Smart Documentation Chatbot

An easy-to-use framework that turns your project documentation into a chatbot you can ask anything.

Introduction

Free TypeScript template for a Smart Documentation Chatbot.

This project indexes your documentation and lets you query it through a chatbot.

It replicates how the Intlayer documentation works.

How it works:

1. We use OpenAI embeddings and store them in a JSON file.
2. We then use vector similarity search to find the documents most relevant to the user query.
3. We return the most relevant documents to the user.

Architecture

The project is divided into three main components:

1. docs
   • We index the docs
   • We store embeddings in a JSON file
2. website (Next.js)
   • Renders and interprets the docs using intlayer and Next.js route handlers
   • Uses OpenAI vector similarity search to find the most relevant documents
   • Returns the most relevant results to the user

How it works

1. List and index all available documentation
2. Generate an embedding for each document using OpenAI
3. Store embeddings in a JSON file
4. When a user asks a question or searches for a doc:
   • Generate a new embedding for the user's query
   • Use vector similarity search to find the most relevant documents
   • Retrieve the top matching documents
5. Return the most relevant documents to the user

Demo

Check out a demo for my Intlayer project here.

Installation

Clone the Repository

git clone https://github.com/aymericzip/smart_doc.git
cd smart_doc

Install Dependencies

npm install -g pnpm
pnpm install

Configure the .env files

You can find two .env template files in the project:

• website/.env.template
• docs/.env.template

Rename these files to .env and fill the variables with your own values.

Build the project

pnpm run build

Start the project

cd website
pnpm run start