Job Tracker

A dashboard for running a job search without drowning in tabs. Applications move through eight status stages with priority and ~30 fields of metadata per role — who you talked to, tech stack, culture notes, red flags.

How it works

• Paste a job URL and a Supabase edge function scrapes it with Firecrawl, then force-calls a structured extraction tool — the model must return typed JSON matching a strict schema, never prose.
• Job boards (Workable, Greenhouse, Lever, Ashby) are client-rendered SPAs, so the scraper waits three seconds for hydration; without the wait you scrape an empty React shell. LinkedIn blocks scraping entirely — the function detects that and asks you to paste manually instead of failing silently.
• Upload a CV or cover letter and it's parsed client-side (pdfjs for PDF, mammoth for DOCX), then scored 0–100 against the job description, with strengths, gaps and missing keywords.
• All AI runs in Deno edge functions behind a configurable OpenAI-compatible endpoint, so keys never reach the browser and the model can be swapped by env var.

Status

Deployed and in maintenance — the commit log is bug fixes, not features, which is what shipped looks like.