Trail Buddy: An Open-Source AI Field Guide for Getting Outside

Описание и применение

Trail Buddy — это веб‑приложение, которое позволяет пользователям быстро записывать наблюдения во время прогулки и находить похожие записи в природе, используя модель MiniLM в браузере без отправки данных на сервер.

Задача

People want a quick way to note observations while walking and find related nature guide entries without sending data to external services.

Результат

Trail Buddy enables quick, private, and offline matching of field observations to a nature guide, eliminating the need for external APIs and reducing data privacy concerns.

Рабочий процесс

• 1. User opens the web app.
• 2. User types a description of an observation (e.g., a bird call).
• 3. The MiniLM model in Transformers.js generates an embedding for the input.
• 4. The app generates embeddings for each field‑guide note and compares them using cosine similarity.
• 5. The closest match is displayed to the user.
• 6. The note is stored in the browser’s local storage for future reference.

Доказательства

It uses Transformers.js to run the open‑weight MiniLM sentence embedding model in the browser.

Источник

The first model load needs an internet connection to download the library and model. After that, inference happens in the browser; this version doesn’t guarantee fully offline use.

Источник