FirstAid AI Review
A Streamlit triage chatbot using Gemini 2.5 Flash, with a local outbreak radar and clinic finder.
Each review covers deployability, value versus commercial alternatives, and privacy model. Tools that can run locally were started and exercised; mobile or backend-dependent tools were assessed from published builds, source code, and deploy guides. Ratings reflect what we were able to verify.
FirstAid AI — Open Source Symptom Triage Chat
A Streamlit triage chatbot using Gemini 2.5 Flash, with a local outbreak radar and clinic finder. Last updated: 2026-06-21.
One-sentence verdict: A compact privacy-minded triage prototype that demonstrates a clean LLM + local SQLite + maps flow, but not a substitute for clinically validated symptom checkers.
What the System Is
FirstAid AI is a 129-line Streamlit prototype with three documented components:
- Symptom checker chat — a conversational interface powered by Gemini 2.5 Flash. The system prompt asks clarifying questions, lists potential conditions, and provides next steps with a disclaimer that it is an AI, not a doctor.
- Outbreak radar — an anonymized SQLite log of symptom category plus zip code, aggregated to show local clusters when case counts reach a threshold.
- Clinic finder — a text box for zip code or city that generates a Google Maps search link.
The chat history lives only in st.session_state, so it is lost on page refresh. The outbreak radar stores only symptom category and location; it does not persist the full chat transcript.
| Key data | |
|---|---|
| Category | Symptom Checker |
| Language | Python / Streamlit |
| License | Unknown |
| Self-hosted | Yes |
| AI provider | Google Gemini 2.5 Flash |
| Database | SQLite (outbreak radar only) |
| Code size | ~129 lines |
How to Install and Deploy
cd /data2/docker/going-global/repos/FIRST-AID-APP
pip install streamlit google-generativeai
echo 'GEMINI_API_KEY = "your-key"' > .streamlit/secrets.toml
streamlit run app.py --server.port 8502
Then open http://localhost:8502.
There is no Docker Compose for this project.
How to Test
The documented test flow is:
- Open
http://localhost:8502. - Enter a symptom description in the chat.
- Observe whether the model asks follow-up questions before producing a triage response.
- Submit a zip code to the outbreak radar and verify it appears in the local SQLite database.
- Use the clinic finder to generate a Google Maps URL.
The app is intentionally lightweight. It does not store medical histories, integrate with EHRs, or provide clinical validation.
Privacy & Compliance
FirstAid AI is not HIPAA compliant and is explicitly not a medical diagnosis tool. The chat sends symptom descriptions to Google Gemini. The outbreak radar stores only anonymized category and zip code. Do not enter PHI that you are not comfortable sending to an external LLM.
FirstAid AI vs Commercial Symptom Checkers
| Dimension | FirstAid AI | Commercial Symptom Checker (e.g., Ada Health, K Health) |
|---|---|---|
| Cost | Free / self-hosted | Freemium or subscription |
| Data model | Chat ephemeral; radar anonymized and local | Cloud-hosted full history |
| Clinical validation | None | Often medically reviewed |
| Setup effort | Install Python deps + Gemini key | Mobile app signup |
| Offline use | No (needs Gemini API) | No |
| Customizability | Full source code | Closed |
Who Should Use It
- Developers learning how to wire a chatbot, SQLite, and maps into a Streamlit app.
- Teams that need a starting point for a privacy-first triage interface.
- Experimenters who understand the limits of LLM-based health advice.
Who Shouldn't Use It
- Anyone seeking actual medical diagnosis or triage.
- Users who need persistent chat history or longitudinal symptom tracking.
- Organizations that require HIPAA-compliant handling of health data.
FAQ
Is FirstAid AI a medical diagnosis tool?
No. It is a prototype that provides general information based on an LLM. It explicitly disclaims being a doctor.
Does FirstAid AI store my chat history?
No. Chat history is kept in Streamlit session state and disappears when you refresh the page. Only anonymized symptom category and zip code are stored in the local SQLite radar.
What do I need to run it?
Python, the required packages, and a Google Gemini API key. There is no Docker Compose.
Verdict
FirstAid AI is a compact, privacy-minded triage prototype. It demonstrates a clean pattern for LLM-powered symptom intake with local anonymized logging. It is not a substitute for clinically validated symptom checkers or professional medical advice, but it is a credible starting point for developers.
Ratings: Deployability 4/5 · Value vs Commercial 2/5 · Privacy Compliance 2/5