.cursor),
)
self.cache[memory_bank.identifier] = index
async def unregister_memory_bank(self, memory_bank_id: str) -> None:
await self.cache[memory_bank_id].index.delete()
del self.cache[memory_bank_id]
@ -189,9 +183,9 @@ class PGVectorMemoryAdapter(Memory, MemoryBanksProtocolPrivate):
banks = load_models(self.cursor, VectorMemoryBank)
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),
index = self._create_bank_with_index(
bank,
PGVectorIndex(bank, bank.embedding_dimension, self.cursor),
)
self.cache[bank.identifier] = index
return banks
@ -214,14 +208,13 @@ class PGVectorMemoryAdapter(Memory, MemoryBanksProtocolPrivate):
index = await self._get_and_cache_bank_index(bank_id)
return await index.query_documents(query, params)
self.inference_api = inference_api
async def _get_and_cache_bank_index(self, bank_id: str) -> BankWithIndex:
if bank_id in self.cache:
return self.cache[bank_id]
bank = await self.memory_bank_store.get_memory_bank(bank_id)
index = BankWithIndex(
bank=bank,
index=PGVectorIndex(bank, ALL_MINILM_L6_V2_DIMENSION, self.cursor),
)
self.cache[bank_id] = index
return index
index = PGVectorIndex(bank, bank.embedding_dimension, self.cursor)
self.cache[bank_id] = self._create_bank_with_index(bank, index)
return self.cache[bank_id]

8
llama_stack/providers/remote/memory/qdrant/__init__.py
View file

@ -4,12 +4,16 @@
# This source code is licensed under the terms described in the LICENSE file in
# the root directory of this source tree.
from typing import Dict
from llama_stack.providers.datatypes import Api, ProviderSpec
from .config import QdrantConfig
async def get_adapter_impl(config: QdrantConfig, _deps):
async def get_adapter_impl(config: QdrantConfig, deps: Dict[Api, ProviderSpec]):
from .qdrant import QdrantVectorMemoryAdapter
impl = QdrantVectorMemoryAdapter(config)
impl = QdrantVectorMemoryAdapter(config, deps[Api.inference])
await impl.initialize()
return impl

12
llama_stack/providers/remote/memory/qdrant/qdrant.py
View file

@ -21,6 +21,7 @@ from llama_stack.providers.remote.memory.qdrant.config import QdrantConfig
from llama_stack.providers.utils.memory.vector_store import (
BankWithIndex,
EmbeddingIndex,
InferenceEmbeddingMixin,
)
log = logging.getLogger(__name__)
@ -100,11 +101,14 @@ class QdrantIndex(EmbeddingIndex):
return QueryDocumentsResponse(chunks=chunks, scores=scores)
class QdrantVectorMemoryAdapter(Memory, MemoryBanksProtocolPrivate):
def __init__(self, config: QdrantConfig) -> None:
class QdrantVectorMemoryAdapter(
InferenceEmbeddingMixin, Memory, MemoryBanksProtocolPrivate
):
def __init__(self, config: QdrantConfig, inference_api: Api.inference) -> None:
self.config = config
self.client = AsyncQdrantClient(**self.config.model_dump(exclude_none=True))
self.cache = {}
self.inference_api = inference_api
async def initialize(self) -> None:
pass
@ -120,7 +124,7 @@ class QdrantVectorMemoryAdapter(Memory, MemoryBanksProtocolPrivate):
memory_bank.memory_bank_type == MemoryBankType.vector
), f"Only vector banks are supported {memory_bank.memory_bank_type}"
index = BankWithIndex(
index = self._create_bank_with_index(
bank=memory_bank,
index=QdrantIndex(self.client, memory_bank.identifier),
)
@ -140,7 +144,7 @@ class QdrantVectorMemoryAdapter(Memory, MemoryBanksProtocolPrivate):
if not bank:
raise ValueError(f"Bank {bank_id} not found")
index = BankWithIndex(
index = self._create_bank_with_index(
bank=bank,
index=QdrantIndex(client=self.client, collection_name=bank_id),
)

8
llama_stack/providers/remote/memory/weaviate/__init__.py
View file

@ -4,12 +4,16 @@
# This source code is licensed under the terms described in the LICENSE file in
# the root directory of this source tree.
from typing import Dict
from llama_stack.providers.datatypes import Api, ProviderSpec
from .config import WeaviateConfig, WeaviateRequestProviderData # noqa: F401
async def get_adapter_impl(config: WeaviateConfig, _deps):
async def get_adapter_impl(config: WeaviateConfig, deps: Dict[Api, ProviderSpec]):
from .weaviate import WeaviateMemoryAdapter
impl = WeaviateMemoryAdapter(config)
impl = WeaviateMemoryAdapter(config, deps[Api.inference])
await impl.initialize()
return impl

18
llama_stack/providers/remote/memory/weaviate/weaviate.py
View file

@ -19,6 +19,7 @@ from llama_stack.providers.datatypes import MemoryBanksProtocolPrivate
from llama_stack.providers.utils.memory.vector_store import (
BankWithIndex,
EmbeddingIndex,
InferenceEmbeddingMixin,
)
from .config import WeaviateConfig, WeaviateRequestProviderData
@ -82,10 +83,14 @@ class WeaviateIndex(EmbeddingIndex):
class WeaviateMemoryAdapter(
Memory, NeedsRequestProviderData, MemoryBanksProtocolPrivate
InferenceEmbeddingMixin,
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 = {}
@ -135,11 +140,10 @@ class WeaviateMemoryAdapter(
],
)
index = BankWithIndex(
bank=memory_bank,
index=WeaviateIndex(client=client, collection_name=memory_bank.identifier),
self.cache[memory_bank.identifier] = self._create_bank_with_index(
memory_bank,
WeaviateIndex(client=client, collection_name=memory_bank.identifier),
)
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
@ -160,7 +164,7 @@ class WeaviateMemoryAdapter(
if not client.collections.exists(bank.identifier):
raise ValueError(f"Collection with name `{bank.identifier}` not found")
index = BankWithIndex(
index = self._create_bank_with_index(
bank=bank,
index=WeaviateIndex(client=client, collection_name=bank_id),
)