DEV Community

Nischay Shukla
Nischay Shukla

Posted on

Week1 Dev Challenge

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass

What I Built

Trailcard turns a voice memo from a walk into a printable one-page field card. You record birdsong on your phone and put it in your pocket. Afterwards, one command gives you:

a checklist of the birds identified by their calls
a short journal entry about the walk
a “listening challenge” for next time

You print it, put it in your pocket, and go back outside. The screen is only used for about a minute. It’s for beginner birders, hikers, and anyone who wants to notice more on a walk without staring at a phone while doing it.

Code

Trailcard

Record birds on a walk, print a field card afterwards. Bird calls are identified offline with BirdNET (open source), and a local open-weight model (via Ollama) writes the journal entry.

Run locally: pip install birdnet-analyzer, ollama pull llama3.2, then python trailcard.py ./recordings --lat 26.85 --lon 80.95

Deploy: push to GitHub, then Render > New > Blueprint (uses render.yaml).




How I Built It

BirdNET-Analyzer (open source) identifies bird calls from audio.

I pass the latitude, longitude and current week so it only considers birds that are plausible for that place and season.

An open-weight model served by Ollama (Llama 3.2 by default) writes the journal entry. Its prompt contains only the species BirdNET detected and tells it not to add other birds or facts.
Everything runs locally. After the one-time model downloads, no internet connection or API key is needed, which suits trails with no signal.
If Ollama isn’t running, the tool still produces a card with a template journal entry.
The whole tool is a single readable Python file, trailcard.py. The repo also includes an optional web wrapper and Render Blueprint config that I haven’t deployed.

Why Does Open Innovation Matter?

Offline use. A closed API can’t identify a bird on a ridge with no signal. Open models can run on your own laptop.
Privacy. Your recordings and your location never need to leave your machine.
Contribution. It’s one small file, so people can add regions, other taxa such as frogs and insects, or other languages.

Top comments (0)