WanderLog: an offline field journal that answers "where did I see that kingfisher last time?"

Description and use

WanderLog is a local‑first field journal that records 30‑second voice memos, parses them into structured sightings and walks, stores data in SQLite, and answers natural‑language queries entirely offline using open‑weight models.

Problem

People who observe nature outside often forget details such as location, time, and context of sightings.

Outcome

WanderLog enables quick, privacy‑preserving capture of nature observations via voice, automatic parsing, and semantic search, all running locally with minimal latency (e.g., 244 ms cold load, 3 ms embedding).

Workflow

• Record a 30‑second voice memo on the trail
• Run npm run log-voice -- <audio> at home to transcribe, parse, and store the data
• The memo becomes a walk entry and linked sightings in SQLite
• Ask a question via npm run demo "<question>" and the local model returns a cited answer
• All operations run on the local machine with no network calls

Evidence

I built WanderLog — a local‑first field journal that turns my coding agent into a naturalist's notebook.

Source

No internet required. After the first run, the embedding model is cached and every query is answered with zero network calls.

Source

Measured on my MacBook (no GPU fan spin‑up, no cloud): cold model load 244 ms, one embedding 3 ms, full semantic search over a 16‑entry journal 9 ms.

Source