About

’m building an MVP for a legal-tech product focused on organizing user-uploaded documents and generating AI-powered summaries and pattern analysis. The project is built in Lovable and uses: React + Vite (Lovable scaffold) Supabase (Auth, DB, Storage, Edge Functions) OpenAI Responses API At the moment, my document analysis pipeline does not work reliably, especially for PDFs, Word docs, images, and text files. OpenAI returns either: “Document appears corrupted” or a generic 400 error or no readable content This means OpenAI is not receiving the document content correctly — usually due to URL issues, incorrect request shape, or Lovable generating conflicting code. I need an engineer to fix this end-to-end.