Evidence-Backed Lead Generation on a Zero-Cost Stack
The architecture behind Neurix Intelligence — a queue between a cloud API and a local worker, a private search engine, and proof stored with every lead.
- Published
- Nov 2026
- Reading
- 5 min
- Topics
- architecture, job queue, supabase, data pipelines
The constraint
Searching a country's worth of businesses means reading a very large public dataset — Overture Maps — and that needs about 1 GB of memory. Free cloud tiers offer around half of that, and search engines block cloud IPs far faster than a home connection.
Split by job, not by layer
The web app (Vercel) and API (Render) only display data and enqueue work. A worker on a local machine claims jobs from a queue in Supabase, does the heavy discovery and verification, and writes results straight back. If no worker is online, jobs simply wait — the site keeps working.
Proof, not opinions
Each lead stores why it qualifies: the website check result, search-engine findings from a private SearXNG instance, and the public contact channels found. Regional outreach rules are flagged per lead, so the team knows what's allowed before reaching out.
Takeaway
A queue in the database is often all the infrastructure a small team needs. It decouples a fast, cheap front end from slow, resource-hungry work and makes runs resumable for free.