# 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