Sehat Agent
Link to open source: https://github.com/Sachinyadav844/Sehat-Ai.git
SehatAgent is a real-time AI-powered healthcare video assistant that interacts with patients through live video calls, understands symptoms using advanced Large Language Models (LLMs), performs intelligent healthcare triage, analyzes risk levels, verifies safety, and autonomously coordinates healthcare actions.
Unlike traditional healthcare chatbots that only provide text-based suggestions, SehatAgent functions as an AI healthcare agent capable of conducting doctor-like conversations, gathering complete symptom information, assessing severity, generating healthcare reports, booking appointments, and escalating critical cases to healthcare professionals.
The system is designed specifically for resource-constrained and rural healthcare environments where timely access to medical guidance and healthcare services is often limited.
How SehatAgent Works
1. Real-Time AI Video Consultation
The user starts a live video call with the AI healthcare assistant.
Using TruGen API, SehatAgent creates an AI-powered doctor avatar capable of:
- Real-time video interaction
- Natural voice conversation
- Human-like communication
- Facial expression synchronization
- Multilingual communication
The consultation experience feels similar to talking with a healthcare professional.
2. Doctor-Like Symptom Collection
Instead of asking fixed questions, SehatAgent conducts a dynamic consultation.
The AI:
- Listens carefully to patient symptoms
- Understands context
- Maintains conversation memory
- Generates follow-up questions based on responses
Example
Patient:
I have fever and chest discomfort.
SehatAgent:
How long have you been experiencing these symptoms?
Are you having difficulty breathing?
Have you experienced dizziness or weakness?
The AI continues gathering information until it has enough clinical context for assessment.
3. Intelligent Symptom Analysis
Once sufficient information is collected, the system performs:
- Symptom extraction
- Context analysis
- Severity evaluation
- Medical pattern matching
- Risk assessment
The AI identifies:
- Symptom combinations
- Warning signs
- Potential emergency indicators
- Uncertainty levels
4. Healthcare Triage & Risk Classification
SehatAgent automatically classifies patients into:
| Risk Level | Meaning |
|---|---|
| Low Risk | Minor symptoms |
| Medium Risk | Consultation recommended |
| High Risk | Urgent medical attention required |
| Critical Risk | Emergency escalation required |
This process helps prioritize patients efficiently.
5. Safety Verification Layer
Before performing any healthcare action, SehatAgent activates its Safety Verification Engine.
The system checks:
- AI confidence score
- Response reliability
- Uncertainty levels
- Potential hallucinations
- Safety constraints
Only after successful validation does the system proceed with healthcare actions.
This safety-first architecture is one of the major innovations of the project.
6. Autonomous Healthcare Actions
Unlike traditional AI assistants, SehatAgent performs real healthcare workflows.
Possible actions include:
Appointment Booking
- Doctor appointment scheduling
- Telemedicine consultation booking
Healthcare Routing
- Directing patients to appropriate healthcare services
Report Generation
- Creating structured consultation reports
Emergency Escalation
- Triggering emergency workflows
- Connecting patients with healthcare professionals
7. Emergency Escalation System
If the AI detects:
- Chest pain
- Breathing difficulty
- Stroke indicators
- Severe weakness
- Critical symptom combinations
The system automatically:
- Generates a high-risk alert
- Escalates the case
- Connects the patient with a doctor
- Routes to the nearest district hospital or healthcare facility
- Creates an emergency consultation report
This reduces delays during medical emergencies.
8. Multilingual Healthcare Support
SehatAgent supports multiple languages.
The user can select:
- English
- Hindi
- Regional Indian languages
Based on the selected language:
- Entire interface changes
- AI speaks in the chosen language
- Reports are generated in that language
- Healthcare instructions are localized
This improves accessibility for diverse populations.
9. AI Healthcare Report Generation
At the end of every consultation, SehatAgent automatically generates a structured report.
Report Includes
- Patient information
- Symptoms reported
- Symptom duration
- Risk classification
- AI observations
- Safety assessment
- Recommended next steps
- Escalation status
- Appointment details
This report can be shared with healthcare professionals for continuity of care.
Multi-Agent AI Architecture
SehatAgent uses multiple specialized AI agents working together.
Symptom Understanding Agent
Collects and understands patient symptoms.
Severity Assessment Agent
Calculates risk scores and identifies emergencies.
Action Orchestration Agent
Handles appointments, reports, and healthcare workflows.
Safety-Aware Escalation Agent
Detects unsafe situations and manages escalation.
Supervisory Validation Agent
Reviews outputs before execution to ensure safety and consistency.
Key Innovations
Real-Time AI Doctor Video Calls
Human-like healthcare conversations through AI avatars.
Multi-Agent Healthcare Intelligence
Multiple AI agents collaborate like a healthcare support team.
Safety-Aware Healthcare AI
All actions are validated before execution.
Autonomous Healthcare Workflow Execution
The AI performs actions rather than only giving suggestions.
Emergency Escalation Framework
High-risk patients are connected directly to healthcare providers.
Multilingual Accessibility
Healthcare support in multiple languages.
Rural Healthcare Focus
Designed for low-resource and underserved environments.
Technology Stack
Frontend
- React 19
- TypeScript
- Vite
- Tailwind CSS
- Framer Motion
Backend
- Node.js
- Express.js
- Socket.IO
- WebSockets
AI Layer
- OpenAI API
- LangChain
- RAG (Retrieval-Augmented Generation)
- Multi-Agent LLM Architecture
Video AI
- TruGen API
Database
- PostgreSQL
- Prisma ORM
Queue & Caching
- Redis
- BullMQ
Python AI Services
- FastAPI
- Uvicorn
- Pydantic
Security
- JWT Authentication
- bcrypt
- Helmet
Vision
SehatAgent aims to become a safe, intelligent, multilingual healthcare access layer that enables real-time AI-assisted healthcare consultation, risk assessment, and coordinated healthcare action execution for millions of users, especially in underserved and resource-constrained.
This build was uploaded as a hackathon project
1
