AI-Powered Smartphone Decision Support System

An intelligent full-stack decision support system that helps users identify and rank smartphones using computer vision, multi-criteria decision analysis, and optional AI-generated explanations.

Features

• Identify smartphone models from uploaded images or live camera.
• Rank devices with Direct Weighting + TOPSIS or optional Pairwise AHP + TOPSIS.
• Compare raw specifications side-by-side with bar charts.
• Show transparent scoring inputs, sensitivity analysis, and comparison results.
• Generate natural-language explanations with OpenRouter or Groq when configured.
• Durable per-analysis URLs that survive refresh.
• Run a React frontend and FastAPI backend together from the project root.

Tech Stack

• Backend: FastAPI, SQLite, SQLAlchemy, pytest, ONNX Runtime
• Frontend: React, Vite, TypeScript, Recharts, Framer Motion, Axios
• Vision: ONNX Runtime with 3 trained YOLO models

Project Structure

src/
  api/                API endpoints and schemas
  database/           Database models, connection, and seed data
  decision_engine/    Preference validation, AHP, and TOPSIS analysis
  vision/             Image transforms, model lifecycle, and inference
  assistant/          Shared provider adapters, context, and fallback behavior
  catalog/            Active catalog and retirement-aware queries
  runtime/            Validated settings and capability configuration
  saved_analysis/     Durable ranking persistence operations
frontend/
  src/                React application source
  e2e/                Playwright browser journey tests
tests/                Unit and integration tests
computer_vision/      Training artifacts and exported models
docs/                 ADRs, verification evidence, defense artifacts

Environment Setup

Copy .env.example to .env and add only the values you plan to use:

OPENROUTER_API_KEY=your_openrouter_key

If OpenRouter or Groq keys are missing, tests remain offline and the app uses local fallback answers. Training notebooks read ROBOFLOW_API_KEY from the environment and do not contain a key.

Install

From the project root:

python -m venv .venv
.\.venv\Scripts\activate

npm run install:all

Run

Start the backend and frontend together:

npm run dev

Expected local services:

• Backend: http://127.0.0.1:8000
• Frontend: http://localhost:5173

Test

Run the authoritative local verification command:

npm run verify

Architecture terms and decisions are documented in CONTEXT.md

Database Migrations

Upgrade an existing database:

python -m alembic upgrade head

Fresh databases use the same command before startup seeding. Application startup also runs pending migrations automatically.