A Full-Stack Native Android Application that leverages Google Gemini Pro to estimate vehicle market values in real-time.
Buying a used car is confusing. Static calculators use outdated math, while human experts are expensive. I wanted to build a bridge: An app that "thinks" like an expert. By using a Large Language Model (LLM), this app considers brand value, luxury status, and market depreciation logic rather than just crunching raw numbers.
Unlike standard calculators, this app sends a complex prompt to Google Gemini Pro. It analyzes the Make, Model, Year, Engine, and HP to determine a price based on "Luxury Status" and "Indian Market Trends."
Reliability is key. I engineered a Dual-Engine System:
- Primary: Queries the AI for a high-accuracy market prediction.
- Fallback: If the AI service is down or rate-limited, the system automatically switches to a Mathematical Depreciation Algorithm (Math Fallback) to ensure the user always gets a result.
The Android app is built with Kotlin Coroutines. It performs network requests on background threads (Dispatchers.IO), ensuring the UI remains buttery smooth (60fps) and never freezes while waiting for the server.
Includes a Server-Side Rendered (SSR) web dashboard (dashboard.html). It uses Jinja2 templates to log and visualize every request coming from the mobile app in real-time.
The Python Flask backend maintains a live log of all incoming traffic and AI inference requests. This allows for debugging and performance monitoring in real-time.
![Server Logs]
Terminal output showing successful API requests and template rendering.
graph LR
A[📱 Android App] -- JSON Post Request --> B[🌐 Python Flask Server]
B -- Prompt Engineering --> C[🤖 Google Gemini API]
C -- Text Response --> B
B -- Regex Parsing --> B
B -- JSON Response --> A