danny-avila/rag_api

This project is a backend service that lets your application store documents and then search through them intelligently, meaning it finds relevant content based on meaning, not just keyword matches. The core idea is "retrieval-augmented generation" (RAG), which is how chatbots and AI assistants pull in specific information from your uploaded files before answering a question. Rather than a chatbot giving generic answers, it can cite and reference your actual documents. The project was built primarily to power document handling for LibreChat, an open-source AI chat platform, but it can serve as a document search layer for any application. When you upload a file, the service breaks it into smaller chunks, converts each chunk into a mathematical representation called an embedding (which captures the meaning of the text), and stores those embeddings in a database. Later, when someone asks a question, the system converts the question into an embedding too, then finds the stored chunks that are mathematically closest in meaning. Those relevant chunks get returned so your AI can use them to craft a grounded answer. Everything is organized by a unique file ID, so you can add, search, or remove documents on a per-file basis. The main audience is someone building an AI chat application or knowledge base who needs their assistant to reference specific uploaded documents. For example, a company might upload its HR policies and product manuals, then let employees ask a chatbot questions that draw answers from those files. A legal tech startup could let lawyers search through case documents by meaning rather than exact wording. The file-level approach means you can delete one document without affecting others, which matters when users manage their own files. The service supports multiple embedding providers (OpenAI, Hugging Face, Google, AWS Bedrock, Ollama, and others) and can use either PostgreSQL with the pgvector extension or MongoDB Atlas as its storage backend. It also handles practical concerns like batch processing for large files and memory limits in cloud environments, making it deployable in resource-constrained settings.