NaturalDex
A privacy-first nature collection game application that encourages users to go outdoors and photograph plants, animals, or fungi. It uses a locally running visual AI model to identify species, allowing users to collect field-guide cards and accumulate experience points. The application operates entirely offline without relying on any cloud services.
Timeline (last 90 days)
NaturalDex is a nature field guide game developed by developer Kevin Forsgren during Hacktoberfest 2026, encouraging users to go outdoors to identify real species and collect achievements.
NaturalDex is essentially a mobile-first web application powered by local AI and a self-hosted architecture, adopting a PWA architecture that is fully offline, privacy-first, and independent of cloud services.
NaturalDex uses Ollama to run the local vision model llava-phi3 for species identification, with image classification never sent to any cloud service.
NaturalDex's technology stack includes React 18 + Vite 6, TailwindCSS v4, FastAPI + Pydantic v2 (based on Python 3.14+), SQLAlchemy 2.0 + SQLite, Ollama, and JWT + Argon2.
NaturalDex requires users to take three photos from different angles to provide to the local vision model, in order to improve species identification accuracy and address common misidentification issues with single-photo recognition.
The NaturalDex system generates three personalized tasks per day, covering the three categories of plants, animals, and fungi respectively.
NaturalDex designed a challenge-response QR code pairing protocol, combined with an ngrok tunnel, enabling phones on a cellular network to complete pairing with a laptop on a local Wi-Fi network.
NaturalDex achieves idempotency by setting a database unique constraint on the observation-record-to-collection-card association table, intercepting duplicate submissions at the database write stage to prevent duplicate card collection and false XP accumulation.
NaturalDex is open-sourced under the GPL-3.0 license, with code hosted on GitHub (KevinForsgren/NaturalDex).
All Facts (9)
NaturalDex is a nature field guide game developed by developer Kevin Forsgren during Hacktoberfest 2026, encouraging users to go outdoors to identify real species and collect achievements.
50%UnverifiedNaturalDex is essentially a mobile-first web application powered by local AI and a self-hosted architecture, adopting a PWA architecture that is fully offline, privacy-first, and independent of cloud services.
50%UnverifiedNaturalDex uses Ollama to run the local vision model llava-phi3 for species identification, with image classification never sent to any cloud service.
50%UnverifiedNaturalDex's technology stack includes React 18 + Vite 6, TailwindCSS v4, FastAPI + Pydantic v2 (based on Python 3.14+), SQLAlchemy 2.0 + SQLite, Ollama, and JWT + Argon2.
50%UnverifiedNaturalDex requires users to take three photos from different angles to provide to the local vision model, in order to improve species identification accuracy and address common misidentification issues with single-photo recognition.
50%UnverifiedThe NaturalDex system generates three personalized tasks per day, covering the three categories of plants, animals, and fungi respectively.
50%UnverifiedNaturalDex designed a challenge-response QR code pairing protocol, combined with an ngrok tunnel, enabling phones on a cellular network to complete pairing with a laptop on a local Wi-Fi network.
50%UnverifiedNaturalDex achieves idempotency by setting a database unique constraint on the observation-record-to-collection-card association table, intercepting duplicate submissions at the database write stage to prevent duplicate card collection and false XP accumulation.
50%UnverifiedNaturalDex is open-sourced under the GPL-3.0 license, with code hosted on GitHub (KevinForsgren/NaturalDex).
50%