Back
2025 Solo 3 weeks

AgentForge

AI No-Code Chatbot Builder — A multi-tenant SaaS platform for creating and deploying AI-powered chatbots using RAG.

ReactTypeScriptSupabasepgvector

// the problem

Why build this?

Businesses need custom AI chatbots trained on their own data, but building RAG systems requires significant engineering expertise and infrastructure knowledge.

// the solution

How I solved it

Created a no-code platform where users upload documents, which are automatically chunked, embedded, and stored in pgvector for semantic search. The chatbot uses RAG to provide accurate, context-aware responses from the uploaded knowledge base.

// architecture

System Design

React + TypeScript frontend
Supabase backend with PostgreSQL + pgvector
Document ingestion pipeline
Embedding generation
Semantic search
LLM response generation

// features

Key Features

1

No-code chatbot creation with document upload

2

RAG pipeline with chunking, embeddings, and pgvector semantic search

3

Embeddable chat widget with theme customization

4

Analytics dashboard tracking conversations and satisfaction

5

Automated fallback to human agents for unresolved queries

6

Multi-tenant architecture with isolated data per user

// challenges & learnings

What I Learned

Optimizing document chunking strategy for different file types to maintain context

Implementing efficient pgvector similarity search at scale

Building a real-time embeddable widget that works across different host websites

Interested in this project?

Check out the code on GitHub or try the live demo.