josephjelson06/football-intelligence-assistant

Football Intelligence Assistant is a small learning project that builds a working football chat application while teaching the core pieces of LangChain, a framework for connecting language models to external tools and data. The project is intentionally kept simple, aimed at helping one developer study prompts, chains, tools, retrieval augmented generation (a technique where the AI looks up information in a document store before answering, often called RAG), memory, and LangGraph, a way to lay out multi-step AI workflows as a graph of connected steps. The backend runs on FastAPI, a Python web framework, and uses LangChain and LangGraph to manage the AI logic. It calls Groq as the language model provider, uses Jina to generate text embeddings, which are numerical representations of text used for search, and stores those embeddings in ChromaDB, a local vector database. Three endpoints are exposed: a health check, a chat endpoint, and an endpoint for ingesting documents into the knowledge base. The frontend is a single chat page built with Next.js and Tailwind CSS, styled in a dark football analysis color scheme, with the earlier version's dashboard and extra navigation stripped out to keep things focused. The README lays out a suggested learning path through six backend files, from prompts to tools to retrieval to memory to the graph definition to the chat handler, so a reader can trace how the AI assistant is assembled step by step. This project suits someone who wants a compact, readable example of how LangChain's pieces fit together in a real chat application, using football as the subject matter rather than a generic demo.