Richmond-24/African-Business-Intelligence-AI

Enterprise-grade RAG Agent built with Next.js, FastAPI, LangGraph, PostgreSQL, Qdrant, Docker, and Google Gemini. # 🚀 Africa Business AI Production RAG Agent ### Enterprise-Grade Retrieval-Augmented Generation (RAG) Platform Build AI-powered knowledge assistants capable of ingesting, indexing, retrieving, and reasoning over your documents using **Google Gemini**, **LangGraph**, **FastAPI**, **Next.js**, **PostgreSQL**, and **Qdrant**. --- # 📖 Overview Production RAG Agent is a full-stack enterprise-ready Retrieval-Augmented Generation (RAG) system designed to build intelligent AI assistants over private documents. Instead of relying only on an LLM's internal knowledge, the application retrieves relevant information from your uploaded documents using semantic search, injects that context into the prompt, and generates grounded responses with source citations. The project emphasizes production engineering practices including resilient document ingestion, deterministic chunking, duplicate detection, retry handling, crash recovery, and scalable vector search. --- # ⭐ Highlights - Enterprise-grade RAG Architecture - Multi-format Document Ingestion - Google Gemini Integration - LangGraph-based AI Pipeline - FastAPI Backend - Next.js + TypeScript Frontend - PostgreSQL Metadata Storage - Qdrant Vector Database - Adaptive Semantic Chunking - Duplicate Document Detection - Resume-safe Indexing - Streaming AI Chat - Source Citations - Docker Deployment - Production Health Monitoring --- # ✨ Features ## 📂 Document Processing Supports uploading and indexing: - PDF - DOCX - PPTX - XLSX - CSV - TXT - Markdown Each uploaded document is automatically: - Validated - Parsed - Chunked - Embedded - Indexed - Stored --- ## 🧠 Adaptive Semantic Chunking Production-friendly chunking pipeline featuring: - Recursive chunk splitting - Configurable chunk size - Chunk overlap - Deterministic Chunk IDs - Duplicate detection - Resume interrupted indexing --- ## 🤖 Retrieval-Augmented Generation Instead of sending the whole document to the LLM: 1. User asks a question. 2. …