Project Description

🚀 Try it live: This portfolio itself uses an instance of the Chatbot AI! You can interact with it by clicking the chat bubble in the bottom right corner.

A production-ready, multi-tenant RAG (Retrieval-Augmented Generation) platform. It enables businesses to deploy intelligent chatbots trained on their proprietary knowledge base (PDFs, DOCX, Web pages). Each tenant benefits from an isolated vector database, a configurable AI persona, and a secure, embeddable widget.

The system leverages FastAPI for a high-performance asynchronous backend, React for both the admin dashboard and the chat widget, and a modern data architecture using Supabase and Pinecone.

Key Features

Advanced RAG Architecture

Document Processing: Automated pipeline using LlamaParse V2 for high-fidelity PDF and DOCX ingestion.
Smart Web Scraping: Scraping system with SSRF validation and content hashing to keep the knowledge base in sync efficiently.
Hybrid Search: Semantic retrieval with Pinecone and intelligent re-ranking based on query intent.

Multi-Tenancy & Security

Data Isolation: Dedicated Pinecone namespaces and Row-Level Security (RLS) in Supabase for every tenant.
Secure Widget: API Key authentication, domain allowlisting (CORS), and abuse protection.

Admin Dashboard

Comprehensive management of knowledge sources (file uploads, URL indexing).
Chatbot “Persona” configuration (tone, base instructions, custom rules).
Conversation analytics and security event logs.

Detailed Tech Stack

Backend: Python 3.12+, FastAPI, LangChain.
Frontend: React 19, TypeScript, TailwindCSS, Vite.
AI/ML: OpenAI GPT-4o-mini, text-embedding-3-small.
Infrastructure: Docker, Render.com, Cloudflare Pages.