Backend, infrastructure, security, and deployment for NIV AI - Heet

[advikdivekar]

Own everything that makes the system stable, secure, and accessible via a live production URL. This includes authentication, APIs, data persistence, cost logic, report generation, and deployment. The goal is to ensure the frontend and AI layers run smoothly on top of a reliable backend that judges can access without failure.

1. regenerate firebase service account key and remove any previously exposed keys
2. add credentials to .gitignore and ensure they are never committed
3. audit repository for any leaked secrets or sensitive configs
4. enforce SKIP_AUTH=false for production environment
5. implement safe development bypass inside auth middleware without affecting production flow
6. remove /ws-test or any unsecured testing routes from backend
7. verify firebase authentication works end-to-end with real tokens
8. build docker image for backend service
9. push image to container registry
10. deploy backend to cloud run (region: asia-south1)
11. configure environment variables for production (SKIP_AUTH, USE_OLLAMA, ENVIRONMENT)
12. move sensitive values (gemini api key, firebase credentials) to secret manager
13. connect cloud run service with secrets securely
14. test all endpoints on deployed live url
15. configure CORS properly
16. allow all origins in development
17. restrict to frontend domain in production using environment variable
18. verify requests from frontend work without CORS issues
19. create /india/defaults endpoint
20. return repo rate, gst rules, stamp duty rates, and last updated timestamp
21. connect endpoint to constants defined in backend
22. ensure frontend and backend calculations stay consistent
23. implement PMAY subsidy calculation logic based on income brackets
24. compute subsidy impact (NPV or effective reduction)
25. include eligibility flag and subsidy value in response
26. integrate subsidy into affordability and cost calculations
27. extend user input schema to include annual_income
28. auto-derive from monthly income when not provided
29. update all dependent calculations and validations
30. persist analysis outputs (presentation, verdict) to firestore after processing
31. ensure session data survives server restarts
32. create helper function to save structured results
33. update report generation flow to read from firestore instead of in-memory state
34. ensure reports can be generated even after deployment restarts
35. add direct PDF download endpoint returning raw file bytes
36. set correct headers for file download in browser
37. ensure frontend can trigger download instantly
38. test PDF generation flow thoroughly across multiple sessions
39. ensure no dependency on temporary or in-memory state
40. write complete backend README
41. include setup steps, environment variables, API documentation, and deployment instructions
42. add architecture overview and link to live deployment
43. test full backend integration with frontend
44. validate /analyze, websocket roundtable, /report, and auth flows
45. fix any response mismatches or errors
46. ensure backend handles errors gracefully (invalid input, failed calls, missing data)
47. return clean structured error responses
48. verify system stability under repeated usage
49. ensure no crashes, memory issues, or broken sessions during demo
50. finalize backend readiness for demo video and judging (stable, secure, deployed, fully functional)