aiRag AI Assistant

Upload your file (.txt, .pdf) — or drop a .key file to use an existing vector DB.

About This Project

Introduction: aiRag is a retrieval-augmented generation (RAG) chatbot built using FastAPI, LangChain, Hugging Face Embeddings, FAISS vector DB, and OpenAI API. It showcases how document-aware AI assistants can be deployed serverlessly and interactively.

Challenge: Implement a contextually aware chatbot that retrieves domain-specific information from indexed documents to supplement its responses.

Tech Stack: Python backend (LangChain, FAISS, HuggingFace Embeddings, Transformers, OpenAI), served via FastAPI on RunPod. Frontend: plain HTML/CSS/JS.

How It Works: User input is processed and passed into a RAG chain. FAISS retrieves relevant chunks from a vector store, which are fed along with the user prompt into a language model for answer generation.

Usage: Ask a question related to the knowledge base. aiRag will combine real-time understanding with document context to generate a precise response.

Limitations: Subject to the quality of retrieved documents. Performance depends on chunking and embedding granularity. No tool use or code execution.

Recommendations: Enhance retrieval quality via embedding optimization, hybrid search (BM25 + vectors), and multi-hop reasoning. Can be extended with chat memory and feedback loops.