Custom Chatbot Builder & Hosting Platform

Creators can upload documents for knowledge base, configure guardrails and tone, then deploy via API or an embeddable web widget. The system is multi‑tenant with isolated data for each user, session memory per conversation, and allows users to set system prompt and score threshold.

FastAPI React Qdrant Vector DB RAG AWS (EC2 · S3 · CloudFront · Cognito)
🔗 View Live Demo
File upload interface for chatbot knowledge base

Upload Files

Easily upload documents (Markdown, PDF, CSV) to build your chatbot's knowledge base

System prompt configuration interface

Set System Prompt

Configure your chatbot's personality, tone, and behavior with custom system prompts

Chatbot interface for asking questions

Ask Questions About Knowledge Base

Interact with your chatbot and get intelligent responses based on uploaded content

Debug interface for chunk scoring and threshold settings

Debug & Monitor

Check which knowledge chunks pass the relevance threshold and fine-tune performance

Upload Files
Set System Prompt
Ask Questions
Debug Interface

Architecture & Stack

  • Frontend: React Web App
  • Backend: FastAPI on EC2, Open AI API: Embedding & Chat
  • Data: Qdrant (vector DB), Redis (session memory)
  • Storage: AWS S3 for document storage / retrieval

Challenges

  • Deciding how to score chunk revelevancy, how many chunks to feed into chat context
  • Once a user deletes a document from a project, the embeddings from this doc are removed from the vector DB

What I Learned

  • RAG pipelines rely heavily on document parsing and storage
  • How to manage knowledge bases, allow user deletions / changes

Links