# 104008: Run response stream failed

Identifier: `run_response_stream_failed`



An AI Transport run's response could not be streamed to the channel: the source event stream threw, or an underlying publish failed mid-stream.

## What you should do

Re-attempt the [step](https://ably.com/docs/ai-transport/concepts/steps.md) that was producing the response. Under a [durable-execution framework](https://ably.com/docs/ai-transport/features/durable-execution.md) the retry re-runs it and supersedes the failed attempt; without one, the run has failed, so produce the reply again with a fresh run, which on a chat client is a regenerate. The partial response already on the channel is not completed on its own.

Inspect the `cause` for the underlying failure; it carries its own error code, and shows whether the response stream or a publish was at fault.

## Why it happens

A run's response is streamed to the channel event by event. This error is raised when that pipe fails: the source event stream throws, or one of the underlying publishes fails after streaming has begun. The same code is also used as a fallback when a run ends in error but no specific code survived on the wire. The originating error is preserved as the `cause` where available.

## What you'll see

The error is reported with code 104008 and status 500. Messages take the form `unable to pipe response for run ...; ...` or `unable to complete run; ...`.