AI Agents / Agent state and durable workflows
AI Agents: Run identity
Create a stable run identifier and use it across tools, checkpoints and artifacts.
Environment and inputs: Durable agent workflows; LangGraph illustrates checkpoints and review interruptions.
When to use it
A restarted process must distinguish unfinished work from operations that already committed. Design state transitions together with external operation receipts, and use a disposable backend to exercise recovery before relying on it for real records.
Practical example
Illustrative durable state envelope. Store it using your framework or database transaction strategy.
{
"schema_version": 1,
"run_id": "demo-run-001",
"revision": 3,
"status": "waiting_for_tool",
"pending_operation_id": "lookup-002",
"artifacts": []
}Check the result
Resume after a worker restart and verify every new event retains the original run identity.
Save and compare results
Save the test input, configuration and result. Fix the failing boundary, repeat the same test and compare it with the saved working configuration.
Scope and limits
These exercises have not been tested on your systems. Check feature support and configuration names against the installed release and linked documentation.