AI Chat Platform
A SaaS platform for building custom AI assistants on top of a company's internal knowledge base. Its RAG (Retrieval-Augmented Generation) pipeline ingests internal documents, FAQs, product manuals, and more to deliver accurate answers.
Project Overview
A SaaS platform that lets companies easily build AI assistants powered by their internal knowledge. Without any complex AI infrastructure, teams can upload documents and deploy a custom chatbot in just a few clicks.
Key Features
RAG Pipeline
When a document is uploaded, the platform automatically extracts its text, splits it into chunks, and converts them into vector embeddings. When a user asks a question, it retrieves semantically similar document fragments and uses them as context.
const retrieveContext = async (query: string) => {
const embedding = await embedModel.embed(query);
const relevantChunks = await vectorStore.similaritySearch(embedding, 5);
return relevantChunks.map((chunk) => chunk.content).join("\n\n");
};
Real-Time Streaming Responses
AI responses are streamed in real time using SSE (Server-Sent Events). Users see results immediately instead of waiting for the full response to be generated.
Multi-Tenant Architecture
Each company's data is fully isolated. Row-Level Security (RLS) and per-tenant vector indexes guarantee both data security and search performance.
Tech Stack
- Frontend: Next.js 15, React 19, Tailwind CSS, Zustand
- Backend: Next.js API Routes, LangChain.js
- AI: Claude API (Anthropic), OpenAI Embeddings
- Database: PostgreSQL (Supabase), pgvector
- Caching: Redis (Upstash)
- Deployment: Vercel, Docker
Results
- Achieved an average response accuracy of 92%
- Handles 10,000+ queries per day on average
- Takes 5 minutes on average from document upload to chatbot deployment