Skip to main content

Deprofile

Decomposing Clinical Profiles for Mental Health Patient Simulation

Overview​

Deprofile role-plays a patient built from a decomposed clinical profile: a clinical assessment (symptoms, risks, Big Five), assessment and counseling dialogue records, and a social-media timeline (symptoms and life events). At simulation time, clinical positive/negative symptoms are rendered as natural-language descriptions (with the symptom timeline woven into the positive symptoms), and the life-event timeline is injected as rendered memory cards.

The character file consumed by this client is produced by the Deprofile generator, which assembles and validates these sources into a single DeprofileCharacter record. See that page for the data pipeline, resource files, and output format.

Key Features​

  • Decomposed profile: Combines a clinical profile with matched assessment/counseling dialogues and a social-media timeline.
  • Timeline memory: Life events are surfaced from processed memory cards (timeline_memory.life_event.cards[*].card_text), falling back to the raw timeline when cards are unavailable.
  • Localized rendering: Symptom descriptions and relative-time phrasing are drawn from localized constants (en / zh).

Usage​

CLI​

patienthub simulate client=deprofile

Simulate a specific character in the file by index:

patienthub simulate client=deprofile client.data_idx=0

Python​

from patienthub.clients import get_client

client = get_client(agent_name='deprofile', lang='en')
response = client.generate_response("How have you been feeling this week?")
print(response.content)

Configuration​

OptionTypeDefaultDescription
prompt_pathstrdata/prompts/client/deprofile.yamlPath to the role-play prompt file
data_pathstrdata/characters/deprofile.jsonPath to the generated character file
data_idxint0Index of the character in the file
max_dialogue_snippetsint10Max assessment/counseling snippets injected
max_timeline_itemsint50Max symptom/life-event timeline items injected

Common API-model options (model_type, model_name, temperature, max_tokens, max_retries, lang) are inherited from the shared client configuration — see the overview.

Character Data Format​

data/characters/deprofile.json is a JSON array of DeprofileCharacter records. Each record carries the clinical profile, both dialogue records, the selected social user's symptom and life-event timelines, and the processed timeline_memory (graph, episodes, and memory cards). The full schema and an annotated example are documented in the Deprofile generator → Output Format.

To produce or extend this file, use the generator:

from omegaconf import OmegaConf
from patienthub.generators import get_generator

config = OmegaConf.create({
    "agent_name": "deprofile",
    "profile_id": "0069",
    "output_path": "data/characters/deprofile.json",
})
generator = get_generator(agent_name="deprofile", configs=config, lang="zh")
character = generator.generate_character()

See Also​