The 72-hour runbook we use for new client deploys. Hour-by-hour, with the eval gates, rollback triggers, and the Slack handoff template.
We have shipped enough client agent deploys that we run them off a single 72-hour runbook. Below: the full runbook, the eval gates between phases, and the rollback triggers. The stack we actually use Below is the production stack we ship for this category. This one ships in 72 hours and survives a real-world Monday morning. Pipecat for voice, a thin Express service for chat, deployed on Fly.io or Render.
This one ships in 72 hours and survives a real-world Monday morning. Pipecat for voice, a thin Express service for chat, deployed on Fly.io or Render. We avoid Vercel for anything stateful. Default to Claude 3.7 Haiku for the chat layer and Claude 3.7 Sonnet for the planning and tool-use steps. Fall back to GPT-4.1-mini when Anthropic has capacity pressure.
The 72-hour runbook we use for new client deploys. Hour-by-hour, with the eval gates, rollback triggers, and the Slack handoff template. It is filed under AI Infrastructure because that is where operators looking for this problem actually start, and it is written from production work rather than from a content calendar.