Just a congrats on a really, really helpful tool. It has completely changed the accuracy of of the output. What a time-saver.

Implement support for a custom base_url OpenAI-compatible endpoint. TLDR; 1. A request from the PromptDC extension would first be sent to an OpenAI-compatible endpoint, in my case, that's LiteLLM proxy. 2. The LiteLLM proxy would then forward that request directly to OpenAI-compatible (PromptDC backend) using my OpenAI key. 3. Upon receiving my request, PromptDC backend enhances the prompt with own instructions, and then forwards the enhanced request to OpenAI using my OpenAI key observed at step 2. 4. Subsequently, everything would flow in reverse: the response from OpenAI would go to your backend, then to LiteLLM, and finally arrive back at PromptDC extension.

How can we get this tool to properly work with cascade? thanks!