Psyche
Psyche is an LLM-based psychiatrist agent driven by a structured system prompt, designed for intake-style clinical interviews.
Overview
| Property | Value |
|---|---|
| Key | psyche |
| Type | LLM-based |
| Focus | Structured intake-style interview |
How It Works
- Prompt Loading: Loads a system prompt from
data/prompts/therapist/psyche.yaml. - Session Init: Initializes the conversation with that system prompt.
- Response Generation: Each
generate_response()call appends the client message and asks the model for the next reply.
Usage
CLI
uv run python -m examples.simulate therapist=psyche
Python
from patienthub.therapists import get_therapist
therapist = get_therapist(agent_name="psyche", lang='en')
response = therapist.generate_response("I've been struggling with sleep and low mood.")
print(response.content)
Configuration
| Parameter | Type | Default | Description |
|---|---|---|---|
agent_name | string | "psyche" | Therapist identifier |
prompt_path | string | data/prompts/therapist/psyche.yaml | Path to prompt file |
model_type | string | "OPENAI" | Model provider key |
model_name | string | "gpt-4o" | Model identifier |
temperature | float | 0.7 | Sampling temperature |
max_tokens | int | 8192 | Max response tokens |
max_retries | int | 3 | API retry attempts |
lang | string | "en" | Language key for prompt loading |
Use Cases
- Intake-style interview simulations (psychiatrist role)
- Prompt-driven therapist baseline
- Evaluating client agents in a clinical assessment context