Tianji

Why we included this project

Tianji teaches the full arc of LLM application development, from prompt writing to fine-tuning, using Chinese social situations as the working example throughout. Instead of stopping at API calls, it walks through building a RAG knowledge base with LangChain or LlamaIndex, standing up an agent on MetaGPT with tool calling, and running LoRA or full fine-tuning with Transformers and Xtuner. The repo bundles Chinese conversational datasets plus scripts for generating and cleaning your own training data, so you do not have to hunt for examples. The finished 'send greetings' models show what the fine-tuning workflow produces end to end, which makes it easier to see how the stages connect. It is a learning resource rather than a production-ready product, and for teams new to LLM development that is exactly the point.