Passport output

A passport is the explainability layer around extraction. It helps answer:

• Which dictionary snapshot was used?
• Which aliases matched?
• Which canonical values were accepted?
• Which values were filtered out?
• Which warnings should a reviewer see?

Example shape

{
  "snapshot": {
    "version": "infra_incidents@2026-05-11-v1",
    "source": "file"
  },
  "alias_matcher_backend": "aho_corasick",
  "accepted": [
    {
      "slot": "TOOL",
      "value": "kubernetes",
      "source": "alias",
      "matched_text": "k8s"
    }
  ],
  "filtered_out": [],
  "warnings": []
}

Why it matters

For internal search and terminology governance, it is not enough to return a score. Teams need to review why a term was found, whether it came from a trusted alias, and whether a proposed dictionary change would make search noisier.

Design direction

SkeinRank should keep passport output bounded and useful:

• enough detail for debugging;
• stable enough for tests;
• compact enough for UI review;
• explicit about filters and warnings.