Interviewer

The Interviewer NPC conducts structured scripted interviews with clients by sequentially delivering questions from a JSON file.

Overview

Property	Value
Key	interviewer
Type	Rule-based
Focus	Structured Interviews

How It Works

1. Data Loading: Loads a list of questions from a JSON file at configs.data.
2. Sequential Delivery: Each generate_response() call records the incoming message and returns the next question in the list.
3. Completion: Returns "[No more questions]" when the question list is exhausted.

Usage

Python

from omegaconf import OmegaConf
from patienthub.npcs.interviewer import InterviewerNPC

config = OmegaConf.create({"data": "data/evaluations/surveys/default_survey.json"})
interviewer = InterviewerNPC(configs=config)

question = interviewer.generate_response("start")
print(question)

Configuration

Parameter	Type	Default	Description
data	string	data/evaluations/surveys/default_survey.json	Path to the questions JSON file

Data Format

The questions file is a JSON array of strings:

[
  "How have you been feeling this week?",
  "Have you experienced any changes in sleep or appetite?",
  "How would you rate your mood on a scale of 1–10?"
]

Use Cases

• Pre-session intake: Administer a standardized questionnaire before therapy begins
• Post-session evaluation: Collect structured feedback from client agents
• Research data collection: Gather consistent responses across simulated sessions
• Training simulations: Practice interview skills with simulated clients

Example: Intake Before Therapy

from omegaconf import OmegaConf
from patienthub.clients import get_client
from patienthub.npcs.interviewer import InterviewerNPC

interviewer = InterviewerNPC(
    configs=OmegaConf.create({"data": "data/evaluations/surveys/default_survey.json"})
)
client = get_client(agent_name="patientPsi", lang="en")

question = interviewer.generate_response("start")
while question != "[No more questions]":
    print("Interviewer:", question)
    answer = client.generate_response(question)
    print("Client:", answer.content if hasattr(answer, "content") else answer)
    question = interviewer.generate_response(answer.content)