Discovery
Finds fresh job postings matching Pontoon's targeting rules, ready to hand off for contact enrichment.
Ready to run
Set your parameters below and click Start.
0 of 60 found
0%
Start a run to see live activity here...
Guide
How to use PARA, where your data lives, and what to do when something looks off.
1. What PARA actually does right now
PARA searches job boards for fresh postings matching Pontoon's real targeting rules (company size, industry, freshness) and saves the qualifying ones for review.
Know exactly what stage this is at today:
- ✓ Discovery works — PARA finds real job postings matching your rules
- ✓ Data is saved safely — every posting lands in the database the moment it's found
- ✗ Not yet connected to RAT — PARA does not yet find contacts or push finished leads into your RAT tool. That's the next phase.
2. Where everything lives
| Piece | What it is | Where it lives |
|---|
| The Dashboard | The page you open and click Start on | pontoonpara.pages.dev |
| The Backend | The engine that does the searching | Railway — pontoon-para-backend-v2 |
| The Database | Where results and history are stored | Supabase — pontoon-ats (same project as RAT) |
You never need to touch the backend or database directly for normal use — the dashboard is the only thing you interact with day to day.
3. How to run a discovery search
- Fill in Industries — comma-separated (e.g., Construction, Manufacturing, Legal, Accounting & Finance)
- Days back — how far to search (max 7). An older posting that's still open is a good sign, not a bad one.
- Max employees — the hard cap on company size (default 500)
- Target leads — how many qualifying postings you want this run to find
- Check the job boards you want this run to use — core boards are checked by default
- Click Start discovery run and watch the live feed
4. Understanding the status colors
| Status | What it means |
|---|
| Idle | Nothing running, ready to start |
| Running | Actively searching right now |
| Healthy | Run finished, hit its target, zero errors |
| Minor issues | Finished with some errors, but found most of what you asked for |
| Needs attention | Something went wrong — check the red banner for the exact reason |
5. How to confirm data is actually being saved
Every [found] line in the live feed represents a real database write that already happened, not just a display update. Click any row in Recent runs to expand it and see the actual companies, positions, and confidence levels saved for that run, pulled directly from the database.
6. Where the data sits, technically
| Table | What's in it |
|---|
| para_seen_jobs | Every job URL ever processed — prevents duplicate results across runs/days |
| para_discovery_runs | Full history of every run: settings used, counts, final status |
| para_discovered_postings | The actual postings found — company, position, location, industry, confidence, notes |
None of this currently writes into crm_leads (RAT's table) — that connection is planned but not yet built.
7. Common situations and what they mean
"Needs attention" with a message about API usage limits — your Anthropic account hit a spending cap, not a bug. Check console.anthropic.com → Billing, raise the cap or wait for the reset date shown in the message.
Live feed looks stuck — give it 30-60 seconds, each search round takes real time. If truly stuck for minutes, check the browser Console (F12) for a red error.
"No job boards selected" — check at least one board before starting.
Fewer postings than target — often just means PARA ran out of new qualifying postings for that scope that day, not a malfunction. Check the run's error log first.
8. What's planned next (not built yet)
- Contact enrichment integrated into this same pipeline — finding 3-4 hiring manager/HR contacts per posting
- Writing finished leads into RAT's crm_leads table, tagged ai_pending, for your team to review
- Scaling up to real daily volume (800-1,000 leads/day) once accuracy is proven at smaller scale