Changelog.

Better Research Saves

• Saved research notes now include more of the full conversation, especially when you continue chatting after the first result.
• Completed follow-up findings are folded into saved research artifacts instead of being left behind as loose checklist items.
• Research can still be saved as a full note, brief, comparison matrix, or notebook.
• Saving research now shows a clearer live preview while Canto prepares the final document.

Edit Open Notes From Canto Chat

• Canto Chat can now help revise the note currently open beside it.
• You can ask Canto to rewrite, tighten, restructure, or expand part of the open note.
• Proposed edits are shown with a preview before you approve them.
• The handoff between Canto Chat and the open document feels smoother and more direct.

EPUB Book Support

• EPUB files are now supported as source files.
• You can add EPUB books to your Library or drop them into chat.
• Canto can preview book covers when available.
• Book text is read with chapter awareness, making summaries and study guides easier to follow.
• Broad prompts like “summarize this book” or “make a chapter-by-chapter guide” now work better with EPUB sources.

New Local Model Options

• Added new Gemma 4 local model choices.
• These models give you more private, on-device options for writing, reasoning, coding, and image-aware work.
• Older local model selections are guided toward the newer supported options.

Smaller Improvements

• The Memory window now fits better on narrower screens.
• Long rows in the Memory panel wrap properly instead of overflowing.
• Research save, follow-up, and stop/resume controls are clearer while work is running.
• Citation and source links are more consistent across chat-created notes.
• EPUB files appear more naturally in Library and supported-file flows.

Research Planning

• Canto now drafts a research plan before longer research begins.
• You can review, adjust, approve, or cancel the plan before Canto starts.
• Deeper web research now asks for approval before using external research credits.
• Cancelling a plan no longer leaves behind a messy half-started run.

Live Research Progress

• Research runs now show a clearer source trail while Canto works.
• You can see whether Canto is reading from your notes, Library files, or the web.
• Deep research now shows live status and elapsed time, so longer runs feel less stuck.
• You can stop a research run, narrow the topic, or change source types mid-run.

Research Results

• Completed research now gets a dedicated results card.
• Results can include unverified claims and follow-up tasks.
• Follow-up tasks are shown as an interactive checklist.
• You can run follow-up tasks when you are ready.

Save Research Outputs

• You can now save completed research as a Research note.
• You can also save research as a Research brief.
• You can save research as a Comparison matrix.
• You can save research as a Notebook when the work belongs with code, calculations, or step-by-step analysis.
• Saving research is now more reliable and no longer depends on the AI manually calling a note-creation tool.

Better Research Notes

• Saved research notes now have a cleaner, more paper-like structure.
• Research notes can include a real table of contents when useful.
• Notes now use cleaner numbered sections, methodology, limitations, and follow-up tasks.
• Saved notes are better at preserving important tables, numbers, comparisons, prices, figures, and other details from the research chat.
• Research note titles are now more specific and less likely to simply repeat the original question.

Cleaner Citations and References

• References are now de-duplicated, so the same source does not appear over and over.
• Web sources now show a clearer clickable link back to the original page.
• Vault sources can open the related note or notebook.
• Library file sources can open the related file.
• Unresolved citation markers are cleaned up instead of being left as stray text.
• Citation numbering is more consistent between the body of the note and the reference list.
• Saved notebooks now get a references section too.

Manual Citations

• Added a new /cite command in the editor.
• You can cite a web URL while writing.
• You can search for and cite another note from your vault.
• New citations continue the numbering already used in the document.

Source and Link Handling

• Saved research notes now index web source links for future search and memory use.
• Source links in saved notes show clearer indexing status.
• Bare or accidental web-looking text is less likely to turn into unwanted links.
• Task-list style items from research saves now become real checkboxes when appropriate.

Document Agent Improvements

• When you ask the document agent to edit selected text, it is better at keeping the edit focused on that selected passage.
• Moving from a saved research note into the document agent is smoother.
• If a saved note is not open yet, Canto offers to open it first.

Composer and Trial UI Polish

• The trial usage pill is now shared across chat composers for a more consistent look.
• Research controls are better integrated into both the home composer and in-session chat composer.

Theme Polish

• Canto now remembers your last light theme and last dark theme separately.
• Switching between light and dark returns to the theme you last used in each mode.

Model and Context Improvements

• Ollama Cloud models can now use more of the available conversation and source context.
• Long research and chat sessions have more room to include relevant context.

Smaller Fixes

• Research save flows are more dependable across notes, briefs, matrices, and notebooks.
• Research progress messages are clearer during long-running web research.
• Saved research notes avoid duplicate metadata sections.
• Created research notes behave better when edited afterward.
• The embedding and link indexing inspector is clearer when research saves create background work.

🤝 Keep Going in the Document Agent

• A friendly handoff strip — after Canto Chat creates or opens a note, a dismissible prompt offers to continue in that note's docked Agent Chat.
• Animated, not jarring — moving from the chat home into a note-and-agent layout now glides instead of snapping.
• Chat can read the open document — when chatting beside a note or notebook, Canto Chat can read and cite the document open next to it.
• Clear web boundaries — if you ask something that needs the web while web access is off, Canto tells you plainly instead of quietly coming up short.

☁️ iCloud Sync, Rebuilt for Real Time

• Real-time updates — changes now stream between your Macs promptly instead of waiting on slow background passes.
• One simple switch — turning on iCloud Drive Sync is all it takes; the faster engine activates automatically.
• Your whole vault stays in sync — no more picking which folders to sync; everything is kept together by default.
• Trustworthy cloud badges — the cloud icons next to your notes now reliably reflect what's safely backed up, including notes restored on a new Mac.
• The sidebar keeps up — when a change arrives from another Mac, your sidebar and file tree refresh on their own instead of showing stale rows.

🩺 Calmer Sync Settings, With Real Repair Tools

• A clear status strip — one compact line tells you what sync is doing, instead of a wall of technical toggles.
• Sync Doctor — a one-tap, non-destructive repair that re-checks your vault against the cloud and recovers anything missing.
• Re-join from iCloud — a second Mac can cleanly join the shared vault already in your iCloud instead of starting a separate one.
• Clean-slate reset — a clearly labeled reset wipes the cloud and local sync state for a fresh start, with a separate hard reset when you want everything cleared.
• Quieter syncing indicator — the "Syncing…" spinner now appears for real work, not every routine background check.

🏷️ Smarter Chat Titles Without the Wait

• An optional tiny helper — download a small on-device model just for naming chats, from Settings → AI Models.
• Never blocks your main model — it loads, names the chat, and unloads on its own.
• Graceful fallback — without it, Canto still gives your chats sensible titles automatically.

⚡ Better Local Models

• Higher-quality downloads — local models now use newer, better-tuned versions for cleaner output.
• Tuned for GLM Flash — the GLM model now runs with sampling settings matched to its recommended defaults.
• Accurate download states — the model picker and settings track downloads live and stop showing stale "loading" or "downloading" states once a model is ready.
• Steadier on tighter memory — a couple of models had their settings adjusted to match what runs reliably in practice.

🧮 Turn Plain Text Into Real Equations

• Conversion requests understood — phrases like "convert this to an equation" are recognized as real edit requests.
• Lands as real math — the result is an editable, rendered math block, not flattened plain text.

🛠️ Smaller Fixes & Refinements

• Notes remember your place — switching note tabs restores your cursor and scroll position instead of resetting to the top.
• Tidier status bar in narrow panes — in split-pane layouts the footer stays on a single row, switching to compact icons instead of wrapping.
• Copy Mermaid errors — when a diagram fails to render, you can copy the error text in one click from both the editor and Agent Chat.
• Cleaner agent activity rail — nested timeline steps line up neatly and the live edit preview is easier to read.
• Bigger vault tidy-ups — the AI Vault organizer can reorganize large vaults in one pass, with a dedicated panel for reviewing and approving the plan.
• Latest building blocks — the editor, app runtime, and build tooling were updated to their newest stable versions for better reliability.

🔄 Improved

A quieter shell

• Floating update pills — update prompts no longer push your editor down with a full-width banner; they float as a compact pill below the tab bar.
• What's New, in the editor — release notes now open as a clean read-only header on a real document instead of a modal-feeling banner.
• Simpler memory status pill — a single CPU icon plus a resident-RAM number, easier to read at a glance.
• Refreshed macOS app icon for a cleaner look in Dock, Finder, and Spotlight.
• Traffic-light-safe drag regions on the home screen — moving the window no longer activates composer controls by accident.

Document Agent Panel parity

• The docked Agent Panel got a UI refresh to align with Canto Chat.
• It now shares the new Library tooling for listing, reading, and attaching assets.
• The smooth streaming "typewriter" reveal that made the Agent Panel feel snappy is now shared with Canto Chat assistant messages.

Polished chat output

• Assistant tables now render with the same restrained styling Canto uses everywhere else.
• Citation tooltips no longer cover up nearby text.
• Citation markers no longer collide with [[wikilink]] syntax.

🔒 Pro Trial

• Vault context and web search are now Pro features — free users see a clear lock with a single nudge to upgrade, instead of silent failures or empty answers.
• One free Pro-flavored query per send is enforced server-side; legacy trial counters now align to a clean 25-query baseline.
• Basic on-device chat without grounding stays unlimited and free.

🐛 Fixes

• Stop button actually stops — pressing Stop now cleanly cancels both agent and plain chat turns mid-stream.
• Hardened external model switching — switching between local and external models mid-session no longer wedges the chat or strands a half-finished reply.
• Stable per-turn citation markers — tool results now carry consistent citation IDs across a turn, so [1] always means the same thing.
• Vision context sized correctly — multi-frame video inference no longer overflows the model's context window when batches stack up.
• Save-as-note now renders real markdown HTML instead of raw text.

Cold-Start Local Models

Canto no longer eagerly loads your local model the moment the app opens.

• Faster launches — startup no longer pays the cost of mapping gigabytes of weights.
• Lower idle memory — a session where you only read or write notes never pays the AI footprint at all.
• Warming, surfaced clearly — when you do send the first prompt, Agent Chat shows a brief warming status before the response begins.
• Adopts already-loaded models — if the runtime is warm from another flow, the agent reuses it instantly instead of reloading.

Send Diagnostics

A new Send Diagnostics option in the Help menu and on Settings → About makes it easy to report a bug without exposing your notes.

• Full preview before sending — the entire support bundle is shown to you as JSON, with secrets and personal content scrubbed out.
• Pick what to include — toggle whether to attach a screenshot, recent crash metadata, memory state, update state, and database health info.
• Save locally or upload — save a copy to disk for your own records, or post it directly to LonelyDuck.
• Crash-aware — if a window or helper process crashes, that event is captured locally so the next diagnostics report has the right context.

Long Conversations Stay Snappy

Big tool results — web research, large file reads, multi-page extracts — used to live in memory for the whole session. They now offload to disk, so chats stay light no matter how long they get.

• Big tool results stored on disk as artifacts in your vault, with a compact summary kept in the chat itself.
• Cached token estimates — the context-usage indicator no longer re-tokenizes everything on every render.
• Reclaim now — a new manual button in Settings → Memory drops idle weights and frees RAM on demand, no restart needed.
• Smarter idle eviction — when Canto unloads an unused model, it now runs a deeper cleanup pass so RAM actually goes back where it belongs.

A Self-Healing Vault

The Memory Links sidecar database (the engine behind Related Notes and semantic search) is now resilient to corruption, and your main vault file actively reclaims unused space.

• Auto-rebuilds a broken index — if the embedding sidecar can't be opened, Canto quietly quarantines the bad files and starts a fresh index instead of refusing to launch.
• One-shot compaction at launch — when your vault has accumulated significant unused space, Canto compacts it on startup so the file shrinks back down on disk.
• No more index bloat — semantic-search tables are rebuilt cleanly each launch instead of growing indefinitely from background indexing.
• Separate dev and production indexes — development builds use their own sidecar file, so testing never collides with your real vault.

Friendlier Embedding Status

The status-bar pill and Memory Links inspector are much better at explaining which part of indexing needs attention.

• Notes vs. links called out separately — error states say things like "links are stuck on a fetch" or "notes need a rebuild" instead of a generic "needs attention."
• Active phases win — if links are actively being fetched or embedded, the pill stays in its working state instead of being overridden by stale errors.
• Real error messages bubble up — the inspector shows a short, truncated reason for the failure right in the lane header.

Web Search, Off Means Off

If you haven't set up a Tavily API key, web research tools used to still appear and fail confusingly. Now it's clean either way.

• Tools hidden when not configured — without a key, all web research tools (search, extract, map, crawl, research, fetch URL) are completely removed from the agent's toolbox.
• Friendly nudge in chat — if your message is clearly asking for research, the agent replies with a short note and a shortcut to Settings → Web Search.
• Polished warning panel — the missing-key banner in Agent Chat has been redesigned to match the rest of the modern chat surface.

The Local Agent Bails Out Instead of Spinning

Occasionally a local model produces a malformed tool reply. Older builds would let the agent keep retrying until the watchdog tripped — 0.7.9 catches the pattern and ends the run cleanly with an explanation.

• Repeat-failure guard — when the same kind of tool failure happens twice in a row, the agent stops and explains rather than looping.
• Clearer tool-error summaries — common validation problems now produce short, agent-readable messages so the model can self-correct on its next turn.
• Cleaner streaming output — internal scaffolding leaked into the streaming preview is stripped more aggressively, so what you see really is what lands in your note.
• Citations done right — the agent now writes [1]-style numeric citations instead of accidentally writing them as [[1]] wikilinks that point nowhere.

Under the Hood

• Unsloth-aware load planner — the local model loader understands sharded files, normalizes legacy IDs, and persists your effective selection so it survives migrations.
• Lifecycle-synced runtime — Agent Chat, Settings, and the memory inspector all stay in sync when the main process loads, sleeps, or evicts a model behind the scenes.
• Pressure-aware idle timers — outside Low Memory Mode, the idle-unload window now adapts to current memory pressure rather than using a single hard-coded value.
• Effective-settings caveats — if your saved weight-pinning options diverge from what's actually live in the runtime, the Memory panel flags it and prompts a quick reload.
• Lower baseline RAM — the embeddings model now idle-unloads after a quiet window even outside Low Memory Mode, and re-warms automatically on the next embed.