This would be perfect for the currently FREE Gemini 1.5 Pro and 2.0 Experimental models!

They already explain why they won't do individual API keys on their website. I don't remember where exactly but the gist of it was that they negotiate directly for access in bulk, and allowing for individuals to connect their API keys conflicts with this somehow; they are currently able to offer so much usage due to the good rates they negotiate, and this wouldn't be possible if they allowed for individual API keys. Not sure how true any of this is as I don't know how it actually works, but seems they already have addressed this request with a "no".

However, it does seem that this response leaves the door open for local models, seeing as they don't need to do any negotiations for us to use them on our own machines.

+ openrouter support.

This will allow users to choose from hundreds of models

+ Ollama support.

Seriously need Local LLM support, along with selecting which LLMs to use on a copilot vs agent basis.

Azure and Bedrock is needed

I would take this further and ask for the ability to point to an Openai compatible endpoint. This would allow you to use litellm which in turn allows companies to use their own security approved solutions such as Amazon Bedrock.

OpenAI compatible API needed

That would be awesome. Imagine Cascades abilities together with o1-preview or 01-mini

Hope it can support OpenAI-compatible API.

qwen 2.5 coder !