Everything you need: the engine, the web UI, an example pack, and the one-command setup. Runs on macOS and Linux.
The complete engine: ingest, datagen, training, retrieval, grounding, eval, serve, and UI. Includes example packs (Ramayana, Atterby).
Download slm-engine.tar.gz ↓
Environment requirements, one-command setup, and platform-specific instructions for macOS, Linux, and CUDA servers.
Read the install guide →
Train your first pack and chat with it in under 30 minutes, following the README alone from a clean machine.
Start the quickstart →
Environments, supported agents and LLMs, usage, pack system, operations, runbook, and known limitations.
Browse the docs →
System requirements
| Component | Minimum | Recommended |
|---|---|---|
| OS | macOS 13+ / Linux x86_64 or arm64 | macOS 14+ / Ubuntu 22.04+ |
| RAM | 16 GB | 64 GB+ (8B model + embeddings + index) |
| Python | 3.11 | 3.11 – 3.13 |
| GPU | None (CPU works) | Apple Silicon / NVIDIA CUDA |
| Disk | 5 GB | 50 GB+ (models + corpora) |
| Model server | — | ollama, vLLM, or MLX |