kany187/RAG-Translation-AI

RAG Translator AI is an intelligent Mbochi-French translation API powered by Retrieval-Augmented Generation (RAG). RAG Translator AI is an intelligent Mbochi-French translation API powered by Retrieval-Augmented Generation (RAG). It combines a vector database (Pinecone) with OpenAI's language models to provide accurate translations from a Mbochi-French dictionary. Unlike a simple lookup tool, it understands natural language questions and retrieves the most relevant dictionary content to generate clear, concise answers. Key features: Conversational API — users can ask multiple questions in a session with memory Intelligent routing — filters off-topic questions automatically RAG pipeline — retrieves relevant dictionary chunks before generating answers Answer refinement — improves responses for clarity and conciseness Tech stack: FastAPI — REST API LangGraph — conversational agent with stateful workflow LangChain — document loading, chunking, embeddings Pinecone — vector database for dictionary storage OpenAI — embeddings (text-embedding-3-large) and LLM (gpt-4o) Unstructured — PDF parsing with French language support Use case: Preserving and making accessible the Mbochi language — a minority language from the Republic of Congo — by providing an AI-powered translation assistant grounded in an authoritative dictionar