Fieldwork

Qualitative research companion — transcription, coding, and analysis powered by local AI.

Fieldwork is a desktop app for researchers, interviewers, and oral historians. It transcribes audio locally using OpenAI Whisper, provides speaker detection and correction tools, and includes an AI-powered Analysis Lab for thematic coding.

Product Modules

Module	Status	What it does
Studio	Alpha	Record, transcribe, edit, speaker correction, export (CSV, TXT, DOCX, PDF)
Lab	Alpha	AI-powered thematic analysis, codebook generation, coded excerpt list with click-to-jump audio
Field	Planned	Interview planning, copilot, practice mode, data provenance

Current Structure

Active app

• src/ Tauri frontend — Transcript Studio, Analysis Lab, settings, and persistence UI.
• src-tauri/ Desktop shell, backend process orchestration, persistence commands, and export commands.
• backend/transcription_server.py Flask + Whisper backend with optional speaker tooling and document generation.
• backend/requirements.txt Python dependency list for the local transcription engine.
• scripts/build-backend.sh PyInstaller sidecar build script for packaged macOS builds.

Reference material

• docs/README.md Start-here index for architecture, implementation, and release docs.
• docs/architecture/ Productization architecture and transfer notes.
• docs/implementation/ Execution plans and ticket-by-ticket build sheets.
• docs/release/ Release checklist and private alpha distribution process.
• docs/planning/ Product naming and roadmap notes.
• docs/archive/ Recovery notes and recovered documentation from the original Whispr app.
• archive/legacy-electron/ Old Electron implementation kept only as historical reference.
• archive/brand/ Previous Whispr-era brand assets kept for reference.
• archive/dev-tools/ Older test scripts and one-off recovery utilities kept out of the active app root.

Run Locally

1. Install the Tauri CLI dependency:

npm install

2. Create the local Python environment and install the Whisper backend:

python3 -m venv .venv
source .venv/bin/activate
python -m pip install -U pip
python -m pip install -r backend/requirements.txt

3. Start the desktop app:

npm run dev

The Tauri app will launch and start the local Python backend automatically on http://127.0.0.1:53721.

Build A Desktop Bundle

npm run build

For a friend-ready macOS build, first freeze the Python backend into a bundled sidecar:

npm run build:backend
npm run build

Or both in sequence:

npm run build:all

Productization Docs

If you are picking the repo up for release engineering or cleanup work, read:

1. docs/README.md
2. docs/implementation/FIELDWORK_PRODUCTIZATION_BUILD_SHEET.md
3. docs/architecture/ARCA_TO_FIELDWORK_TRANSFER_MAP.md
4. docs/release/RELEASE_CHECKLIST.md

Distribution

• The sidecar build currently targets the host machine architecture only
• The packaged backend is large (~1-2 GB) because it includes Python + Whisper + PyTorch
• First packaged launch can take up to a minute while the onefile sidecar extracts
• Whisper models are downloaded on first use to ~/.cache/whisper/

AI Analysis

The Analysis Lab supports three AI providers:

• Ollama (local) — no data leaves the machine
• OpenAI — cloud, requires API key
• Anthropic (Claude) — cloud, requires API key

Configure the provider in the Analysis Lab sidebar, then click "Analyse Themes" to run thematic coding on the current transcript.

Validation

python3 -m py_compile backend/transcription_server.py
node --check src/app.js
cargo check  # inside src-tauri/