Walkworth: Walk it. Say it. Keep it. A spoken trek journal on local open-source AI

Description and use

Walkworth lets walkers record trek observations by voice, store data offline, and automatically generate a verified journal with citations using local models and a Temporal workflow.

Problem

Walkers need a way to record trek observations without interrupting their walk or relying on network connectivity.

Outcome

Walkworth enables offline voice-based trek journaling with automatic transcription, extraction, and verified journal generation, eliminating the need for on-trail network or battery-intensive models.

Workflow

• Start a walk by tapping the big button, speak, tap again; phone buzzes to indicate listening and records GPS fix.
• Notes and photos are stored in IndexedDB and uploaded when network is available.
• On the laptop, whisper.cpp transcribes audio clips locally.
• Gemma 3 via Ollama extracts observations, captions photos, and writes journal lines, each line citing source note IDs.
• Temporal workflow orchestrates import, transcribe, align, extract, caption, compose steps with retries.
• FastAPI and SQLite store data locally; Render hosts the finished static journal.
• The journal is printed, published, or shared, with no fabricated content.

Evidence

Walkworth is a trek journal you speak instead of type.

Source

The phone buzzes when it starts and stops listening, so you never have to look at it.

Source

A Temporal workflow transcribes (ElevenLabs Scribe), lines notes up with the GPX track, picks out what you saw, captions photos…

Source

whisper.cpp (the small model) transcribes every clip. It copes with Hindi, Marathi and English in the same sentence, which is how I talk anyway.

Source