Clinical Scoring

Validate PHQ-9/GAD-7 calculations

Correct score mapping

NLU Intent Classifier

Detect intents accurately

≥ 90 % precision

RAG Response Module

Generate contextual, empathetic text

Relevant responses

Privacy Layer

Ensure data not transmitted externally

100 % local retention

Doctor Recommender

Rank clinicians by location and rating

Top 3 accurate results

Ease of Use

92 %

Interface Design

90 %

Empathetic Responses

95 %

Privacy Confidence

96 %

Overall Satisfaction

93 %

TC-01

User fills PHQ-9 form

Score = sum of answers

Accurate output

Pass

TC-02

User types “I feel anxious lately”

Detect intent = Anxiety Query

Correctly identified

Pass

TC-03

User mentions self-harm keyword

Trigger escalation protocol

Activated

Pass

TC-04

Network disconnect mid-chat

System resumes gracefully

Reconnected without data loss

Pass

TC-05

Request for nearest doctor

Top clinicians ranked by distance & rating

Accurate list displayed

Pass

Processor

Intel Core i5

Intel Core i7 / Ryzen 7

RAM

8 GB

16 GB

Storage

250 GB SSD

512 GB SSD

Network

10 Mbps

25 Mbps

OS

Windows 10 / Ubuntu 22.04

Development environment

Frontend

React + TypeScript

Chat and dashboard interface

Backend

Python, Flask / FastAPI

Logic and API integration

AI Framework

TensorFlow, Transformers

Intent and emotion detection

LLM API

Gemini 2.5 / Google Generative AI

Empathetic response generation [13]

Database

LocalStorage / Firebase

Local user data storage

Security

AES-256, TLS 1.3

Data protection [10]

Cloud

Google Cloud Platform

Hosting and scalability

Tools

VS Code, GitHub, Postman

Development and testing