Inspiration
Every minute matters in a medical emergency, but most people aren’t sure how to judge the urgency of their symptoms or what to do next. We’ve all seen friends or family panic over a sudden health scare, unsure whether to rush to the ER or just rest at home. Our team wanted to create a tool that empowers anyone, anywhere, to get immediate, reliable triage guidance-especially when time and clarity are critical. With the power of Perplexity’s Sonar API, we saw a chance to bring expert-level triage to everyone’s pocket.
What it does
Emerg-Triage is an AI-powered chatbot that acts like a digital triage nurse. Users simply describe their symptoms in their own words-no medical jargon needed. The chatbot listens, asks smart follow-up questions based on clinical best practices, and rapidly assesses the situation’s urgency. If it detects red flags (like severe injuries or loss of consciousness), it immediately advises emergency action. For less urgent cases, it provides tailored next steps, from home care tips to recommendations for seeing a doctor. In every case, Emerg-Triage delivers clear, actionable guidance-fast.
How we built it
We built Emerg-Triage around the Perplexity Sonar API, which serves as the brain of our chatbot. Our frontend is a simple, friendly chat interface built with React and hosted on Firebase. When a user sends a message, our backend (running on Google Cloud Functions) forwards the conversation to Sonar, along with a prompt that instructs it to act as a triage nurse and follow clinical protocols. Sonar interprets the user’s symptoms, generates follow-up questions, and helps determine the urgency of the situation. We used synthetic patient data to test various scenarios and fine-tuned our prompts to ensure safety and clarity. Throughout, we prioritized privacy and data security, using only anonymized information.
Challenges we ran into One big challenge was teaching the AI to ask the right follow-up questions for a wide variety of symptoms-especially when users describe things in unexpected ways. We spent a lot of time refining our prompts and testing edge cases to make sure the chatbot wouldn’t miss critical red flags or give unsafe advice. Another hurdle was ensuring the chatbot’s recommendations were always clear and actionable, without crossing into making diagnoses or medical claims. Finally, integrating real-time, context-aware conversation flow with Sonar’s API took some creative engineering!
Accomplishments that we're proud of
We’re proud that Emerg-Triage can handle both everyday health questions and true emergencies with the same calm, reliable approach. The chatbot’s ability to dynamically probe for more information-just like a real nurse-was a breakthrough for us. We’re also thrilled with how quickly we were able to build a working prototype that feels genuinely helpful and approachable, thanks to the flexibility of Perplexity’s Sonar API.
What we learned
We learned that building trustworthy healthcare AI is as much about careful conversation design as it is about technology. Prompt engineering is both an art and a science, especially when patient safety is on the line. We also discovered how powerful Perplexity’s Sonar API is for real-time, context-sensitive reasoning, and how important it is to test with a wide range of user inputs. Most of all, we saw firsthand the potential for AI to make healthcare guidance more accessible to everyone.
What's next for Emerg-Triage
Next, we want to expand Emerg-Triage’s capabilities to cover more languages and integrate with real electronic health records (EHRs) for personalized triage. We’re also exploring partnerships with telehealth providers, so users can connect directly to a clinician if needed. Ultimately, our goal is to make Emerg-Triage a trusted first step for anyone facing a health concern-anytime, anywhere.
Log in or sign up for Devpost to join the conversation.