lastmile-ai/mcp-agent
dottxt-ai/outlines
getomni-ai/zeroxFine-tuning takes a capable base model and nudges it toward your task using your own examples. Full fine-tuning updates every weight; parameter-efficient methods like LoRA train only small adapters, making it far cheaper. Teams reach for fine-tuning when prompt engineering and retrieval are not enough to get consistent behaviour or a specific style.
Open-source fine-tuning frameworks handle the training loop, data formatting, and adapter management. In practice the dataset matters more than the algorithm: a few hundred high-quality, well-labelled examples usually beat a large noisy set.
