mirror of
https://github.com/meta-llama/llama-stack.git
synced 2025-10-13 14:08:39 +00:00
ecc8a554d2
Some checks failed
Integration Auth Tests / test-matrix (oauth2_token) (push) Failing after 0s
Python Package Build Test / build (3.12) (push) Failing after 1s
Unit Tests / unit-tests (3.13) (push) Failing after 4s
SqlStore Integration Tests / test-postgres (3.12) (push) Failing after 0s
SqlStore Integration Tests / test-postgres (3.13) (push) Failing after 0s
Test External Providers Installed via Module / test-external-providers-from-module (venv) (push) Has been skipped
Python Package Build Test / build (3.13) (push) Failing after 1s
Integration Tests (Replay) / Integration Tests (, , , client=, ) (push) Failing after 3s
Vector IO Integration Tests / test-matrix (push) Failing after 5s
Test External API and Providers / test-external (venv) (push) Failing after 5s
Unit Tests / unit-tests (3.12) (push) Failing after 4s
API Conformance Tests / check-schema-compatibility (push) Successful in 10s
UI Tests / ui-tests (22) (push) Successful in 40s
Pre-commit / pre-commit (push) Successful in 1m23s
Applies the same pattern from https://github.com/llamastack/llama-stack/pull/3777 to embeddings and vector_stores.create() endpoints. This should _not_ be a breaking change since (a) our tests were already using the `extra_body` parameter when passing in to the backend (b) but the backend probably wasn't extracting the parameters correctly. This PR will fix that. Updated APIs: `openai_embeddings(), openai_create_vector_store(), openai_create_vector_store_file_batch()`
102 lines
4.4 KiB
Python
102 lines
4.4 KiB
Python
# 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 collections.abc import Iterable
|
|
|
|
from together import AsyncTogether
|
|
from together.constants import BASE_URL
|
|
|
|
from llama_stack.apis.inference import (
|
|
OpenAIEmbeddingsRequestWithExtraBody,
|
|
OpenAIEmbeddingsResponse,
|
|
)
|
|
from llama_stack.apis.inference.inference import OpenAIEmbeddingUsage
|
|
from llama_stack.apis.models import Model
|
|
from llama_stack.core.request_headers import NeedsRequestProviderData
|
|
from llama_stack.log import get_logger
|
|
from llama_stack.providers.utils.inference.openai_mixin import OpenAIMixin
|
|
|
|
from .config import TogetherImplConfig
|
|
|
|
logger = get_logger(name=__name__, category="inference::together")
|
|
|
|
|
|
class TogetherInferenceAdapter(OpenAIMixin, NeedsRequestProviderData):
|
|
config: TogetherImplConfig
|
|
|
|
embedding_model_metadata: dict[str, dict[str, int]] = {
|
|
"togethercomputer/m2-bert-80M-32k-retrieval": {"embedding_dimension": 768, "context_length": 32768},
|
|
"BAAI/bge-large-en-v1.5": {"embedding_dimension": 1024, "context_length": 512},
|
|
"BAAI/bge-base-en-v1.5": {"embedding_dimension": 768, "context_length": 512},
|
|
"Alibaba-NLP/gte-modernbert-base": {"embedding_dimension": 768, "context_length": 8192},
|
|
"intfloat/multilingual-e5-large-instruct": {"embedding_dimension": 1024, "context_length": 512},
|
|
}
|
|
|
|
_model_cache: dict[str, Model] = {}
|
|
|
|
provider_data_api_key_field: str = "together_api_key"
|
|
|
|
def get_base_url(self):
|
|
return BASE_URL
|
|
|
|
def _get_client(self) -> AsyncTogether:
|
|
together_api_key = None
|
|
config_api_key = self.config.auth_credential.get_secret_value() if self.config.auth_credential else None
|
|
if config_api_key:
|
|
together_api_key = config_api_key
|
|
else:
|
|
provider_data = self.get_request_provider_data()
|
|
if provider_data is None or not provider_data.together_api_key:
|
|
raise ValueError(
|
|
'Pass Together API Key in the header X-LlamaStack-Provider-Data as { "together_api_key": <your api key>}'
|
|
)
|
|
together_api_key = provider_data.together_api_key
|
|
return AsyncTogether(api_key=together_api_key)
|
|
|
|
async def list_provider_model_ids(self) -> Iterable[str]:
|
|
# Together's /v1/models is not compatible with OpenAI's /v1/models. Together support ticket #13355 -> will not fix, use Together's own client
|
|
return [m.id for m in await self._get_client().models.list()]
|
|
|
|
async def openai_embeddings(
|
|
self,
|
|
params: OpenAIEmbeddingsRequestWithExtraBody,
|
|
) -> OpenAIEmbeddingsResponse:
|
|
"""
|
|
Together's OpenAI-compatible embeddings endpoint is not compatible with
|
|
the standard OpenAI embeddings endpoint.
|
|
|
|
The endpoint -
|
|
- not all models return usage information
|
|
- does not support user param, returns 400 Unrecognized request arguments supplied: user
|
|
- does not support dimensions param, returns 400 Unrecognized request arguments supplied: dimensions
|
|
"""
|
|
# Together support ticket #13332 -> will not fix
|
|
if params.user is not None:
|
|
raise ValueError("Together's embeddings endpoint does not support user param.")
|
|
# Together support ticket #13333 -> escalated
|
|
if params.dimensions is not None:
|
|
raise ValueError("Together's embeddings endpoint does not support dimensions param.")
|
|
|
|
response = await self.client.embeddings.create(
|
|
model=await self._get_provider_model_id(params.model),
|
|
input=params.input,
|
|
encoding_format=params.encoding_format,
|
|
)
|
|
|
|
response.model = (
|
|
params.model
|
|
) # return the user the same model id they provided, avoid exposing the provider model id
|
|
|
|
# Together support ticket #13330 -> escalated
|
|
# - togethercomputer/m2-bert-80M-32k-retrieval *does not* return usage information
|
|
if not hasattr(response, "usage") or response.usage is None:
|
|
logger.warning(
|
|
f"Together's embedding endpoint for {params.model} did not return usage information, substituting -1s."
|
|
)
|
|
response.usage = OpenAIEmbeddingUsage(prompt_tokens=-1, total_tokens=-1)
|
|
|
|
return response # type: ignore[no-any-return]
|