Leads-scrapers (website leads)

Same Lead desk as Leads-scraper-, retargeted at website leads instead of digital marketing leads.

A GitHub Action (scripts/scrape.py) uses the official Freelancer.com Python SDK to find people and businesses that posted a project on Freelancer.com asking for a website. Only buyers with a real business name (written in the project text, taken from a website link in it, or from their profile) and located in India are kept; quality matters more than quantity, the daily run keeps searching until it has 50 such leads (or runs out of matching projects). Projects that look like ongoing hiring, agency outsourcing, or non-website work (logo, video, SEO, apps) are skipped, and agencies are removed by name. Each lead links to its Freelancer project post. Freelancer does not give out phone numbers, so you add each phone number by hand on the dashboard. Voice calling, call log, hot/warm/cold grading and auto-call work exactly as in the original.

Optional repo secret FREELANCER_OAUTH_TOKEN (from developer.freelancer.com). Without it the scraper uses the public no-login search endpoint; add the token only if Freelancer starts refusing anonymous requests (tokens expire after 30 days). Optional env vars on the workflow step: FL_COUNTRIES (default in, comma separated codes, empty = worldwide), FL_MAX_AGE_DAYS (default 14), FL_PAGES (default 3).

What differs from the digital marketing version:

• Default Freelancer.com search phrases in lib/core.js (DEFAULT_SETTINGS.terms), editable later in the dashboard Settings tab
• Call variables sent to Rasen/Vapi: service_needed is "Website design and development", job_title falls back to "a web role"
• Call grading prompt describes a website design and development service

Setup is the same as the original: see .env.example, add the Vercel env vars, and set the repo secrets APP_URL and INGEST_SECRET for the workflow in .github/workflows/daily-leads.yml.