Want to talk about data?
Data science, career questions, practical machine learning, and more?
Join @DataTalksClub! The link is in the profile description.
Learn more about our activities in the thread! đź§µ
Agent demos often stop at the agent loop:
Plan.
Call tools.
Return an answer.
Production needs the layer around it.
Workshop with Nicholas Lotz: luma.com/wz20rm8n
1/5
ALT Workshop checklist for managing durable agents in production, outlining workflow persistence, failure recovery, and event details.
Most agent frameworks wrap the same loop:
Call model -> run requested tools -> append results -> call model again.
Stop when the model response has no tool calls left.
That is the basic agentic loop.
Agent demos often stop at the agent loop:
Plan.
Call tools.
Return an answer.
Production needs the layer around it.
Workshop with Nicholas Lotz: luma.com/wz20rm8n
1/5
ALT Workshop checklist for managing durable agents in production, outlining workflow persistence, failure recovery, and event details.
In the workshop, Nicholas will cover:
- Durable workflow state
- Resuming after failures
- Retrying tool calls
- Human approval steps
- Execution traces
- Deployment patterns
4/5
Don't miss our upcoming live free webinar on The Agentic Shift in Data Engineering: From Pipeline Builders to Architects of Controlled Chaos with Nik B. Get ready to learn new techniques and get your questions answered by an expert in the field.
👉 luma.com/hpuirqjd
Your agent calls a tool to create a support ticket.
The request times out.
Should it retry?
If the ticket was already created, the retry can create a duplicate.
Workshop: luma.com/wz20rm8n
1/5
ALT A document discussing tool calling in production environments, outlining ticket creation and retry issues, with a workshop invitation.
Nicholas Lotz will cover how to handle this in production agent workflows:
- Retry policies
- Tool call history
- Timeouts
- Recovery paths
- Execution traces
4/5