Durable Execution meets Durable Sessions: Resilient AI Agents with Temporal and Ably

Most teams building agents with Temporal have solved the backend problem: crashed workflows restart, LLM call failures retry automatically, and long-running tasks complete reliably.

What they haven't solved is the client side -- what happens to the stream when the user's connection drops, when they switch devices, or when two sub-agents are working concurrently and the client needs a single coherent view.

This demo shows how Durable Execution (Temporal) and Durable Sessions (Ably AI Transport) work together to close that gap.

Mike Christensen walks through a live Temporal + Ably integration that demonstrates:

• Resumable streaming: a network drop mid-response reconnects automatically, picking up exactly where it left off — without the agent needing to know or care
• Multi-tab and multi-device sync: two sessions (Alice and Bob) stay fully synchronized over a single Ably channel — no extra plumbing
• Multi-agent fan-in: concurrent sub-agents stream their output into a single multiplexed session layer the client consumes from one place
• LLM failure and retry: Temporal retries the failed activity; Ably structures the retry in the session so the client sees a clean canonical output, not a corrupted stream

Timestamps:

0:08 Introduction -- why Temporal teams need a session layer

0:25 What durable sessions are and how they pair with durable execution

1:02 The resumable multiplex stream: one connection, all session activity

1:33 Demo setup: simple chat app, Temporal backend

2:09 Two workflow runs visible in Temporal

2:20 Resumable streaming demo: killing the network mid-response

2:53 Multi-tab sync: Bob joins Alice's session

3:42 Multi-agent demo: concurrent sub-agent streaming via Temporal child workflows

4:37 Viewing multi-agent activity in Temporal

5:08 Single multiplexed stream across all concurrent sub-agents

5:31 Simulating LLM failure and Temporal retry

6:13 How the session layer structures retried output for the client

6:29 Clean canonical output after retry — no corrupted state

Links:
Ably AI Transport: https://ably.com/ai-transport
AI Transport docs: https://ably.com/docs/ai-transport
Ably GitHub: https://github.com/ably
Ably on LinkedIn: https://linkedin.com/company/ably-realtime/
Ably on X: https://x.com/ablyrealtime
Ably on Facebook: https://facebook.com/ablyrealtime

About Ably:
Ably is an enterprise-grade realtime messaging platform. We help developers build, ship, and scale realtime features — from chat and notifications to AI agent streaming and live data sync. Millions of messages delivered every day to thousands of companies worldwide.

#RealtimeAI #Temporal #DurableExecution #AIAgents #LLMStreaming