Support AWS Bedrock

[pontus-devoteam]

I would like to run this through AWS Bedrock instance, instead of your API.

[rm-openai]

You can use any model or model provider. Easy way is to use an OpenAI compatible endpoint with base_url. Hard way is to build your own instance of Model. Does that work?

[aristide1997]

Unfortunately bedrock is not compatible with OpenAI style api requests unless it’s through the bedrock gateway (3rd party). Is a native bedrock client implementation on the table?

[rm-openai]

Would be hard for the OpenAI team to prioritize this. The SDK has the abstractions to support it (i.e. the Model interface), so we'd be open to PRs adding support via that interface.

[pontus-devoteam]

I will have a look and see how I could contribute on that, one way would be to add support for Lite LLM, they take care of all the variations of schemas, authentication etc.

[pontus-devoteam]

Created a PR for to support this #125

[rm-openai]

Thanks, this is a great PR. I think the right move is to add a separate "extensions" library that is an optional dependency, and this there. Let me talk to the team internally and see what they think!

[pontus-devoteam]

Yeah from a dependancy standpoint I would understand, but its also kinda part of the core of the Agent SDK to give the developer choice of what provider it wants to use, more then offer a base_url and api_key as option, that would not be close to acceptance for many providers in terms of configurations etc.

Also a big hassle Lite LLM solves is the rate limit management for each provider, so on and so forth you get my memo.

Happy to contribute

[pontus-devoteam]

In the meantime if anyone wants to build using Go lang, you could use:
https://github.com/pontus-devoteam/agent-sdk-go

[github-actions]

This issue is stale because it has been open for 7 days with no activity.

[EnggQasim]

Created a PR for to support this #125

I love to see Litellm contribution with openai-agents SDK Thank you @pontus-devoteam