Skip to main content
These three tools let an agent that runs here record its run on a Nanonets Health custom agent, so that agent’s dashboard in Nanonets Health shows the run and each item’s outcome. They are for agents migrated from Nanonets Health custom agents.

​
Authentication and enablement

The tools use the Nanonets Health integration. The platform injects api_key. That key is never part of the model-visible schema, and it scopes every call to the connected Nanonets Health organization. All three tools are disabled by default. Every tool takes agent_id, the Nanonets Health custom-agent id (UUID) whose dashboard the run belongs to. Put it in the agent’s instructions; it cannot be discovered.

​
Flow

  1. Call nanonetshealth_start_agent_run once at the start of the run and keep its run_log_id.
  2. Call nanonetshealth_report_agent_item once for every item, including skips and failures.
  3. Call nanonetshealth_complete_agent_run once, after the last item.
All three are writes to the Nanonets Health database.

​
nanonetshealth_start_agent_run

Inputs:
  • agent_id (required)
  • external_run_id: optional. This platform’s own run or task id, recorded on the run log for cross-reference.
Output:

​
nanonetshealth_report_agent_item

Inputs:
  • agent_id, run_log_id, item_key, status (all required)
  • patient_id, visit_id, appointment_date, reason, decision, error: optional strings
  • item, outcome: optional JSON objects. A string holding a JSON object is also accepted.
status is one of updated, mips_found, completed, skipped, needs_va, accepted, check_required, declined, dry_run, scored, reviewed, failure, failed. Nanonets Health stores failed as failure. Idempotent per run_log_id + item_key: reporting the same key again replaces the earlier row, so a retry never duplicates an item. Output:

​
nanonetshealth_complete_agent_run

Inputs:
  • agent_id, run_log_id (required)
  • status: optional. One of success, partial_success, failure. If you omit it, Nanonets Health derives it from the reported items:
    • failure if every item failed
    • partial_success if some items failed
    • success otherwise
  • output: optional JSON object, such as counts per status.
Output:

​
Expected errors

The tools return Nanonets Health’s own error text: These tools require the Nanonets Health release that adds external runs (fleming PR #467).