GPT for Zen and the Art of Motorcycle Maintenance

Zen and the Art of Motorcycle Maintenance by Robert M. Pirsig is my favorite book in 2022.

With this app, you are able to ask Robert questions and search book content with natural language.

Dataset

Zen and the Art of Motorcycle Maintenance

Implementation

Zen and the Art of Motorcycle Maintenance is scraped and written as a json file ./scripts/zen_motor.json.

The book was split into chunks and each chunk was embedded with OpenAI Embeddings (text-embedding-ada-002), refer to ./scripts/embed.ts, all the embeddings are stored in Supabase database.

When user input a question:

1. Seach mode: the embedding of each chunk would be comparaed with the embedding of the question, and the chunks with close cosine similarity score will be returned.
2. Chat mode: the returned chinks will be used as prompt and fed to OpenAI's API and get an answer.

Running Locally

Here's a quick overview of how to run it locally.

Requirements

1. Set up OpenAI

You'll need an OpenAI API key to generate embeddings.

2. Set up Supabase and create a database

Note: there are also other embedding databases

Check schema.sql for setting up database on Supabase.

Repo Setup

3. Clone repo

git clone https://github.com/shao-shuai/zen_motorcycle.git

4. Install dependencies

npm i

5. Set up environment variables

Create a .env.local file in the root of the repo with the following variables:

OPENAI_API_KEY=

NEXT_PUBLIC_SUPABASE_URL=
SUPABASE_SERVICE_ROLE_KEY=

Thanks

This repo was inspired by paul-graham-gpt created by Mckay Wrigley, he also has an awesome video here, How to create an OpenAI Q&A bot with ChatGPT API + embeddings.