mirror of
https://github.com/meta-llama/llama-stack.git
synced 2025-12-12 20:12:33 +00:00
a0c0e5933a
Introduced a new embedding model with a dimension of 3072 and a context length of 2048 to the GeminiInferenceAdapter class.
22 lines
791 B
Python
22 lines
791 B
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 llama_stack.providers.utils.inference.openai_mixin import OpenAIMixin
|
|
|
|
from .config import GeminiConfig
|
|
|
|
|
|
class GeminiInferenceAdapter(OpenAIMixin):
|
|
config: GeminiConfig
|
|
|
|
provider_data_api_key_field: str = "gemini_api_key"
|
|
embedding_model_metadata: dict[str, dict[str, int]] = {
|
|
"models/text-embedding-004": {"embedding_dimension": 768, "context_length": 2048},
|
|
"models/gemini-embedding-001": {"embedding_dimension": 3072, "context_length": 2048},
|
|
}
|
|
|
|
def get_base_url(self):
|
|
return "https://generativelanguage.googleapis.com/v1beta/openai/"
|