Analyze
Upload a zip of your local campaign folder — boltz_input.md +
boltz_yamls/ (including .boltzmaker_manifest.json) +
boltz_output/ with the completed predictions — and get back a
summary CSV/XLSX and an interactive dashboard. Real GPU inference (run)
isn't run here; this expects predictions you've already computed elsewhere. If any
family sets Apo structure:, also include that referenced file (e.g. its
reference/ folder) so compare-sse has something to compare against —
otherwise it's skipped for that family, not an error.