User (Human-in-the-Loop)

The User client enables human participation in therapy simulations, allowing real people to act as clients/patients in interactive sessions.

Overview

Property	Value
Key	user
Type	Human-in-the-loop
Focus	Interactive Sessions

Description

The User client is a special agent type that delegates client responses to a human operator. This enables interactive training scenarios, human evaluation studies, and live demonstrations where a real person plays the role of a patient while interacting with therapist agents (either AI or human).

Key Features

• Real-time interaction - Human provides responses in real-time
• Full control - Complete flexibility in presenting symptoms and behaviors
• Training mode - Ideal for standardized patient practice
• Evaluation studies - Enables human evaluation of therapist agents

Configuration

YAML Configuration

client:
  agent_type: user

Python Usage

from patienthub.clients import get_client
from omegaconf import OmegaConf

config = OmegaConf.create({
    'agent_type': 'user'
})

client = get_client(configs=config, lang='en')

Parameters

The User client does not require additional configuration parameters as it relies on human input.

Use Cases

• Standardized Patient Training - Medical/psychology students practice with real SPs
• Therapist Agent Evaluation - Human clients evaluate AI therapist quality
• Interactive Demonstrations - Live demos at conferences or training sessions
• Research Studies - Controlled experiments with human participants
• Quality Assurance - Testing therapist agent responses to edge cases

Integration

The User client integrates with PatientHub's interactive interfaces:

• Chainlit Web Demo - Provides a chat interface for human interaction
• Command Line - Text-based interaction through the terminal
• Custom Interfaces - Can be integrated with custom UIs

Example Session

from patienthub.clients import get_client
from patienthub.therapists import TherapistRegistry
from omegaconf import OmegaConf

# Create human client
client_config = OmegaConf.create({'agent_type': 'user'})
client = get_client(configs=client_config, lang='en')

# Create AI therapist
therapist = TherapistRegistry.create("cbt", config={"model": "gpt-4o"})

# Run interactive session
# Human provides client responses, AI provides therapist responses