TrotroLive

We are making trotro transportation information readily available

Comment 1

Inspiration

The inspiration behind TrotroLive came during my time as a media intern at Orange 107.9 in Kumasi. While commuting daily to work via trotro (public minibuses), I often witnessed arguments and confusion about fare prices and route information. These frequent quarrels made it clear that there was a serious information gap. That’s when I got the idea to create a platform that provides clear, reliable, and accessible information on Ghana's trotro system.

What it does

TrotroLive is a smart transportation assistant for Ghana’s trotro system. It provides real-time and verified information on:

  • Routes and destinations
  • Fare prices
  • Trotro stations
  • Trip schedules
  • Route planning
  • Station-to-station trip insights
  • AI-powered chatbot assistance (in English and local languages)

The platform makes it easier for passengers to plan their journeys, avoid fare scams, and access transport information anytime through a web app, mobile app, and chatbot.

How we built it

We built TrotroLive with:

  • Backend: A robust Django REST API that powers the core logic and handles routes, stations, trips, and fare data using GTFS (General Transit Feed Specification).
  • AI Agent (MCP): An advanced AI chatbot using LLMs to answer user queries on WhatsApp and web platforms.
  • Dataset Integration: Leveraged transport data from 7 African cities, including Accra and Kumasi.
  • Mobile Apps: Two apps — one for public users and one for community data collection.
  • Multi-language Support: The AI is trained to respond in both English and local Ghanaian languages.

🚀 Features

Core AI Capabilities

  • Natural Language Question Answering for transport-related queries
  • Route Planning and optimization based on user inputs
  • Fare Estimation with real-time pricing logic
  • Station Information via a comprehensive database

Challenges we ran into

One of our biggest challenges was hosting. We initially deployed the API on a shared cPanel hosting, which caused frequent server crashes and "500 Internal Server Errors" after integrating the AI model. The shared hosting environment wasn’t suitable for our compute-heavy tasks. We’re currently migrating to a VPS on Google Cloud, but faced delays due to lack of a credit card to activate the free trial. Thankfully, we acquired a virtual card and are resolving those issues.

Accomplishments that we're proud of

  • Built a scalable Django backend from scratch
  • Developed an AI-powered chatbot (MCP) for trotro info on WhatsApp
  • Launched two mobile apps — one for the public and one for trotro data collectors
  • Created a community-driven data model that allows users to contribute information
  • Supported by data from over 7 African cities including Ghana’s Accra and Kumasi

What we learned

  • Public transport data in Ghana is either outdated or missing
  • We learned to build our own data pipeline and collect real-world data
  • Importance of community-driven data collection to improve AI accuracy
  • Hosting matters — shared hosting is not ideal for modern AI systems
  • Learned how to integrate AI with traditional backend systems

What's next for TrotroLive

  • Fully migrate from cPanel to VPS (Google Cloud) for better stability
  • Multi-language Support (English & Ghanaian local dialects)
  • Launch the public web app, mobile app, and USSD system
  • Continue training the AI model for improved accuracy and local language understanding
  • Enable open contributions from users, drivers, and local unions
  • Collaborate with transport authorities for live station data integration
  • Expand to other cities and become the go-to platform for trotro information in West Africa
Share this project:

Updates