CAG keyword/entity search

Execute CAG (Cached Augmented Generation) search over cached facts.

How it works:

1. Tokenizes query into keywords and entities
2. Looks up matching facts in cache (Constitution + High-Certainty)
3. Scores by keyword overlap, entity overlap, and POT Score
4. Returns results with confidence metadata

Scoring formula: score = (keyword_score * 0.5) + (entity_score * 0.3) + (pot_score * 0.2)

Confidence suggestions:

• sufficient: Good matches found, likely complete answer
• consider_rag: Partial matches, client may want to also query RAG
• no_match: No meaningful matches, definitely query RAG

Note: Client decides whether to follow up with RAG based on suggestion.

{
  "include_constitution": true,
  "min_pot_score": 0,
  "query": "What is POT Index?",
  "top_k": 10
}

{
  "confidence": {
    "match_quality": 0.92,
    "results_count": 1,
    "suggestion": "sufficient",
    "top_score": 0.92
  },
  "data": {
    "results": [
      {
        "content": "POT Index assigns certainty scores 0-1",
        "domain": "core",
        "fact_id": "const_001",
        "level": "CONSTITUTION",
        "match_details": {
          "entity_score": 1,
          "keyword_score": 0.8,
          "matched_entities": [
            "POT Index"
          ],
          "matched_keywords": [
            "pot",
            "index"
          ]
        },
        "pot_score": 1,
        "score": 0.92,
        "tags": [
          "pot-index"
        ]
      }
    ],
    "total_results": 1
  },
  "metadata": {
    "cache_status": "hit",
    "latency_ms": 12,
    "pot_id": "scipot-core",
    "query": "What is POT Index?",
    "retrieval_strategy": "CAG"
  }
}