Use run data for traceability now and analytics later.

Report storage and indexing

Reports are written as JSON plus HTML, organized into a date-based folder tree. An append-only index.ndjson file at the root lets the interface load large histories instantly without scanning sub-folders.

{reports_path}/
  index.ndjson
  2026/
    04/
      W14/
        AMP-100-SN001-PASS-20260405-180358-abc12345.json
        AMP-100-SN001-PASS-20260405-180358-abc12345.html
  archive/
    2025/
      ...

Report file naming

Every report filename encodes the most important fields at a glance:

{part_number}-{serial}-{RESULT}-{YYYYMMDD}-{HHMMSS}-{uuid8}.json

Example: AMP-100-SN00412-FAIL-20260405-180358-3fdc76ec.json

The report_id field inside the JSON body matches the filename stem and is the stable identifier used across the index and the viewer.

Result Viewer

The Result Viewer is at the Results tab of the main navigation.

If the index file is lost or corrupted, use Rebuild index in the Config panel to regenerate it by scanning the report folders.

Storage configuration

Report storage is configured in config/system_config.json:

{
  "reports": {
    "path": "data/reports",
    "archive_after_days": 90,
    "max_hot_reports": 10000
  }
}

To change the path: Config → Report Storage → Browse or type a path → Save. Takes effect on the next engine start. Existing reports are not moved — run Rebuild index after copying files.

GRR study modes

GRR is a first-class operating mode in Corvus. It runs a structured measurement study to quantify how much observed variation comes from the gauge itself versus the parts being measured.

The hardware qualification profile is activated with:

{
  "mode": "single_op",
  "part_handling": "static",
  "auto_advance": true,
  "skip_prompts": true
}

Part handling modes

%GRR status thresholds (AIAG MSA 4th Ed.)

Study lifecycle

1. Create — configure the study in the GRR screen (parts, trials, operators, sequence, capture steps, tolerance spec).
2. Plan generated — Corvus writes a randomized execution matrix (Fisher-Yates shuffle with consecutive-part avoidance).
3. Start — the study runner fires measurement triggers in plan order, prompting the operator to load or change parts as needed.
4. Data collected — each completed run is appended to results.jsonl immediately. A crash mid-study can be resumed without data loss.
5. Analysis — when all runs are complete, Corvus runs the ANOVA computation and writes report.json.
6. Result panel — the study result shows %GRR, NDC, P/T ratio, and capability indices side-by-side.

Study files live at data/grr_studies/{study_id}/:

Study configuration reference

Capability and measurement concepts

• %GRR — gauge variation as a percentage of total variation.
• NDC — Number of Distinct Categories. ≥ 5 is generally required for a capable measurement system.
• P/T ratio — GRR span relative to the tolerance window. Target < 10%.
• Cpk — short-term process capability using part variation with measurement error excluded.
• Ppk — long-term capability using total variation, including both process and gauge effects.

Capability status thresholds

Interpreting Cpk vs Ppk