5:00 ÖS - 6:00 ÖS (UTC)

Long-running processes fail partway through all the time, whether from a pod restart, an API timeout or a deploy that lands mid-execution. Durable execution means your code survives those failures and resumes exactly where it stopped. This session builds a durable workflow in .NET with Dapr Workflow, runs and debugs it locally with Aspire and the Dapr Dev Dashboard, then moves it to Catalyst's managed durable execution engine with one change to the AppHost. The Catalyst workflow visualizer is used to inspect workflow executions and manage workflows.

7:00 ÖS - 8:00 ÖS (UTC)

Agents are long running, stateful and expensive. When one crashes halfway through a multi-step task, you lose its progress, its memory and the tokens you've already spent. This session shows how durable execution makes Microsoft Agent Framework agents production-ready with Dapr and Catalyst: every step is checkpointed, memory survives restarts, and agents communicate over pub/sub. We then deploy to AKS to show why AKS plus Catalyst is a strong platform for durable agents on Kubernetes. [eventID:27766]

7:00 ÖS - 8:00 ÖS (UTC)

Microsoft Foundry gives you hosting, identity and models for your agents. This session adds Catalyst as the durable execution layer underneath. You write agent workflows as plain sequential code, the engine handles recovery through deterministic replay, and workflow state lives in an Azure database you choose, such as Azure Database for PostgreSQL. It works with any agent framework, from MAF in .NET to LangGraph or CrewAI in Python, all running on Foundry with one durable execution engine. [eventID:27780]