comit

Commit to the people who matter. A privacy-first work-life-balance companion for WhatsApp that turns a chat export into interaction scores, reply-debt tracking, and gentle nudges — all computed locally, with zero data leaving your machine.

comit reads a WhatsApp Export Chat file and answers the questions a busy person actually has:

• Who am I drifting away from that I used to talk to every week?
• Who am I leaving on read — and for how long?
• Is work bleeding into my evenings (and who keeps pinging me at 11pm)?
• Which relationships are thriving, and which are quietly fading?

It's named for the pun it lives by: a small daily commit to the relationships you'd otherwise let drift, and to your own off-hours.

---

Table of contents

• Why
• Features
• Quickstart
• See it in action
• The dashboard
• How it works
• Use your own chats
• Privacy
• Project structure
• Roadmap
• Contributing
• License

---

Why

Messaging apps are great at telling you about unread messages. They're terrible at telling you about unreplied ones, about the friend you haven't spoken to in two months, or about the fact that 64% of your work chatter now lands after you've clocked off.

comit is a tiny, transparent tool that treats your communication like something worth being intentional about — without uploading a single message anywhere. It's built to be read, audited, and forked: one pure analytics engine, a friendly CLI, and a local dashboard.

Features

🎯 Interaction Score	A transparent 0–100 score per contact, blending recency, frequency, reciprocity, your responsiveness, and who-starts-conversations. Graded thriving → steady → cooling → fading.
↩️ Reply Debt	Detects threads where the ball is in your court, and how long it's been there — "you owe Amma a reply, waiting 3 days."
⚖️ Work-Life Balance	Work vs personal split, after-hours work detection, busiest hour, weekday rhythm, and your most frequent late-night pingers.
📈 Momentum	Per-contact trend (recent half vs prior half of the window) — who's heating up, who's going quiet.
✨ Nudges	A ranked, plain-language to-do list that combines all of the above: reply-debts, drifting friends, boundary suggestions, slow-reply warnings.
🔒 Private by design	100% local. No accounts, no servers, no telemetry, zero runtime dependencies, no external network calls — even the dashboard fonts and charts are self-hosted.

Quickstart

You need Bun (≥ 1.1). No install step is required to run — Bun executes the TypeScript directly.

git clone https://github.com/akhil29897/comit.git
cd comit

# Run the full report against bundled, synthetic demo data:
bun run demo

# Or open the visual dashboard at http://localhost:4317
bun run demo:web

The repo ships with synthetic demo conversations (generated by scripts/generate-fixtures.ts) so everything works out of the box. No real chat data is included, ever.

See it in action

bun run demo prints a full report to your terminal:

  comit  ·  commit to the people who matter
  a privacy-first WhatsApp work-life-balance companion

  6 conversations · 321 messages · 90-day window · as of 2026-06-19

── nudges  (what to do) ──────────────────────────────────────
  ⚠  Vinay (work) reached you after hours 44× recently. Consider muting or setting expectations.
  ↩  You owe Amma a reply — waiting 3 days. Last: "your cousin is asking about you"
  ≀  Amma is cooling off: ~10.89/week → ~2.02/week. Reach out?
  ⚠  Work chatter with Vinay is climbing (~1.4/week → ~6.22/week). Protect your off-hours.
  ⏱  Your replies to Karan take ~15h on average. If they matter, tighten that up.

── interaction scores ────────────────────────────────────────
contact      score                grade       last  sent/recv  type
Riya         83.7 ████████████░░  thriving   today      38/36  personal
Vinay        78.8 ███████████░░░  thriving   today      25/24  work
Sourav         71 ██████████░░░░  thriving  2d ago      29/26  work
Amma         69.9 ██████████░░░░  steady    4d ago      44/39  personal
Karan        58.3 ████████░░░░░░  steady    2d ago      19/20  personal

── reply debt  (you owe a reply) ─────────────────────────────
contact   waiting  type      their last message
Amma       3 days  personal  "your cousin is asking about you"
Vinay    13 hours  work      "quick call?"

── work-life balance ─────────────────────────────────────────
  work vs personal   ██████░░░░░░░░░░░░ 35% work
  after-hours work   66 messages (64% of work chatter, outside working hours)
  busiest hour       11 PM
  late pingers       Vinay (44), Sourav (22)

Individual views are available too: bun run comit scores <path>, debts, nudges, balance, and --json for piping into anything.

The dashboard

bun run web serves a local, offline dashboard (no CDNs, no external requests):

Score rings per contact, a work/personal donut, after-hours stats, a weekday rhythm, and the same ranked nudges — all rendered from the exact same pure engine the CLI uses.

On your phone

comit is designed as a WhatsApp-style mobile companion. A clickable iPhone 15 prototype lives in prototype/ — open it in any browser (it's a self-contained, zero-dependency page):

Four screens — Launch, Today (your nudges as chat bubbles), People (relationships as a chat-style list with scores), and Balance (work-life breakdown) — with the familiar green header, chat-wallpaper background, and bottom tab bar.

How it works

WhatsApp .txt export
        │
        ▼
┌──────────────────┐   parse + normalize    ┌─────────────────────────┐
│  Data source     │ ─────────────────────▶ │  Pure analytics engine  │
│  (export parser) │                        │  score · debt · balance │
└──────────────────┘                        │  · trends · nudges      │
        ▲                                    └─────────────────────────┘
        │                                              │
   (future: live                          ┌────────────┴───────────┐
    Baileys adapter —                     ▼                        ▼
    same interface)                    CLI report            Web dashboard

The heart of comit is a pure, deterministic engine in src/core: data in, plain objects out, no I/O and no clock reads. The only thing that touches the world is a small DataSource adapter. Today there's one — the WhatsApp export parser — but the seam is deliberate: a real-time source could be added without changing a line of analytics.

Read more: Architecture · Scoring model · Usage & config

Use your own chats

1. In WhatsApp, open a chat → ⋮ / contact name → Export Chat → Without Media. You'll get a .txt file (the parser handles both iOS and Android formats automatically).
2. Drop one or more exports into a folder, e.g. ./exports/.
3. Run it:

bun run comit report ./exports --me "Your Name" --config comit.config.json

A config file lets you tag contacts as work/personal, set your working hours, and choose the analysis window. See fixtures/comit.config.json and the Usage guide.

Your real exports are git-ignored by default (exports/, *.whatsapp.txt, data/). comit never writes them anywhere but your screen.

Privacy

This is the whole point, so it gets its own page: docs/PRIVACY.md.

The short version:

• Everything runs on your machine. There is no backend.
• comit makes no network requests — not for analytics, not for fonts, not for charts.
• It only reads the files you point it at; it never writes your messages anywhere.
• Zero runtime dependencies means a tiny, auditable supply chain.

Project structure

comit/
├── src/
│   ├── core/                 # pure engine (no I/O) — the testable heart
│   │   ├── analytics/        # interactionScore · replyDebt · balance · trends · nudges
│   │   ├── sources/          # DataSource interface + WhatsApp export parser
│   │   ├── pipeline.ts       # normalize → score → debt → balance → trends → nudges
│   │   └── types.ts          # the shared domain model
│   ├── cli/                  # zero-dep terminal UI
│   └── web/                  # Bun server + offline dashboard
├── test/                     # bun:test suite (parser, analytics, pipeline)
├── fixtures/                 # synthetic demo exports + sample config
├── scripts/                  # deterministic fixture generator
├── docs/                     # architecture, scoring, privacy, usage, roadmap
├── prototype/                # WhatsApp-style iPhone 15 prototype (zero-dep)
└── presentation/             # reveal.js deck (+ exported .pptx)

Roadmap

See docs/ROADMAP.md. Highlights:

• A live (opt-in, clearly-flagged) data source via Baileys, behind the existing DataSource interface.
• Importers for other exports (Telegram, Signal, iMessage).
• Configurable scoring weights and nudge rules.
• A "weekly digest" mode.

Contributing

PRs and ideas welcome — this is built to be forked. Start with CONTRIBUTING.md and our Code of Conduct. Good first issues: new importers, additional nudge rules, and i18n for date formats.

bun test          # run the suite
bun run typecheck # strict TypeScript, no emit

License

MIT © Akhil Tripathi. Use it, fork it, learn from it.

_{Built with care, and with Claude Code as a pair.}