Open source projects
At SERC, we believe conservation technology should be accessible to every program, regardless of budget or technical resources. We build our tools in the open and share them freely so other rehabilitators and researchers can benefit, adapt, and contribute.
turtle-ai
AI-powered monitoring for rehabilitating turtles
turtle-ai is an open-source, locally-run AI monitoring system designed to detect signs of distress in rehabilitating turtles using live RTSP camera feeds. Built for resource-limited rehabilitation environments, the system runs entirely on local hardware with no cloud dependency, capturing still frames every ten minutes and evaluating them using the Gemma 3n model via Ollama. When distress is detected — such as a turtle in a carapace-up position, showing signs of entrapment, or exhibiting unusual inactivity — the system sends a real-time WhatsApp alert via Twilio for immediate intervention.
Presented at the 23rd TSA Annual Symposium, 2025
- ✓ Detects carapace-up positioning, entrapment, unusual inactivity, and aggressive interactions
- ✓ Fully local inference using Gemma 3n and Ollama — no cloud services required
- ✓ Containerized architecture using Docker for easy deployment
- ✓ Real-time WhatsApp alerts via Twilio with debounce logic to prevent alert fatigue
- ✓ Prompt-engineered for turtle-specific distress detection with minimized false positives
- ✓ Open-source and freely available for use by other rehabilitation programs
RTSP cameras stream live enclosure video
Python script captures still frames every 10 min
Gemma 3n (via Ollama) evaluates each frame locally
Distress detected — WhatsApp alert via Twilio
Web app logs observations and human review
ask-serc
RAG chatbot for wildlife rehabilitators and volunteers
ask-serc is an AI-powered chatbot that helps wildlife rehabilitators and volunteers quickly access information about Virginia native animal care. Personified as "Susie," an Eastern Box Turtle ambassador, the system uses retrieval-augmented generation (RAG) to route queries between SERC's own knowledge base for medical and rehabilitation protocols and Google Search for natural history and species identification. Every interaction is logged for volunteer review, helping the team continuously identify and fill gaps in the knowledge base.
- ✓ Intelligent tool routing between SERC knowledge base and Google Search
- ✓ Powered by Gemini via Vertex AI with RAG grounding through Vertex AI Search
- ✓ Knowledge gap tracking — every query logged for volunteer review
- ✓ User feedback system with thumbs up/down response ratings
- ✓ Admin dashboard for reviewing knowledge gaps at /admin
- ✓ Automated Cloud Function syncs Google Drive documents daily
Volunteers upload care documents to shared Google Drive
Cloud Function syncs files to Cloud Storage daily
User query triggers tool selection via Google ADK
Gemini generates a response grounded in SERC knowledge base or Google Search
Exchange logged as a potential knowledge gap for volunteer review
Contribute or collaborate
Our projects are open to contributions from developers, researchers, and wildlife professionals. Whether you want to extend turtle-ai for a different species, adapt it for your own facility, or propose a new project, we welcome your involvement.
We are particularly interested in collaborations that expand the use of conservation technology in resource-limited rehabilitation environments.
Wildlife rehabilitation programs are almost always volunteer-run and resource-constrained. Proprietary tools create barriers to adoption. Open-source conservation technology means any program, anywhere, can benefit from these advances without cost.
It also means the community can improve, audit, and adapt the tools for their own species and environments — advancing the field together.
All projects are open source
Source code, documentation, and setup instructions available on GitHub.