Open-Source AI Stack — Companion Code

Working code for every snippet referenced in The Open-Source AI Stack, the production playbook from AmtocSoft.

Every example is tested in CI. PRs welcome.

Layout

Directory	Chapter	What's in it
ch02_local_serving/	Ch. 2 — Running LLMs Locally	Ollama API client, llama.cpp invocation, hardware sizing helper
ch04_lora_qlora/	Ch. 4 — LoRA & QLoRA	Unsloth training script, eval harness, hyperparameter sweep
ch06_rag_pipeline/	Ch. 6 — Production RAG	Hybrid retrieval, reranking, citation rendering
ch07_domain_ai/	Ch. 7 — Domain AI Stack	Reference implementation: legal contract analyzer
ch08_quantization/	Ch. 8 — Quantization	GGUF/GPTQ/AWQ benchmark scripts
common/	—	Shared utils: logging, config, CLI
scripts/	—	One-off scripts for setup, downloads, etc.
tests/	—	pytest suite that runs every example

Quick start

# Clone
git clone https://github.com/amtocbot-droid/open-source-ai-stack-code.git
cd open-source-ai-stack-code

# Install (uv recommended)
uv venv
source .venv/bin/activate
uv pip install -e .

# Verify
pytest tests/ -v

License

MIT. Use freely in your own projects, including commercial.

If the book taught you something, tell us or book a strategy call. The book funds this code, the code funds the next book.

Maintained by

AmtocSoft · @amtocbot-droid