57 lines
2.0 KiB
Python
57 lines
2.0 KiB
Python
from typing import Any, Optional
|
|
|
|
from langchain_aws import ChatBedrockConverse
|
|
|
|
from .base_client import BaseLLMClient
|
|
|
|
|
|
class BedrockClient(BaseLLMClient):
|
|
"""Client for Amazon Bedrock models.
|
|
|
|
Supports any model available on Bedrock via IAM Role (no API key needed),
|
|
including Claude, Amazon Nova, Kimi, Qwen, GLM, DeepSeek, MiniMax, and more.
|
|
|
|
Authentication:
|
|
Uses boto3 default credential chain: IAM Role (EC2/Lambda), environment
|
|
variables (AWS_ACCESS_KEY_ID / AWS_SECRET_ACCESS_KEY), or ~/.aws/credentials.
|
|
|
|
Model ID formats:
|
|
- Cross-region inference profile (recommended):
|
|
``global.anthropic.claude-sonnet-4-6``
|
|
``eu.anthropic.claude-3-5-sonnet-20240620-v1:0``
|
|
- Direct on-demand (us-east-1 default region only):
|
|
``amazon.nova-lite-v1:0``
|
|
``moonshotai.kimi-k2.5``
|
|
``qwen.qwen3-32b-v1:0``
|
|
``zai.glm-4.7-flash``
|
|
``deepseek.v3.2``
|
|
|
|
Note:
|
|
When specifying a non-default ``region_name``, use region-specific
|
|
inference profile IDs (e.g. ``us-west-2.anthropic.claude-...``),
|
|
as direct model IDs only support on-demand throughput in us-east-1.
|
|
|
|
Example::
|
|
|
|
config["llm_provider"] = "bedrock"
|
|
config["deep_think_llm"] = "global.anthropic.claude-sonnet-4-6"
|
|
config["quick_think_llm"] = "amazon.nova-micro-v1:0"
|
|
"""
|
|
|
|
def __init__(self, model: str, base_url: Optional[str] = None, **kwargs):
|
|
super().__init__(model, base_url, **kwargs)
|
|
|
|
def get_llm(self) -> Any:
|
|
"""Return configured ChatBedrockConverse instance."""
|
|
llm_kwargs: dict = {"model_id": self.model}
|
|
|
|
for key in ("region_name", "max_tokens", "callbacks", "timeout"):
|
|
if key in self.kwargs:
|
|
llm_kwargs[key] = self.kwargs[key]
|
|
|
|
return ChatBedrockConverse(**llm_kwargs)
|
|
|
|
def validate_model(self) -> bool:
|
|
"""Bedrock model IDs are dynamic; skip static validation."""
|
|
return True
|