agent-run-schema

What problem does it solve?

Helps you instrument agent executions by providing a canonical 10-event observability schema that uses a shared agent_run_id to join all run events into a single timeline.

Core Features & Use Cases

• Defines 10 canonical events for agent runs (agent_run_started, task_started, task_completed, user_correction_submitted, approval_granted, approval_denied, tool_call_failed, memory_miss, run_completed, run_failed) with a shared agent_run_id.
• Supports building unified dashboards and sinks (SQLite, Postgres, log streams) to reconstruct run timelines.
• Useful for debugging, monitoring observability, and evaluating agent behavior across tools and memory.

Quick Start

Integrate the reference schema into your agent by emitting the defined 10 events with a shared agent_run_id for each run.