Actiwitee merges GitHub, competitive programming, and local AI-coding sessions into a single heatmap and read-only API, built for portfolios that reflect how you actually work.
config.yaml
sources:
github: [personal, work]
leetcode: [saish_korgaonkar]
codeforces: [saish_k]
agent:
signals: [cursor, hermes-cli]
$ actiwitee collect
github:personal 459 contrib
leetcode 15 contrib
codeforces 6 contrib
codechef 141 contrib
→ store.json updated
$ curl /contributions
{
"total": 622,
"contributions": [
{ "date": "2026-07-25",
"level": 2, "localMinutes": 76 }
]
}
$ actiwitee agent --once
[agent] active: ai-cli, terminal
→ 76 min local today
→ heartbeat stored
Hermes cron · every 5m
Actiwitee is the self-hostable backend behind unified coding activity, from open source commits to contest submissions to late-night AI pair sessions.
Pull from GitHub (multi-account), LeetCode, Codeforces, CodeChef, and more. Weighted merge with percentile-based heatmap levels.
A lightweight agent samples AI CLI, IDE, and terminal activity. Local minutes show up on the graph without inflating commit counts.
Serve a read-only JSON API from Cloudflare Workers + R2. Drop the heatmap into any site with one env var.
One YAML file for sources, weights, CORS, and agent signals. No database, just JSON on disk.
GitHub history back to account creation. Scrollable graph with rich per-day tooltips: contributions, local time, and source breakdown.
Collect on a schedule, publish to R2, and run the agent every few minutes. Set it once and forget it.
Add handles and tokens in config.yaml. Run collect. Actiwitee normalizes dates, deduplicates, and merges into one timeline.
GitHub
Multi-account, private repos via PAT
LeetCode
Submission calendar sync
Codeforces
Contest & practice activity
CodeChef
Practice streaks
Local agent
Cursor, VS Code, Claude Code, Hermes, terminal
Demo
Sample data for development
Three routes, no auth required. Host on Cloudflare Workers + R2 and call from any frontend with one env var.
Live example → codepulse.saish.xyz/healthLiveness check and per-source freshness timestamps.
{ "ok": true, "sources": { "github:personal": "2026-07-25T..." } }/contributions?year=2026Lean heatmap rows, GitHub-compatible {date, count, level} shape.
{ "total": 622, "contributions": [{ "date": "2026-07-25", "count": 3, "level": 2 }] }/activityFull payload with scores, local minutes, and per-source breakdown per day.
{ "contributions": [{ "date": "...", "count": 3, "level": 2, "localMinutes": 76, "breakdown": { "github": 2, "local": 76 } }] }From zero to a live portfolio heatmap. Copy the commands, edit config.yaml for your handles, deploy the Worker, and point your site at the API.
Clone the monorepo, install dependencies, and run with synthetic data. No tokens or network required.
terminal
git clone https://github.com/SaishKorgaonkar/actiwitee.git
cd actiwitee
npm install
npm run build
cd cli
node dist/cli.js collect --demo
node dist/cli.js serve # http://localhost:4787verify
curl localhost:4787/health
curl localhost:4787/activity | headScaffold config files, then add handles and tokens. Secrets stay in .env, never in config.yaml.
GitHub PATs need read:user scope for private contribution counts. Without a token, only public activity is fetched.
terminal
node dist/cli.js init
# edits config.yaml + .envconfig.yaml (excerpt)
sources:
github:
- id: personal
username: your-handle
tokenEnv: ACTIWITEE_GITHUB_PERSONAL_TOKEN
includePrivate: true
leetcode: [{ username: your-lc-handle }]
codeforces: [{ username: your-cf-handle }]
weights: { github: 1, leetcode: 3, codeforces: 3, local: 1 }
levels: { mode: percentile }.env
ACTIWITEE_GITHUB_PERSONAL_TOKEN=ghp_...
ACTIWITEE_GITHUB_WORK_TOKEN=ghp_...The agent samples process, app-focus, and session-log signals on an interval. Local minutes appear on the heatmap separately from commit counts.
macOS only for app-focus signals. Add a signal for every tool you use: Cursor, VS Code, Claude Code, Hermes, terminals, local models, etc.
config.yaml (agent)
agent:
interval: 180
signals:
- { id: cursor, type: app-focus, match: "Cursor", category: editor }
- { id: claude-code, type: session-log, path: "~/.claude/projects", category: ai-cli }
- { id: ollama, type: process, match: "ollama|llama-server", category: local-model }terminal
node dist/cli.js agent # continuous
node dist/cli.js agent --once # single samplePublish merged activity.json to Cloudflare R2 and serve it from a Worker. Your portfolio reads the JSON, no backend on your site.
Routes: GET /health, GET /contributions, GET /activity. CORS is configured in config.yaml for your portfolio origin.
publish locally
node dist/cli.js collect
node dist/cli.js show > data/activity.json
# upload to R2 (see cli/deploy/worker/README.md)deploy worker
cd cli/deploy/worker
npm install
cp wrangler.toml.example wrangler.toml
npx wrangler login
npm run deploy
# attach custom domain, e.g. activity.yourdomain.comPoint any frontend at the API. The contributions shape matches GitHub-style heatmap components.
Use /contributions for lean {date,count,level} rows. Use /activity for per-source breakdowns and local minutes in tooltips.
.env (portfolio)
NEXT_PUBLIC_ACTIWITEE_API_URL=https://activity.yourdomain.comfetch
const res = await fetch(`${API_URL}/activity`)
const { contributions } = await res.json()
// [{ date, count, level, localMinutes?, breakdown? }]Install cron jobs with one script. Agent runs every 5 minutes; collect + R2 upload runs hourly. Your Mac must be awake for jobs to fire.
Requires npm run build first. Uses wrangler for R2 upload after Worker deploy (step 4). launchd or Hermes work too if you prefer those over cron.
install cron (macOS / Linux)
cd cli
# optional: set R2 target before install
export ACTIWITEE_R2_BUCKET=your-bucket
export ACTIWITEE_R2_KEY=actiwitee/activity.json
bash scripts/install-cron.sh
crontab -l | grep actiwiteewhat gets installed
# every 5 min — local heartbeats
scripts/actiwitee-agent.sh
# every hour — collect, merge, upload to R2
scripts/actiwitee-publish.sh
# logs
/tmp/actiwitee-agent.log
/tmp/actiwitee-publish.log
# remove
bash scripts/install-cron.sh --uninstallOpen source and self-hostable. Star the repo, follow the guide above, and wire the heatmap into your portfolio in an afternoon.