alpha8eta / AlphaBeta-GPT

A chatbot trained to act like Amjad Masad, built with LangChain and Next.js.

Amjad GPT

A chatbot trained to act like Adam Breckler, built with LangChain and Next.js.

This is only the frontend part of this project; The OpenAI model with Langchain is hosted in another Repl.

The OpenAI model being used is text-davinci-003, trained with:

• Major parts of the Replit docs
• The Replit blog
• The Replit landing page
• The Replit Employee Organization Chart
• Amjad's personal blog
• Amjad's AmA Repl
• Amjad's Podcasts
• Some of Amjad's Tweets
• Replit's Terms of Service and Community guidelines
• Some random facts the AI should be aware of

All of this data was compiled into a .index file. Zahid Khawaja has an Awesome Tutorial on how to do this if you want to make your own.

Speaking of which, Thanks a lot Zahid, I couldn't have done this without you 🙏

Source Code

Quota & Limits

By default, all users get a total of 10 responses. After you've used up the ten responses, you must add your own OpenAI API key (Open the settings tab) or tip this Repl to increase your quota.

One response is one question asked followed by one answer from the chatbot. Your quota will not be measured in tokens or response length.

• 100 Cycles tip 🍬 - Quota increased by 20 responses
• 500 Cycles tip 🍕 - Quota increased by 100 responses
• 1000 Cycles tip 🌯 - Quota increased by 200 responses

Self-Hosting Instructions

1. Set up the following environment variables:
   • DEFALT_QUOTA_LIMIT - number (set to ten in this repl)
   • MONGO_URI - a MongoDB database URI
   • OPENAI_API_KEY - An OpenAI API key
2. Run yarn setup && yarn dev or yarn build && yarn start
3. Done!

Retrain with new training dataset

1. Change training dataset
2. Run yarn retrain
3. Commit updated files ./vectorStore/*

---

Made with 🧠 & 🔥 by @IroncladDev