Inspiration
When someone is worried about a cough or breathing symptoms, it can be hard to know what details to share or what to ask a clinician. I built Saansathi to help make that conversation easier to prepare for. It is a care-navigation prototype, not a diagnostic tool or a replacement for medical care.
What it does
Saansathi guides someone through a respiratory symptom check-in, turns the details they provide into concise notes for a clinician, and offers general next-step guidance. It also keeps urgent-care guidance visible and reminds users not to enter identifying information.
How we built it
I built the app with Python and Streamlit, with a responsive dark interface. The project includes an optional integration for NVIDIA Nemotron through Nebius Token Factory. When configured, requests are sent through the server-side app; the API key is not exposed in browser code. The hosted demo currently uses prepared sample responses, so its displayed answers are not live Nemotron output.
Challenges we ran into
A key challenge was making the demo useful without a connected model or API key. I added prepared responses so people can still explore the workflow, and made the demo status clear to avoid confusing examples with live AI output. I also learned to keep the product’s scope focused: organize information, explain general next steps, and direct urgent symptoms toward emergency care—without diagnosing or making medical decisions.
Accomplishments that we're proud of
We built a complete, easy-to-follow workflow for checking in on symptoms, preparing notes for a clinician, and asking general questions. Saansathi makes its boundaries visible: it does not diagnose, and it provides urgent-care guidance for severe warning signs. The demo can also be explored with prepared responses when the model is not connected.
What we learned
We learned that a health-related AI prototype needs clear limits and honest status labels. People should be able to tell when a response is a prepared example versus a live model response. We also learned to keep the experience focused on helping someone organize their own information, rather than making decisions for them
What's next for Saansathi
Next, we’d like to test the workflow with users and clinicians, improve accessibility and language support, and refine the urgent-help experience. We’d also like to enable and evaluate the NVIDIA Nemotron integration through Nebius Token Factory, while keeping privacy protections and clear medical boundaries central.
Log in or sign up for Devpost to join the conversation.