README.md

metadata

title: Maya – Klinik Sehat Sentosa (Role‑Play)
emoji: 🩺
colorFrom: green
colorTo: blue
sdk: gradio
sdk_version: 4.44.1
app_file: app.py
pinned: false
hf_oauth: true

Maya – Klinik Sehat Sentosa (Role‑Play)

Overview

• Gradio app where students interview “Maya,” a clinic owner persona, to practice requirements elicitation for an appointment & queueing system.
• “Generate Requirements JSON” creates a structured summary from the chat transcript. No data is persisted.

Run Locally

1. Python 3.9+
2. Install deps: pip install -r requirements.txt
3. (Optional) Model selection (default uses GPT‑OSS via Together): export HF_MODEL=openai/gpt-oss-20b

   If using GPT‑OSS/Together, set provider key:

   export TOGETHER_API_KEY=xxxxxxxxxxxxxxxx
4. Start the app: python app.py
5. Click “Sign in” to grant the app an access token for the Hugging Face Inference API.

Deploy on Hugging Face Spaces

1. Create a new Space (SDK: Gradio).
2. Add files (app.py, requirements.txt, README.md). Push via Git.
3. Users click “Sign in” to authorize Inference API use.
4. (Optional) Set Space Variables:
   • HF_MODEL (e.g., openai/gpt-oss-20b)
   • TOGETHER_API_KEY (required for GPT‑OSS via Together)

Persona & Scope

• Persona: Maya is friendly, busy, non‑technical; answers concretely; asks clarifying questions when prompts are vague.
• Scope: scheduling, queue order, reminders, daily counts only. No billing, insurance, or EMR details.
• Constraints: low digital literacy, intermittent internet, must run on existing Android phones, small budget.

Usage Tips

• Ask clear, specific questions (who, what, when, where, how, constraints, edge cases).
• Click “Generate Requirements JSON” to see a structured summary of established requirements.
• JSON fields: actors, goals, constraints, functional & non‑functional requirements, user stories, edge cases, assumptions, open questions, acceptance criteria.

Notes

• Uses Hugging Face Inference (huggingface_hub.InferenceClient) with streaming responses.
• No database — state is in memory per session.
• If the JSON fails to parse, the raw text is shown instead.