Phase Atlas Conditioning Quickstart

Run PAC on a sample event set using CLI or Python. Phase Atlas is the measurement surface; PAC is the conditioning layer applied to external event outputs. Capital Impact can then translate baseline vs conditioned results into capital-relevant tail metrics under the same event set.

Before you run

  • Use a frozen Phase Atlas bundle with its manifest and hash references
  • Prepare a sample event set with stable event identifiers
  • Select a declared PAC mode from the release manifest

CLI flow (illustrative)

# 1) Resolve a frozen Phase Atlas release (read-only)
ptem phase-atlas resolve   --freeze-tag freeze-2026-01-27   --out manifests/phase_atlas_manifest.json

# 2) Apply PAC conditioning to an external event set
ptem pac run   --phase-atlas-manifest manifests/phase_atlas_manifest.json   --events examples/sample_events.csv   --mode declared_default   --out out/conditioned_events.csv   --receipt out/conditioning_receipt.json

Python flow (illustrative)

from ptem import pac

result = pac.run(
    phase_atlas_manifest="manifests/phase_atlas_manifest.json",
    events_csv="examples/sample_events.csv",
    mode="declared_default",
)

result.write_outputs(
    conditioned_csv="out/conditioned_events.csv",
    receipt_json="out/conditioning_receipt.json",
)

Output contract

  • Conditioned portfolio outputs
  • Manifest-backed run metadata
  • Deterministic conditioning receipt for replay and audit
  • Optional capital_impact_report.json and one-page capital impact sheet

Boundary conditions

  • Non-forecasting outputs
  • No track modification
  • No event generation
  • No write-access to Phase Atlas source artifacts