Ask HN: How are you preventing runaway LLM workflows in production?

Re:Ask HN: How are you preventing runaway LLM workflows in p... by guerython on Monday March 02, 2026 @07:05PM (#47226095)

Plan doc now drives everything. Every request becomes a four-bullet plan (inputs, connectors, guards, metric) before any workflow runs so the agent has a concrete target instead of just questions.

Each connector includes a watcher expecting {status:'ok'}, logs the sessionId, and enforces guard thresholds: retries>3/30s, loop depth>4, or cost>2x baseline. When a guard trips we pause the plan, stream the watcher log to the manual gate, and only let the next run continue after a human approves the diff. That keeps automation fast but keeps the ops team in control.

[ Parent | on HN ]

Re:Ask HN: How are you preventing runaway LLM workflows in p... by HenryM12 on Tuesday March 03, 2026 @06:23PM (#47240566)

Interesting approach. Once you built the plan/watch system, did it mostly hold up, or have you had to keep revisiting it as workflows started spanning more tools and steps? Roughly how long did it take to implement?

[ Parent | on HN ]