An overview of the Agdir Satellite API architecture, including the two-queue processing system and how analysis results are generated from satellite imagery.
The Agdir platform uses a two-queue architecture to process satellite analysis requests:
The first queue handles satellite data acquisition. When an analysis is requested, the system checks whether the required satellite imagery is already available in the Agdir data store. If not, it is downloaded from satellite data providers (primarily Copernicus/Sentinel). This step corresponds to the SentinelQueue and SentinelConsumer statuses.
Once satellite data is available, the request moves to the processing queue where the actual analysis is performed. This includes index calculation, ML model inference, image generation, and statistical computation. This corresponds to the NonSentinelQueue and NonSentinelConsumer statuses.
Every analysis in the Agdir API is a request. You submit a request (mutation), the system processes it asynchronously, and you retrieve the results (query) when processing is complete. This design allows the system to handle long-running analyses without blocking your API calls.
Key characteristics:
Results are delivered as images showing intrafield variation over time. For each cloud-free satellite pass during your date range, the system generates:
New satellite data becomes available as satellites pass over your field location. Sentinel-2 satellites have a revisit frequency of 2–7 days depending on latitude. However, not every pass produces usable data — cloud cover can block the satellite's view of the ground.
This means new analysis results can be generated frequently, but you should not expect data for every single day. For continuous daily coverage, use daily vegetation estimates.