Tight coupling with Anthropic even while developing agents with other frameworks or tools.

[khitab]

There is a strong dependency on ANTHROPIC_API_KEY.

For example, when someone wants to develop an agent with, let's say, Gemini, it gives the impression that by following the example, one can create the agent. By importing the proper libraries and setting appropriate environment variables, it should be doable. But when executing, an error is encountered.

The README does say that ANTHROPIC_API_KEY is required. Should the adapter be flexible enough to work with any LLM provider? How about supporting the popular ones?

1. Anthropic (already supported)
2. Google / Gemini
3. Open AI

Please provide your inputs. I can try decoupling if that is a better design and makes the adapter more flexible.

[khitab]

@jinooos, @mariagorskikh & @ayushchopra96 - please review and suggest. Thanks.

[rdata121]

Thanks, @khitab — we’ve addressed this in our PR.
The adapter can now supports multiple LLM providers out of the box, including Anthropic, OpenAI, Gemini, Groq, Mistral, Cohere, and Grok.

You just need to specify the llm_provider name and set the corresponding API key (e.g., OPENAI_API_KEY, GEMINI_API_KEY, etc.).
If a user doesn’t configure these, the system gracefully falls back to the original Anthropic-based flow to maintain backward compatibility.

This should make the adapter much more flexible and future-proof for anyone working with different LLM ecosystems. [https://github.com//pull/18)]

[khitab]

Good work, @rdata121
Thank you.