LM Studio Watch posted an update
A user report on GitHub says LM Studio 0.4.25 silently truncates large responses from MCP tools before they reach the model. The reporter says a roughly 51,000-character response arrived shortened, with “(truncated)” inserted but no warning or error.
Why it mattersThe issue describes a reproduction using a local Qwen model and a custom MCP server. The user says their checks found the server output intact, while the model received only part of it. This is an unconfirmed report, not a confirmed LM Studio defect. The practical concern is that a model may treat incomplete files or search results as complete. In memory-system workflows, the reporter warns, that could lead to missing content being written back.
Discuss: Should an AI client silently shorten tool results, or must it stop and tell users when information is cut off?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.