Noza Agricultural Intelligence

Ingabe turns drone images, satellite data and weather into plain answers for the people who decide on Rwanda's fields: farmers, insurers, agronomists and agricultural scientists. You upload a drone flight or a field boundary and ask Sage, Ingabe's assistant, what you want to know. Sage finds the right data, puts it on the map, and answers in words, for your role.

What it does

• Drone images first. Upload an orthophoto and Sage posts a first look (greenness and crop cover). Ask for more and it finds stressed patches, compares two flights, and scores an insurance trigger from NDVI flights.
• Vegetation, drought and crop risk. NDVI and other Sentinel-2 indices for any district, sector or cell; drought, vegetation alerts, crop growth stage and yield-risk trends per district.
• Weather and insurance. Past rainfall (CHIRPS, AgERA5), a four-model 16-day forecast, dry spells, and season reports with trigger status for insurers.
• Soil and water. iSDAsoil properties, FAO WaPOR soil moisture and evapotranspiration, and Sentinel-1 radar through clouds (NDVI estimates, flood extent).
• Rwanda built in. All districts, sectors, cells and villages, with outlines that follow the zoom.
• A memory for your organization. The Brain stores reports, policies and field records, so Sage can read and quote them.

Data sources and update times are in the training manual.

Running it locally

Ingabe runs on your own machine with Docker Compose.

1. Install Docker and git, then clone this repository.
2. Copy the settings template and fill it in:

   cp .env.example .env
   

   • Sage needs an LLM key: OPENAI_API_KEY with OPENAI_BASE_URL (OpenRouter by default).
   • Sign-in uses WorkOS: the WORKOS_* settings.
3. Build, start and check everything:

   scripts/deploy.sh
   

4. Open http://localhost:8000.

scripts/deploy.sh --check-only re-checks a running stack without rebuilding it. Agents and contributors: start with AGENTS.md and CODING_STANDARDS.md.

Documentation

• Training manual: using Ingabe and Sage, for every role.
• Local runtime architecture

Security

Please don't report security problems in a public issue. Contact the maintainer privately through GitHub.

License and origin

Ingabe is licensed under the GNU Affero General Public License v3. It began as a fork of Mundi by Bunting Labs, also AGPL-3.0, and has since been extended and adapted for agriculture in Rwanda.

Developed by NozaLabs.