Tool: reasoning_autopilot

Purpose

Run Auto mode: a safe default deep-reasoning flow from natural-language intent.

User promise: tell ReasonKit what you need; it chooses the right reasoning path.

For new clients, prefer run_thinking_mode with mode: "Auto". This tool remains available as the compatibility Auto workflow.

Input

• intent
• optional: session_id, quick_session_id, goal, constraints, profile
• optional: expand_count, beam_width
• optional: policy_override

Output

• user_mode
• internal_method
• mode_rationale
• safety_notes
• estimated_cost
• mode_contract
• deliberation_id
• routed intent plan
• executed_tools
• verification and routing summary (including policy_blocked flag if consensus is blocked by unresolved critical assumptions/claims)
• audit id

Behavior

• Routes intent first, then executes:
  • start_deliberation
  • expand_thoughts
  • score_thoughts
  • prune_thoughts
  • verify_thoughts
  • run_reasonkit_pipeline
  • consensus_answer
  • export_reasoning_audit
• Returns auditable outputs and policy-blocked signal.
• Labels heuristic outputs and does not invent evidence or semantic confidence.