[
https://issues.apache.org/jira/browse/CAMEL-25556?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Claus Ibsen reassigned CAMEL-25556:
-----------------------------------
Assignee: Claus Ibsen
> camel-jbang-mcp - a write the same as the last refused one says so, and the
> hint for an option at the column of its EIP shows the lines to write
> ------------------------------------------------------------------------------------------------------------------------------------------------
>
> Key: CAMEL-25556
> URL: https://issues.apache.org/jira/browse/CAMEL-25556
> Project: Camel
> Issue Type: Improvement
> Components: camel-jbang
> Reporter: Claus Ibsen
> Assignee: Claus Ibsen
> Priority: Minor
> Fix For: 4.24.0
>
>
> In the local-model benchmark (qwen3.6 35B through the camel-jbang-mcp server)
> a model wrote steps: at the column of - split: (so YAML reads it as a second
> key of the step) and got the hint of CAMEL-25529 ("steps: lines up with
> split: ... indent steps: and the lines under it two spaces more, so steps:
> lines up with expression:"). It then sent byte for byte the same content
> again sixteen times (three edits, thirteen writes) until it ran out of tool
> calls; two steps of one run were lost this way.
> The model does not see the problem in its own lines, so no wording reaches
> it; what a model reliably does is copy lines it is given (CAMEL-24906).
> * The message of an option at the column of its EIP ends with the lines to
> write: "Write lines 8-10 as:" and the lines two spaces further right.
> * A write (or an edit) refused with the same content as the last refused
> write of the file says so: "the content is the same as your last write of it,
> refused 3 times now for the same problems. Change what the errors say; where
> an error shows the lines to write, copy them." A write that goes through, or
> other content, starts over.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
