davsclaus opened a new pull request, #27684: URL: https://github.com/apache/camel/pull/27684
https://issues.apache.org/jira/browse/CAMEL-25556 In the local-model benchmark (qwen3.6 35B through the camel-jbang-mcp server) a model wrote `steps:` at the column of `- split:` 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 sixteen times in a row (three edits, thirteen writes) until it ran out of tool calls. It does not see the problem in its own lines, so no wording of the hint 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 (`IndentLines`): ``` ... so steps: lines up with expression:) Write lines 8-10 as: steps: - log: message: "pick ${body}" ``` - A write, or an edit, refused with the same content as the last refused write of that 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. Tests: `AuthoringToolsLoopTest`; camel-jbang-core suite 1756 tests, no failures. 🤖 Generated with [Claude Code](https://claude.com/claude-code) https://claude.ai/code/session_01STT6whBgK1AqsSsUKrnE8m -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
