yanliu1111 / supabase-chatgpt-your-pdf

🌱 Tutorial Study: Delving into a supabase and ChatGPT-based "chatting your files" software, built on the MVC architecture and utilizing pgvector.

Workshop: pgvector to Prod in 2 hours

Create a production-ready MVP for securely chatting with your documents.

☑️ Learning from 👉 HERE

☑️ I kept my learning 📝notes in different branches, just follow the GR great tutorial and you will learn everything.

☑️ From this tutorial you will learn:

• Supabase
• Database migration
• The structure of how to use retrival augmented generation (RAG) to process your files, and query smaller meaningful document sections.
• You will learn pgvector, how to index your text, and create edge functions to process the markdown files into document_sections table.
• You will learn Embedding from GR ❤️, highly recommend his Embedding teaching.
• The interesting step is generating the ueser message embedding in the frontend. Then determine the similarity with markdown files embedding.

📑Notes files:

The first time following the learning using git branches, I like this way, time efficient and more focus on the knowledge and code logic. The code edits in the following Branches:

• Step-2
  • Readme
  • app/files/pages.tsx
  • supabase/functions/process/index.js
  • supbase/seed.sql
  • supabase/migrations/documents.sql
    
• Step-3
  • Readme
  • supabase/migrations/embed.sql
    
• Step-4
  • Readme
  • next.config.js
  • app/chat/pages.tsx
  • supabase/migrations/match.sql
  • supabase/functions/chat/index.tsx
    
• Step-5
  • Generate DB schema typescript type, add <Database> generic and type error check.