taykso/slurpee-inference

Gemma 3 4B IT QAT (Q4_0) running via llama.cpp server.

API endpoint: /v1/chat/completions (OpenAI-compatible)