# Host directory path where GGUF models are stored
MODELS_PATH=/pwspool/software/ai/llama_cpp

# Filename of the default model to load
MODEL_FILE=Llama-3.1-8B-Instruct-Q4_K_M.gguf

# Optional API Key for securing endpoints (leave blank if internal-only, or set a secure token)
API_KEY=your_super_secret_api_key_here

# Performance and hardware tuning
GPU_LAYERS=999
CTX_SIZE=4096
THREADS=6