Tool: verify_thoughts

Purpose

Evaluate claims against supplied evidence and produce claim-evidence status. This is the evidence core of Test mode.

Current implementation mode: heuristic. The server classifies the evidence records supplied by the caller; it does not retrieve external sources or fact check claims by itself.

Input

  • deliberation_id
  • node_ids: string[]
  • critical_claims: string[] (legacy convenience path)
  • claims (preferred):
    • text: string
    • critical: bool
    • evidence: EvidenceItem[]
      • source: string
      • tier: tier1 | tier2 | tier3
      • independence_group: string
      • supports: bool
      • contradictory: bool
      • unambiguous: bool
  • method: cove | triangulation | hybrid
  • policy_override (optional):
    • min_independent_groups: integer (1-8)
    • allow_tier3_for_independence: boolean
    • require_tier1_unambiguous_for_critical: boolean
    • fail_closed_on_critical_unresolved: boolean

Output

  • verification_matrix
  • claim_status_summary
  • revisions
  • analysis: provenance and limitations for the verification run

Behavior

  • CoVe/triangulation method names select the classification route, but current execution remains caller-evidence based.
  • Verification policy:
    • verified if at least one unambiguous Tier-1 supporting source exists and no contradiction.
    • verified if there are >= 3 independent supporting groups from Tier-1/Tier-2 evidence and no contradiction.
    • source_conflict if any contradictory evidence exists.
    • data_deficit otherwise.
  • When no evidence is supplied, the server marks claims with internal heuristic evidence and normally returns data_deficit for strict policies.

Errors

  • Missing critical claims
  • External evidence collection is unavailable in the current build.