.planning/STATE.md

Project State

Project

• Name: Job Automation Agent
• Type: Python/Streamlit job automation pipeline
• Status: Active — production-ready with 5 platforms

Current Architecture

• Platforms: LinkedIn, Indeed, Glassdoor, Remotive, WeWorkRemotely (+ Naukri disabled)
• AI Models: 10-model NVIDIA API pool for parallel assessment
• ATS: Hybrid scoring (70% JD keyword match + 30% resume quality)
• Output: DOCX + PDF resumes, Google Sheets, Excel reports
• Dedup: SQLite 30-day history + cross-platform (title, company) dedup

Key Decisions

• All scrapers extend BaseScraper from src/scrapers/base.py
• Job dataclass is the universal job format
• Platform cap is per-platform total, not per-query
• Resume generation runs in parallel with ThreadPoolExecutor