ShishirPatil/gorilla ? reverse-engineered prompt

Build me a project like Gorilla that helps a language model pick the right API or tool from a natural language question, then returns the exact function call it should make.

I want it to include a way to run inference with a Gorilla style model, plus an evaluation setup so I can test how well it handles function calling against a benchmark of APIs. If there are already example datasets or leaderboard style results in this repo, wire those up so I can reproduce them locally and try my own prompts too.

It should feel like a practical toolkit for tool calling, not just a paper demo, so include a simple way to chat with the model, check its output, and compare models or prompts. If anything depends on current docs or model formats, look up the latest info online before you implement it.