Durable Execution Agents
Web page · https://martinfowler.com/articles/durable-execution-agents.html
First seen by Oriel on .
AI summary
The article argues that durable execution makes long-horizon agents reliable by logging external steps, replaying programs after failures, and rebuilding state without repeating completed work.
- Stock Replenisher demonstrates an agent coordinating quotes, approval, ordering, and delivery across days or weeks while handling changing prices, replies, and timeouts.
- Its agentic loop combines a goal, tools, state, and a model, while coded allowed checks enforce business rules such as requiring approval before ordering.
- Durable runtimes persist every external step’s inputs and outputs; after crashes, recorded steps return logged results and local code replays to reconstruct state.
- The Temporal implementation uses @workflow.run, workflow.execute_activity, and workflow.wait_condition; Cloudflare Workflows provides corresponding step.do and step.waitForEvent primitives.
- Long-running agents face distributed-system failures including crashes, failed calls, duplicate effects, concurrency, changing conditions, code deployments, growing history, and untrusted input.
#durable execution #llm agents #distributed systems #temporal
Generated by gpt-5.6-sol. Check the original for details.
Metric trends
Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained
Points
Score reported by the source.
- Hacker News: 4, latest observation ; +1 over observed period since
Comments
Comment count reported by the source.
- Hacker News: 0, latest observation ; 0 over observed period since
Aliases
| Source | External ID | URL |
|---|---|---|
| Hacker News | 50462904 | https://martinfowler.com/articles/durable-execution-agents.html Read discussion |