The automatic re-analysis routine was implemented in the HTTP route handler. Because of this, asking the Librarian to for example refine prose using editProse will not lead to re-analysis and has to be triggered manually.
Also worth flagging: the green dot indicator does not go stale even when editing prose manually; the invalidate helper within FragmentEditor invalidates the list of fragments and proseChain but doesn't invalidate the ['librarian-analysis-index', storyId] query key which controls the indicator. The 10-second polling interval (refetchInterval: 10_000) in ProseChainView would invalidate it, but unless the model is very slow it will likely have run re-analysis within this time.
The automatic re-analysis routine was implemented in the HTTP route handler. Because of this, asking the Librarian to for example refine prose using
editProsewill not lead to re-analysis and has to be triggered manually.Also worth flagging: the green dot indicator does not go stale even when editing prose manually; the
invalidatehelper withinFragmentEditorinvalidates the list offragmentsandproseChainbut doesn't invalidate the['librarian-analysis-index', storyId]query key which controls the indicator. The 10-second polling interval (refetchInterval: 10_000) inProseChainViewwould invalidate it, but unless the model is very slow it will likely have run re-analysis within this time.