Clinical AI Patient Simulator

Deploy AI-powered patient simulations for medical students to practice clinical skills including history-taking, diagnosis, and treatment planning.

Features

• Realistic Patient Interactions: AI agents simulate diverse patient scenarios with authentic emotional responses
• Voice-First Interface: Natural conversational practice via LiveKit real-time audio
• Cultural Competency: Patients with diverse backgrounds and health beliefs
• Avatar Support: Optional Tavus avatar for visual patient representation
• Dynamic Profiles: Load different patient personas for varied training scenarios

Prerequisites

Infrastructure

The VM you run Lattice on must have a running Docker socket:

# Add lattice user to Docker group
sudo adduser lattice docker
sudo systemctl restart lattice

API Keys

Configure your AI and voice providers:

# Required
export OPENAI_API_KEY="your-openai-key"
export LIVEKIT_URL="wss://your-livekit.livekit.cloud"
export LIVEKIT_API_KEY="your-livekit-key"
export LIVEKIT_API_SECRET="your-livekit-secret"

# Optional - for avatar support
export TAVUS_API_KEY="your-tavus-key"
export TAVUS_REPLICA_ID="your-replica-id"
export TAVUS_PERSONA_ID="your-persona-id"

Architecture

This template provisions:

• Pre-built container image with LiveKit agents SDK
• Voice pipeline (STT, LLM, TTS) configured via environment variables
• Dynamic patient persona injected at runtime via Terraform variables
• Optional Tavus avatar integration
• Persistent storage for session data

Patient Profile Configuration

Customize the patient persona via Terraform variables:

# terraform.tfvars
patient_name               = "Maria Santos"
patient_age                = 58
patient_gender             = "female"
patient_cultural_background = "Filipino-American"
patient_diagnosis          = "Type 2 Diabetes with hypertension"
patient_emotional_state    = "Anxious about medication side effects"
patient_health_literacy    = "Moderate"
patient_symptoms           = ["fatigue", "increased thirst", "frequent urination"]
patient_medical_history    = ["Hypertension diagnosed 5 years ago"]
patient_medications        = ["Lisinopril 10mg daily"]
patient_allergies          = ["Penicillin - causes rash"]
patient_introduction       = "Hello doctor. I'm here because my sugar levels are too high."

Use Cases

• Medical Student Training: Practice clinical interviews and examinations
• OSCE Preparation: Standardized patient encounters for exam preparation
• Communication Skills: Improve bedside manner and patient communication
• Diagnostic Reasoning: Practice differential diagnosis and clinical decision-making

Getting Started

1. Create a new workspace from this template
2. Configure your API credentials in the workspace
3. Optionally customize persona.json with your patient profile
4. Launch the agent to start clinical simulations

Built for medical education and clinical skills training.