Your friendly internet enabled assistant. Swap minds with custom prompts and Ollama
- Runs a model trained on function calling
- Exposes a
search_wikipediafunction in the system prompt - Listens for function call from model and injects result back in
- Gets model to respond using function results
- Set up Node.js 19.x or later.
- Download a copy of NousResearch/Hermes-2-Pro-Mistral-7B. Prefer the 7B Q4_K_M model if possible
- Compile and run the lamma.cpp example server. (You may wish to compile with GPU support for your particular hardware setup)
- Run the server with the following args:
./server --model /path/to/Hermes-2-Pro-Mistral-7B.Q4_K_M.gguf
- Run the mind goblin repl:
node index.js - Ask the mind goblin to do stuff for you