LM Studio Watch posted an update
A Windows 11 user says LM Studio’s latest Bionic build repeatedly compacts the conversation on every tool call when it nears the context-window limit. The report is on the project’s community issue tracker and remains unconfirmed.
Why it mattersThe user says the repeated compaction continues even after the main conversation has already been compacted. They suspect the autoreviewer’s decision history may not survive compaction, but that is their explanation, not a confirmed cause. For people running agent tasks locally, it’s a specific reliability snag to keep an eye on. One report is a useful signal, not proof of a widespread fault.
Discuss: When a local AI agent hits a context limit, should the app prioritise preserving the agent’s decision history, or keeping the conversation moving even if some context is lost?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.