Torba

Know what's being sprayed. Torba checks pesticide risks, bans, and safer options. It also helps understand H and P phrases and precautions. Built for the tools people already have.

Comment

Inspiration

Every year, the Moroccan Poison Control Center records hundreds of pesticide exposure cases. Behind each case is a farmer, a family member, or a community worker facing serious health risks. One of the main reasons is simple: safety information isn't accessible when and where it's needed. Labels are in technical French or English, and most farmers don't have reliable internet access to look up safety data. Torba was built to solve this problem.

What it does

Torba is an offline, easy-to-use tool that puts verified pesticide safety data directly into farmers' hands. Farmers can search any pesticide by name or brand and instantly get:

  • Risk level (HIGH, MEDIUM, LOW)
  • Health effects
  • Practical safety tips
  • Safer alternatives
  • Ban status in Morocco The app also includes a soil color checker based on INRA research, helping farmers assess soil health and make better agricultural decisions. Torba works entirely offline and runs on my laptop that 100% is within the considered public.

How we built it

  • Python for the backend logic
  • Streamlit for the web interface
  • SQLite for the local database
  • Pandas for data processing
  • Custom translation system for Arabic and French

The app is designed to run on the ADTC Standard Laptop with only 8GB RAM. It uses approximately 155 MB of RAM, well under the 7 GB limit.

Challenges we ran into

  • Making the app truly multilingual: I added Arabic and French translations directly into the database so that content, not just interface labels, appears in the farmer's chosen language.
  • Ensuring offline functionality: All data is stored locally with no external API calls.
  • Optimising for low-memory devices: The app uses only 155 MB of RAM. -I experimented with TinyLlama and Llama 3.2 for conversational features, but the models were too heavy for my network and PC. They also didn't fit the ADTC hardware criteria. So I decided to focus on what actually works: a lightweight, reliable database-driven tool that doesn't depend on large models.

Accomplishments that we're proud of

This is the first big project I've built that actually means something for my community. Seeing it work in three languages and knowing it could help a farmer make a safer decision is what I'm most proud of.

  • A fully working offline app that farmers can use without internet access.
  • Full support for English, Arabic, and French.
  • Translation of pesticide safety data into Arabic and French.
  • Soil color checker based on verified INRA research.
  • Runs with minimal resources.

What we learned

I learned that building for real users means making trade-offs. A working, offline, multilingual tool that fits the hardware is more useful than a flashy AI feature that fails in the field. I also learned that translations matter beyond just the interface. Farmers need the actual content in their language, not just the buttons.

What's next for Torba

  • Expand the pesticide database with more products and translations
  • Add offline voice input for farmers who cannot read
  • Include crop-specific recommendations
  • Add weather data for safe application timing
  • Integrate with local agricultural extension services

Built With

  • agriculture
  • arabic
  • english
  • french
  • multilingual
  • offline-first
  • pandas
  • pesticide
  • python
  • safety
  • sqlite
  • streamlit
Share this project:

Updates

Submission history