A small RAG backend needs a low-cost place to run

A small system was built for learning and a personal portfolio. It uses caching to avoid repeating work and to control how many requests can arrive in a short period.

Supabase provides the , while Groq generates the answers. Groq was chosen because the Gemini service kept returning errors for an unknown reason.

The backend was packaged as a for , but its setup screen appeared to require a paid upgrade for that method. Other options, including Gradio, are still being considered for hosting the personal experiment.

Key points

  • The RAG system is a small learning and portfolio project.
  • Caching and are included to control repeated work and excessive requests.
  • Supabase stores the searchable data, and Groq generates answers.
  • Gemini could not be used because its service kept returning errors.
  • appeared to require payment for the , so alternatives such as Gradio are being considered.
Read original