[
https://issues.apache.org/jira/browse/CAMEL-24859?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Claus Ibsen resolved CAMEL-24859.
---------------------------------
Fix Version/s: 4.23.0
Resolution: Fixed
> camel-jbang-mcp - camel_write_file reports the outcome of the reload in dev
> mode
> --------------------------------------------------------------------------------
>
> Key: CAMEL-24859
> URL: https://issues.apache.org/jira/browse/CAMEL-24859
> Project: Camel
> Issue Type: Improvement
> Components: camel-jbang
> Reporter: Claus Ibsen
> Priority: Major
> Fix For: 4.23.0
>
>
> Found by the stepwise half of the round-2 local-model benchmark (ten
> camel-jbang-examples built step by step through the camel-jbang-mcp server,
> five runs each, 2026-09-21).
> After a write to a project whose integration runs in dev mode, the tool
> answers "written; an integration running the file in dev mode reloads it
> now". Whether the reload succeeded is not part of the answer. In the series
> the model wrote a route the validator accepted but the runtime rejected (a
> cron endpoint without its name, a ${...} placeholder in a timer option), the
> reload failed with a WARN in the log, and the model reported success and
> moved on, or thrashed until its tool budget was gone. A person watching the
> console sees the WARN; an agent reading the tool answer does not.
> When the file belongs to a running integration in dev mode, the tool should
> wait a few seconds for the reload record ("Routes reloaded summary" or "Error
> reloading routes from file" in the log, or the reload event) and include the
> outcome in its answer: reloaded, or the error with the validator's report
> (CAMEL-24851 prints it in the log already). The single biggest improvement
> for the edit-run loop seen in the series: the model acts on what the tool
> tells it.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
