Skip to content

memory.md

Latest commit

 

History

History
47 lines (34 loc) · 2.08 KB

memory.md

File metadata and controls

47 lines (34 loc) · 2.08 KB

GeoDocs Memory

Key Lesson: Karnataka Site IP Blocking

The Karnataka land records site (landrecords.karnataka.gov.in) blocks all cloud datacenter IPs.

  • Railway, Vercel, AWS, Cloudflare Workers (as origin) — all blocked with HTTP 522/504
  • Regular user devices (Indian ISPs, mobile networks) — NOT blocked

The correct approach: run PDF extraction client-side in the user's browser.

  • Browser makes ASP.NET form POSTs through the Cloudflare Worker (for CORS headers only)
  • CF Worker: https://geodocs-proxy.harshag954.workers.dev
  • Client-side extractor: lib/clientPdfExtractor.ts
  • Entry point: lib/api.ts → fetchPdfUrl() tries client-side first, falls back to server API

Do NOT try to reach the Karnataka site from any server/cloud function. It will always fail.

Architecture

User Browser
  └─> lib/api.ts → fetchPdfUrl()
        └─> lib/clientPdfExtractor.ts (client-side, primary)
              └─> fetch() to geodocs-proxy.harshag954.workers.dev (CF Worker, CORS only)
                    └─> landrecords.karnataka.gov.in (Karnataka site)
        └─> /api/get-pdf-url (server fallback, will fail for cloud-blocked IPs)

PDF Extraction Flow (ASP.NET form POST sequence)

  1. GET /service3/ → grab hidden form fields (__VIEWSTATE etc.)
  2. POST district selection (__EVENTTARGET: ddl_district)
  3. POST taluk selection (__EVENTTARGET: ddl_taluk)
  4. POST hobli selection (__EVENTTARGET: ddl_hobli)
  5. POST village search (btnSearch: Search)
  6. Parse window.open('FileDownload.aspx?file=...') from response HTML
  7. Return full https://landrecords.karnataka.gov.in/service3/FileDownload.aspx?file=... URL

Infrastructure

  • Railway: https://geodocs-production.up.railway.app — hosts Next.js static export + Express API server
  • Cloudflare Worker: https://geodocs-proxy.harshag954.workers.dev — CORS proxy only, deployed via cf-proxy/
  • Railway env vars: CF_PROXY_URL, NEXT_PUBLIC_CF_PROXY_URL (both set)
  • No PDF_BACKEND_URL or PROXY_URL — both were removed (dead ngrok, broken Bright Data)