With this app, you allow users to train Chat-GPT3 on their own data. The bot will only answer using their documents, and if it can’t find the answer, it will simply say “I don’t know”. Users can upload a PDF file, or extract text from an image. if the document is too large, the app will split it into small chunks, fixing the token limit issue. Check a demo from here.

The app also comes with art generation feature, allowing users to choose aspect ratio and art style with Stable Diffusion.

A third party API is needed, since chatting with documents require other libraries such as LangChain, Stable Diffusion, and the library to extract text from image. but no worries, you’ll only need 2 keys, and they’re low cost

Screens:(11 total)

• On Boarding
• Login / Create Account
• Home Screen
• History
• Normal chat 
• Image generation
• Profile
• Subscriptions
• Document splitter
• Edit document content
• document chat

Features

• Train your documents on your own documents
• Split the large documents into small chunks
• Allow users to add their own OpenAI’s key
• Generate AI art with Stable Diffusion
• Google Sign In
• Push Notifications with firebase
• Admob Ads
• Delete Account
• Normal Chat (chat with normal GPT)
• Subscriptions
• Extract text from images