Skip to main content

Example prompts

Give Clarity AI a concrete objective, identify the relevant workspace or entity, and state constraints such as wavelengths, classes, dates, or required output.

Explore and analyze data​

  • Summarize the sources and analysis layers in this workspace. Flag missing metadata or compatibility constraints.
  • Generate PCA for this image using 400-900 nm. Report stored explained-variance ratios if they are available; otherwise state that variance metadata is unavailable and describe only the component score maps.
  • Compare the selected region with reference spectra and report the closest matches with limitations.
  • Run a matched filter for this target spectrum and explain how the background choice affects interpretation.
  • Stage an operation chain that thresholds this layer, preview it, and wait for my approval before committing it.

Find and organize data​

  • Search Planetary Computer for imagery over this workspace extent between these dates with cloud cover below 20%.
  • Import this local data into the current workspace and tell me when ingestion finishes.
  • Find images compatible with this dataset version and explain the compatibility criteria.

Work with labels, datasets, and models​

  • Import this GeoJSON as labels on the selected image, preserve its feature attributes, and report when the import finishes.
  • Create a draft dataset version from these compatible images, apply the specified wavelengths, and leave it unfinalized for review.
  • Analyze this dataset version and identify class imbalance or preprocessing concerns.
  • Create a model version from this dataset, apply these training settings, and ask before starting training.
  • Analyze this model version's metrics and prepare a model report.

Conduct research​

  • Conduct deep research on whether this sensor and wavelength range can distinguish the target materials. Treat unvalidated methods as exploratory.
  • Review the literature for validated approaches to this question, form competing hypotheses, and stop after the methodology-validation phase for approval.