{
  "execution": "Actual offline Python retrieval tests",
  "documents_loaded": 4,
  "test_questions": 5,
  "tests_passed": 5,
  "status_counts": {
    "evidence_found": 3,
    "conflict": 1,
    "unknown": 1
  },
  "abstentions": 2,
  "model_calls": 0,
  "disclosure": "These are fixed functional test cases, not a general accuracy benchmark. No natural-language answer generation was run.",
  "results": [
    {
      "test_id": "Q1",
      "test_passed": true,
      "question": "How do I invite a teammate?",
      "method": "Deterministic term overlap over manually curated fact metadata",
      "generation": "not_run",
      "ranking": [
        {
          "fact_key": "team.invite_steps",
          "score": 2,
          "matched_terms": [
            "invite",
            "teammate"
          ]
        }
      ],
      "evidence": [
        {
          "id": "KB-001",
          "title": "Team invitations",
          "approved_on": "2026-08-20",
          "status": "approved",
          "path": "docs/KB-001-team-invitations.md",
          "fact_key": "team.invite_steps",
          "quote": "To invite a teammate in the fictional Demo Workspace, open Settings, choose Team, select Invite, and enter the teammate's email address. Only workspace administrators can send invitations."
        }
      ],
      "fact_key": "team.invite_steps",
      "status": "evidence_found",
      "abstained": false,
      "message": "Relevant approved excerpt retrieved. No generated answer or interpretation is included."
    },
    {
      "test_id": "Q2",
      "test_passed": true,
      "question": "When is email support available?",
      "method": "Deterministic term overlap over manually curated fact metadata",
      "generation": "not_run",
      "ranking": [
        {
          "fact_key": "support.hours",
          "score": 3,
          "matched_terms": [
            "available",
            "email",
            "support"
          ]
        }
      ],
      "evidence": [
        {
          "id": "KB-002",
          "title": "Email support hours",
          "approved_on": "2026-08-26",
          "status": "approved",
          "path": "docs/KB-002-support-hours.md",
          "fact_key": "support.hours",
          "quote": "For the fictional Demo Workspace, email support is staffed Monday to Friday, 09:00–17:00 UTC, excluding published holidays. Requests received outside those hours enter the next staffed queue. These hours do not promise a resolution or response time."
        }
      ],
      "fact_key": "support.hours",
      "status": "evidence_found",
      "abstained": false,
      "message": "Relevant approved excerpt retrieved. No generated answer or interpretation is included."
    },
    {
      "test_id": "Q3",
      "test_passed": true,
      "question": "Are monthly invoices due in 14 or 30 days?",
      "method": "Deterministic term overlap over manually curated fact metadata",
      "generation": "not_run",
      "ranking": [
        {
          "fact_key": "billing.invoice_due",
          "score": 4,
          "matched_terms": [
            "days",
            "due",
            "invoice",
            "monthly"
          ]
        }
      ],
      "evidence": [
        {
          "id": "KB-003",
          "title": "Billing policy",
          "approved_on": "2026-08-27",
          "status": "approved",
          "path": "docs/KB-003-billing-policy.md",
          "fact_key": "billing.invoice_due",
          "quote": "For the fictional Demo Workspace, monthly invoices are due 30 calendar days after the issue date."
        },
        {
          "id": "KB-004",
          "title": "Finance operations guide",
          "approved_on": "2026-08-28",
          "status": "approved",
          "path": "docs/KB-004-finance-guide.md",
          "fact_key": "billing.invoice_due",
          "quote": "For the fictional Demo Workspace, monthly invoices are due 14 calendar days after the issue date."
        }
      ],
      "fact_key": "billing.invoice_due",
      "status": "conflict",
      "abstained": true,
      "message": "Approved sources disagree for this fact. Ask the document owners to resolve the conflict; the newer date alone does not supersede a source."
    },
    {
      "test_id": "Q4",
      "test_passed": true,
      "question": "Where is customer data hosted?",
      "method": "Deterministic term overlap over manually curated fact metadata",
      "generation": "not_run",
      "ranking": [],
      "evidence": [],
      "status": "unknown",
      "abstained": true,
      "message": "No matching approved fact was found. No answer can be supported by this small corpus."
    },
    {
      "test_id": "Q5",
      "test_passed": true,
      "question": "Can I get support on Saturday?",
      "method": "Deterministic term overlap over manually curated fact metadata",
      "generation": "not_run",
      "ranking": [
        {
          "fact_key": "support.hours",
          "score": 2,
          "matched_terms": [
            "saturday",
            "support"
          ]
        }
      ],
      "evidence": [
        {
          "id": "KB-002",
          "title": "Email support hours",
          "approved_on": "2026-08-26",
          "status": "approved",
          "path": "docs/KB-002-support-hours.md",
          "fact_key": "support.hours",
          "quote": "For the fictional Demo Workspace, email support is staffed Monday to Friday, 09:00–17:00 UTC, excluding published holidays. Requests received outside those hours enter the next staffed queue. These hours do not promise a resolution or response time."
        }
      ],
      "fact_key": "support.hours",
      "status": "evidence_found",
      "abstained": false,
      "message": "Relevant approved excerpt retrieved. No generated answer or interpretation is included."
    }
  ]
}