Amazon Bedrock for LLM / Voice assistant and image recognition

[ToBeHH]

Describe the feature

As there is already an AWS integration, I would love to see it supporting Amazon Bedrock. Amazon Bedrock is an interface to various LLMs (Anthropic, Meta’s llama, Mistral, Amazon Nova and about 17 more providers). The LLMs are then executed in a private environment and you get charged only by the tokens you use. It would be a nice way to run big LLMs in a private environment in the cloud.

Use cases

• No need to install new integrations when trying out new LLMs
• Private LLM in the cloud
• Executing LLMs, I cannot use elsewhere
• Can be used for voice assistant and image recognition (multi-modal models)

Anything else?

Amazon provides a nice Python library (well, I guess you know, as some AWS services are already supported) which also supports Bedrock.

I saw, @paul-enno tried already in 2024, but the GitHub PR was never merged. Maybe someone can build up on this: Amazon Bedrock Conversation Agent Integration by paul-enno · Pull Request #112216 · home-assistant/core · GitHub ?