VKC Visa Assessment Engine (P1)

Goal

Provide “비자변경 가능성(%)” outputs at scale (cover target-customer visas) while keeping maintenance data-driven:

• rules/weights live in DB (versioned + effective dates)
• code is a stable evaluator + storage + admin activation workflow

Non‑negotiable

• No hardcoding rules in TypeScript. Rules belong to DB rows as JSON.

Core data model (minimum)

• visa_catalog: visa codes / labels / categories
• visa_transition_rules: fromVisa→toVisa ruleset JSON + version + effectiveFrom + status(pending|active)
• visa_assessment_models: score→percent mapping / weights JSON + version + status
• visa_assessments: user assessment history (probabilityPercent, grade, missing, risks, createdAt)

Required interfaces

• Ruleset JSON schema (validate before activation)
  • .codex/skills/vkc-visa-assessment-engine/references/ruleset-schema.json
• API response schema (persist + return)
  • .codex/skills/vkc-visa-assessment-engine/references/response-schema.json

API + admin workflow (recommended)

• User:
  • POST /api/visa/assess (auth + 1/day + save visa_assessments)
• Admin:
  • Manage visa_catalog
  • Create/update visa_transition_rules (pending)
  • Activate a ruleset version (switch to active)
  • Manage visa_assessment_models version and activation

STEP3 visa registry (SoT)

• “전체 커버” visa code registry lives in: docs/STEP3_SOT_RESOURCES.md
• Implementation rule: registry is complete first → rulesets are filled in iteratively (v1→vN)