Skip to main content

Generators

Generators in PatientHub create synthetic client profiles, therapy scenarios, and conversation data for training and research purposes.

Overview​

Generators enable the automatic creation of:

  • Client Profiles - Synthetic patient backgrounds and presenting problems
  • Therapy Scenarios - Structured situations for simulation
  • Conversation Data - Training data for AI models
  • Character Files - Complete character specifications for client agents

Available Generators​

GeneratorKeyDescription
AnnaAgent GeneratorannaAgentCreates character files for AnnaAgent clients
ClientCast GeneratorclientCastGenerates diverse client characters
Psyche GeneratorpsycheCreates psychologically rich client profiles

Usage​

Command Line​

# Generate with defaults
uv run python -m examples.generate

# Override generator type
uv run python -m examples.generate generator=psyche

In Code​

from patienthub.generators import GeneratorRegistry

# Create a generator
generator = GeneratorRegistry.create("annaAgent", config={
    "model": "gpt-4o"
})

# Generate a client profile
profile = generator.generate({
    "disorder": "depression",
    "severity": "moderate"
})

Generation Types​

Client Profile Generation​

Create detailed client backgrounds including:

  • Demographics
  • Presenting problems
  • History and background
  • Personality traits
  • Communication style
profile = generator.generate_profile({
    "disorder": "anxiety",
    "age_range": "25-35",
    "gender": "any",
    "background": "professional"
})

Character File Generation​

Generate complete character specification files:

character = generator.generate_character({
    "profile": profile,
    "format": "json",
    "include_dialogue_examples": True
})

Scenario Generation​

Create therapy scenarios:

scenario = generator.generate_scenario({
    "type": "initial_session",
    "client_profile": profile,
    "therapeutic_approach": "CBT"
})

Configuration​

Generator Settings​

generator:
  type: annaAgent
  config:
    model: gpt-4o
    temperature: 0.8
    diversity: high
    output_format: json

Output Formats​

Generators support multiple output formats:

  • JSON - Structured data for programmatic use
  • YAML - Human-readable configuration files
  • Markdown - Documentation-friendly format

Batch Generation​

Generate multiple profiles at once:

from patienthub.generators import GeneratorRegistry

generator = GeneratorRegistry.create("clientCast")

# Generate 10 diverse profiles
profiles = generator.batch_generate(
    count=10,
    diversity_constraints={
        "disorders": ["depression", "anxiety", "ptsd"],
        "age_range": [18, 65],
        "balance_gender": True
    }
)

Creating Custom Generators​

You can create custom generators:

from patienthub.generators.base import Generator

class MyCustomGenerator(Generator):
    def __init__(self, config):
        super().__init__(config)
        # Initialize your generator

    def generate(self, params):
        # Generate content
        return generated_content

Then register it:

from patienthub.generators import GeneratorRegistry

GeneratorRegistry.register("my_generator", MyCustomGenerator)

Best Practices​

  1. Diversity - Ensure generated profiles represent diverse populations
  2. Realism - Validate generated content against clinical knowledge
  3. Ethics - Review generated content for appropriateness
  4. Versioning - Track generated content versions for reproducibility

See Also​