vault-ai

Why we included this project

If you want to ask natural-language questions about a private set of documents, this project shows you the whole pipeline in one place. A Go server does the processing while a React frontend handles uploads: feed it PDFs, EPUBs, or plain text, and it splits the content into chunks, embeds each chunk through OpenAI, and stores the vectors in Pinecone. Ask a question and the same path retrieves the relevant sections and passes them to the model, so answers arrive with the filename and the exact passage they came from. Because the codebase is small, reading the upload and question handlers is a quick way to see chunking, embedding, and retrieval working together without a lot of abstraction.