Door Knocking Notes

A privacy-first, fully on-device web app for door knockers to record conversations and transcribe them locally — then share the day's notes as one message. Nothing is ever sent to the cloud by the app itself.

Features

🎙️ Record & Transcribe

• Record audio of doorstep conversations on any smartphone (iOS Safari, Android Chrome)
• Transcribe speech to text entirely on-device using Whisper Base (~80 MB, downloads once)
• Works offline after the first model download

🗂️ Session Blocks

• The first recording creates a block; every start/stop adds a session to it
• Each session is transcribed individually
• Tap Finish block & share to see a nicely formatted list of every transcribed conversation with its date and time

📤 Share & Wipe

• Share to any app via the native share sheet (WhatsApp, Notes, email, …) — or copy to the clipboard
• After a successful share, the block and all of its sessions are deleted from the device

🛡️ Privacy & Security

• ✅ 100% on-device processing — no audio or transcripts ever leave your device via the app
• ✅ Automatic PII redaction — names, addresses, phone numbers, postcodes are stripped before anything is stored
• ✅ No audio retention — audio is discarded in memory after transcription
• ✅ No backend, no accounts, no analytics

📱 Progressive Web App (PWA)

• Install to home screen like a native app
• Works offline after initial setup
• No app store friction

Tech Stack

Layer	Technology	Notes
Frontend	TypeScript + Vite	Fast dev experience, tiny bundle
Recording	MediaRecorder API + getUserMedia	Cross-browser audio capture
Transcription	whisper-web / @xenova/transformers	Whisper Base, WASM + WebGPU
Storage	IndexedDB (idb)	On-device blocks, sessions & transcripts
Sharing	Web Share API	Native share sheet; clipboard fallback
Hosting	Static (GitHub Pages)	No server at all

Quick Start

Prerequisites

• Node.js 18+
• npm

Development

npm install
npm run dev

Opens https://localhost:5173 with HTTPS (required for microphone access on phones).

Build for Production

npm run build
npm run preview

Output in dist/ ready for static hosting. Pushes to main deploy automatically via GitHub Pages.

Architecture

┌────────────────────────────────────────┐
│ Smartphone Browser (iOS Safari /       │
│ Android Chrome)                        │
├────────────────────────────────────────┤
│ PWA Frontend                           │
│ ┌──────────────────────────────────┐   │
│ │ Microphone → Audio Recording     │   │
│ └──────────────────────────────────┘   │
│ ┌──────────────────────────────────┐   │
│ │ Whisper Worker → Transcription   │   │
│ │ (Runs in Web Worker, keeps UI    │   │
│ │  responsive)                     │   │
│ └──────────────────────────────────┘   │
│ ┌──────────────────────────────────┐   │
│ │ PII Redaction (rules)            │   │
│ └──────────────────────────────────┘   │
│ ┌──────────────────────────────────┐   │
│ │ IndexedDB → Blocks, Sessions,    │   │
│ │ Redacted Transcripts             │   │
│ └──────────────────────────────────┘   │
│ ┌──────────────────────────────────┐   │
│ │ Web Share API → formatted notes  │   │
│ │ then delete block locally        │   │
│ └──────────────────────────────────┘   │
└────────────────────────────────────────┘
         (no backend, no network calls
          after the model is cached)

Usage

First run

1. Welcome screen explains the privacy model
2. Download the Whisper transcription model (~80 MB) — required, happens once

A day of door knocking

1. Tap ● Record at the first door — this creates a session block
2. Stop when the conversation ends; the session is transcribed and saved
3. Repeat at each door — every recording joins the current block
4. When finished, tap Finish block & share
5. Review the formatted notes, then share to any app (or copy)
6. After sharing, everything is deleted from the device automatically

Reviewing mid-block

• The Sessions view lists recorded sessions with timestamps
• Tap one to see its redacted transcript or delete it individually