Results, Evidence, and Export¶
简体中文 · Previous: Orbit correction · Next: Import and migration
Studio result pages are read-only projections of local evidence, not a second compute authority.
Every refresh rereads canonical bytes, verifies hashes, parses the schema, and checks project and
Engine identity. Partial, modified, or crossed records are withheld as UNVERIFIED.
Screening Results center¶
Results → Screening Results unifies v0.2–v0.5 radiation slices, equilibrium, matched distribution, and LMA. Grouping is:
CURRENT: project ID, content, and physics revision match the open project.STALE: same project ID but different content/physics; readable historical evidence.INCOMPATIBLE: no comparable open project or a different project.UNVERIFIED: necessary route/artifact identity is incomplete and withheld.
Filter by type, double-click to open, copy a full evidence value, or select two results for comparison.
Comparison rules¶
Only two artifacts of the same type can be compared. Project/schema mismatch rejects comparison. Revision, grid, or physical-semantic mismatch permits side-by-side review only. Absolute differences appear only after every gate passes. Studio does not interpolate, invent percentages, or merge positive and negative LMA branches.
A trustworthy A/B comparison should also hold Engine source, target, seed, profile, warnings, and qualification boundary fixed.
Specialized stores¶
- v0.8 nonlinear results reopen within Nonlinear Dynamics;
- v0.9 orbit results reopen within its workspace;
- v0.7 Touschek formal evidence appears in the corrected Screening Study panel;
- v0.6 exists only in its quarantined historical view;
- v0.1 tracking opens/exports from Tasks.
Stores are isolated; renaming a job or copying files cannot promote evidence across them.
Revision index and portable packing¶
Formal stores still reverify their own schemas, while Studio also publishes exact project
snapshots and result bytes into a SHA-256 content-addressed object store. .vmaproj keeps only
revision, calculation-input dependency, and object references, so lattice edits preserve old
results without making the normal working file grow with DA/FMA, tracking, or particle arrays.
The calculation-input digest covers an exact dependency ledger: project snapshot, upstream
artifacts, configuration/profile or durable exact Submit request. Only a fully equal digest is a
reuse candidate. Results belonging to a different physics revision remain visible as STALE.
Use File → Pack Project with Indexed Results… when results must travel with the project. This creates a separate self-contained package and leaves the lightweight working file unchanged. Opening a portable package verifies and restores every embedded object; the matching specialized store must still pass its schema and Engine-identity gates before physics promotion.
What to check first in Evidence¶
- Qualification/Validity and
machinePrediction/designQualified. - Project ID, content SHA, physics revision.
- Engine version/build/source and capability digest.
- Input/output artifact bindings.
- Warnings, checks, convergence.
- Assumptions and applicability.
- Headline numerical values.
Target route is local Studio metadata, not signed Engine identity.
Export surfaces¶
| Page | Export |
|---|---|
| Lattice Geometry | PDF, SVG, PNG |
| Nonlinear plots | PNG, SVG |
| Nonlinear Evidence | CSV, canonical JSON |
| Tracking Tasks | artifact JSON, evidence manifest, checksum sidecar |
| Project | lightweight .vmaproj save, canonical JSON export, portable indexed-result package |
| Result trees | copy one complete evidence value |
Keep sidecars and hashes with exported data. A screenshot alone is not a traceable calculation record. Refresh in Screening Results performs local verification; it does not rerun calculations.
Physical element/order/parameter/RF changes and Cell copy create a new physics revision. Old results remain historical but are not default prerequisites. Layout-only Cell rename/ungroup does not stale physics.