Infernum

"From the depths, intelligence rises"

Local LLM inference CLI for running large language models on your machine.

Quick Start

# Install
cargo install infernum

# Set your model
infernum config set-model TinyLlama/TinyLlama-1.1B-Chat-v1.0

# Start chatting
infernum chat

Features

• Local Inference: Run LLMs entirely on your machine - no API keys, no cloud
• Standard API: Industry-standard /v1/* server at localhost:8080
• Interactive Chat: Full-featured CLI chat with history and session management
• Multi-Backend: CPU, CUDA (NVIDIA), and Metal (Apple Silicon) support
• Smart Caching: Models download once via HuggingFace Hub
• Streaming: Real-time token-by-token output

Commands

# Chat interface
infernum chat [--model MODEL] [--system PROMPT]

# Start API server
infernum server [--port PORT]

# Download a model
infernum pull meta-llama/Llama-3.2-3B-Instruct

# List available models
infernum list

# System diagnostics
infernum doctor

# Configuration
infernum config set-model MODEL
infernum config get-model

GPU Support

# Install with CUDA support
cargo install infernum --features cuda

# Install with Metal support (macOS)
cargo install infernum --features metal

Architecture

Infernum is built from specialized components:

• abaddon: High-performance inference engine
• malphas: Model orchestration and lifecycle management
• stolas: Knowledge retrieval and RAG capabilities
• beleth: Agent framework for autonomous AI
• asmodeus: Model fine-tuning and adaptation
• dantalion: Observability and telemetry
• infernum-server: HTTP API server

Documentation

Full documentation available at infernum.daemoniorum.com

License

Licensed under either of Apache License, Version 2.0 or MIT license at your option.