Loading workspace…
AI intelligence · CHMv2 first
Execute a scientific workflow from real predicted canopy height. Vegetation evidence is optional.
Meta/WRI CHMv2 · source date 1 March 2018 · predicted height, not LiDAR ground truth. The fixed replay question does not pretend to answer arbitrary new questions with a recorded model call.
Stages perform real source validation and computation. Recorded context is attributed to its original run; no animation stands in for live agent work.
Planner → analyst → critic → strategist → synthesizer. Primary source is mandatory; optional NDVI failure is a recorded omission.
Deterministic analysisNative scientific measurements, derived fractions, citation checks and report validation execute now.
Recorded Muse contextOnly the CHM-cited finding is selected from the earlier Baseten response. It is not a recorded five-agent run.
Local live · disabled by defaultSeparate authenticated developer CLI uses existing Meta/Baseten adapters with explicit opt-in and bounded run budgets. Public production cannot invoke it.
Protected production demonstration remains unavailable until approved spending safeguards exist. No operational flight guidance or crop/yield classification is provided.
AI intelligence
Inspect the scientific workflow. See what is implemented before you run it.
The Baseten research agent validates structured findings, selected citations and provenance. Protected development access is required.
Guardrails
Explicit evidence selection · source confidence retained · no invented citations · no unlabelled synthetic outputs.
Research results include their real duration, model ID and token usage. Other agent stages have no backend execution history.