knowledge-packs

crystal amber-framework llm apple-silicon mlx evaluation fine-tuning knowledge-pack local-llm lora qlora synthetic-data

Fine-tune a small local model on your docs and prove it actually learned: versioned knowledge packs (curated sources + generated training JSONL + contamination-guarded held-out eval + JSON manifest), the packsmith Crystal CLI, MLX LoRA recipes for Apple Silicon, and measured before/after eval reports.