📝 docs: Memory tab Dreaming Memory layout and inline help Document Fixed memory vs Dreaming Memory sections, title help icons, and dynamic memory nested under the dreaming group.
📝 docs: last dream run is the latest scheduled attempt Clarify that the Memory tab shows the latest dream attempt (not a historical manual Regenerate) and a failure hint when that attempt failed.
📝 docs: dream-owned dynamic memory UI and Compact under Chat Preference - Memory tab: drop manual Regenerate/Restore; dynamic memory is edit-only (Save, Copy, Clear) with the dream as the automatic writer - Document the last dream run status row under the dream schedule - Compact group (Assist preset, auto-compact, threshold) now lives under Assistant → Chat Preference - How to set up guide moved under Fixed memory with fixed-only copy
📝 docs: distinguish configured vs expanded history limit in popover Co-authored-by: Cursor <cursoragent@cursor.com>
📝 docs: qualify dream settled markerKeyHash to enqueue_failed Worker and malformed-job dream_scheduler_settled events do not emit markerKeyHash; document hash correlation only for dispatcher enqueue failures.
📝 docs: document dream settled markerKeyHash and enqueue_failed Clarify that dream_scheduler_settled is a terminal dispatcher or worker event, not only a completed job outcome, and document markerKeyHash correlation plus enqueue_failed diagnostics.
📝 docs: document server-side daily memory dream Clarify that the dream runs on the Graphile worker at UTC time with no open browser, and that compaction-debug dream events are server-emitted.
📝 docs: scope idempotent topic create to source session
📝 docs: cross-session topic hydration and idempotent create
📝 docs: note scoped auto-topic compaction, recovery, titles, and draft tokens
📝 docs: note auto-topic compaction and post-compaction fences Document that auto-created topics compact before enqueue, account/clear fences are rechecked after compaction, and large-window overhead uses skill XML plus per-user templates.
📝 docs: note durable pre-send compaction and history-window fields
📝 docs: clarify next-request context window and pre-send compaction Document history-window popover fields, large-context expand, assist-level summary caps, pre-send message_count, and new CHATHUB_COMPACTION_DEBUG fields.
📝 docs: note hide flush drains the generation-debug client queue
📝 docs: document event_drop_summary for generation-debug event_dropped is now only the attached-then-missed smoking gun; SSE replay volume is event_drop_summary.
📝 docs: lead wiki Home with the 2.0 story instead of a comparison dump
📝 docs: mark ChatHub 2.0 catalog and v2.x versioning
📝 docs: document composer-scoped PASTED chips and Shift-paste bypass
📝 docs: explain compact PASTED chips and topic cards
📝 docs: note best-effort Artifacts storage cleanup
📝 docs: document Artifacts batch delete and native-ratio cards
📝 docs: drop the model suffix from the cache status line
📝 docs: note flat durable usage on the token popover
📝 docs: note grouped usage and totals-only cache display
📝 docs: note cache help icon on the compact token popover
📝 docs: clarify most recent reported cache on the token popover
📝 docs: document last-response cache hit rate on the token popover
📝 docs: document CHATHUB_COMPACTION_DEBUG for early compact diagnosis Explain the 80% token gate vs message-count/daily/manual, and how to read the compaction debug stream when compact looks early.
📝 docs: clarify History Compress thinking-off and translation cap
📝 docs: empty compaction summaries fail fast instead of eight retries Document History Compress as the summarizer model, thinking-model empty summaries, and that background compaction does not retry empty content the way title-transcript races do.