Support AWS bedrock LLM

[taranjeet]

🚀 The feature

https://python.langchain.com/docs/integrations/platforms/aws

Motivation, pitch

• We need to support more llms so that users have the flexibility of choosing whatever LLM they want to use.
• Langchain provides a standard interface for llms. This helps in better code maintenance and also helps in better observability.

[KapilM26]

@taranjeet please assign this to me

[krrishdholakia]

Hey @taranjeet @KapilM26 @sidmohanty11 you can do this with ChatLiteLLM in Langchain

from langchain.chat_models import ChatLiteLLM
from langchain.schema import AIMessage, HumanMessage, SystemMessage

chat = ChatLiteLLM(model="bedrock/anthropic.claude-v2")

messages = [
    HumanMessage(
        content="Translate this sentence from English to French. I love programming."
    )
]
chat(messages)

Benefit over using the bedrock integration? Bedrock supports:

• Anthropic
• Cohere
• Amazon Titan
• AI21

Each has their own prompt formatting and input params (langchain doesn't natively map this).

Langchain Docs: https://python.langchain.com/docs/integrations/chat/litellm LiteLLM Docs: https://docs.litellm.ai/docs/providers/bedrock

You can also pass in provider-specific params: https://docs.litellm.ai/docs/completion/input#provider-specific-params