Context Switch Radar

A personal productivity observability tool that detects and visualizes hidden context switches across apps, tabs, and tasks.

Why this is portfolio-worthy

• Not another CRUD app: it models attention as a graph.
• Demonstrates event pipelines, feature engineering, and UX storytelling.
• Can evolve into a real product for developers/designers.

MVP implemented in this repo

• FastAPI backend with local SQLite storage
• Event ingestion API (POST /events)
• Event list API (GET /events)
• Fragmentation score API (GET /metrics/fragmentation)
• Rule-based score using switch ratio + rapid-switch ratio

Core idea

Collect local activity events (window title changes, tab focus, IDE/git actions) and compute a Context Fragmentation Score by hour/day.

Tech stack

• Backend: Python + FastAPI
• Storage: SQLite
• Frontend: placeholder (src/frontend) for upcoming dashboard

Quick start

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
uvicorn src.backend.main:app --reload

Health check:

curl http://127.0.0.1:8000/health

Create event:

curl -X POST http://127.0.0.1:8000/events \
  -H 'Content-Type: application/json' \
  -d '{
    "source":"vscode",
    "title":"Fix login bug",
    "task_key":"bugfix-login",
    "occurred_at":"2026-02-26T00:10:00+00:00"
  }'

Get score:

curl 'http://127.0.0.1:8000/metrics/fragmentation?hours=24'

Project structure

• src/backend: API + scoring logic
• src/frontend: dashboard placeholder
• docs: architecture docs
• tests: API and metric tests

Roadmap

1. Add activity collectors (VS Code window events, browser tab events)
2. Add hourly heatmap endpoint for dashboard
3. Add sequence-model baseline and compare with rule-based score
4. Add privacy controls and redaction rules for window title data