mesalogo
9836a4cc9d
Keep interactive questions and answered cards at their original position in conversation history, including remote reception, recovery, and reopening. Preserve legacy history placement without a database migration. Move supplementary guidance into shared accessible contextual-help popovers across pages and settings. Add keyboard, modal, accessibility, and Electron regressions, plus bounded opt-in live canvas-analysis validation. Remove unused Agent, Runtime, and in-memory graph-search helpers, obsolete mocks and browser copy; reuse the existing package-size formatter. Keep production execution paths, protocol schemas, and released data migrations. Validation: targeted cleanup tests, typecheck, lint, shared Linux Host attach/read/re-attach, and added help/Electron checks passed. Final full-suite verification passed 4,858 tests with 67 skipped and zero failures (412 files passed, 9 skipped). Focused help checks passed 54 tests, including Electron style comparisons and Chromium accessibility descriptions. Details and validation boundaries are recorded in feature docs. Release note: 修复交互式问答在回答后排到后续输出末尾的问题，重新打开会话仍保留提问位置。页面和设置中的补充说明改为标题或字段旁的按需帮助，支持鼠标与键盘操作。
Magic Notes
Magic Notes is the local notes, checklist and AI-comment workbench. Desktop actions and authorized Agent/MCP tools use the same desktop store: SQLite for metadata and indexes, and local files for entry bodies and binary assets. The open workbench subscribes to successful writes so users can see external changes without leaving the page or clicking a refresh button.
The workbench opens on a responsive note-card overview. Opening a note replaces the overview with a continuous record stream, a narrow left index and independent right AI comments. Both side panes can collapse. Index items show one real first-page canvas thumbnail with page count, or a text summary; clicking scrolls without replacing the editor or losing a draft. Opening a note shows a blank composer above the history, with text/canvas selection and Save. Creating a record resets that composer; editing a saved record requires its explicit Edit action. The separate To-dos view keeps search, status filters and grouped tasks on one page, with direct completion and a stable left list/right detail layout for instructions, source content and AI comments. Narrow containers use on-demand detail with an explicit return to the mounted list. One content-sized button switches to the other view in the page header immediately before New note; status filters sit beside search and wrap when needed. Note detail removes the surrounding frame and stream padding while keeping editor borders and the AI divider. Returning from a source note restores the task context and protects unsaved editor changes. A successful save followed by no further edits does not prompt again on return. Narrow note layouts use an index drawer and bounded bottom AI pane.
Documents
- Technical design: file storage, schema 38 migration, backup boundaries, database notifications, process boundaries, canvas and AI contracts, refresh lifecycle and regression coverage.
- Progress: implementation and validation evidence.
- UI design system: authoritative workbench interaction rules, including selection and draft preservation.
- Canvas integration: React controller API, PeopleLib core limits, asset adaptation and rendering lifecycle.
An entry is a saved rich-text or paged-canvas record within a note. Its body lives
in notes/<noteId>/entries/<entryId>.json beside note-local binary assets;
note.json is a derived membership manifest. SQLite retains metadata, indexes,
revisions, todos and comments. API reads hydrate assets back into the existing
content format. A todo is a checklist item derived from rich text
or the canvas Quill flow. An unsaved draft belongs to the open editor; it is not
written to storage by a background refresh.
Backups must coordinate SQLite and the adjacent notes/ tree while note writes
are paused or the app is closed. A database-only backup cannot restore these
bodies. There is no product one-click backup action; see the
backup contract.
Canvas Scope
The integrated PeopleLib editor uses Fabric for paged annotations and Quill for body text flowing across pages. It supports paper templates, pen/highlighter, object erasing and selection/transforms, floating text, images, PDF backgrounds with native text extraction, and raster PDF export. Saves are manual. Undo/redo belongs to the current annotation page or Quill body mode; it is not a global document history. There is no infinite canvas or PNG download button. Editors and saved viewers support view zoom and responsive Fit width without changing saved content or export dimensions; see the canvas zoom reference.
Drafts, saved entries and todos sourced from canvases support AI comments using
the default model's image-input capability. The Canvas pages to send setting
selects the first 1-8 pages in current canvas order, defaulting to 1. It counts
pages in one canvas, not entries, and leaves full-canvas saving and export intact.
Image-capable models receive complete captures and extracted text for that prefix.
Text-only models receive the same page range as an explicitly labelled
text fallback; a purely visual canvas entry or draft reports that an image-capable
default model is required. Canvas comments are requested manually in immediate
mode, or after a manual save in after-save-auto mode. Analysis failure does not
undo a successful save. Detailed retention rules are in the
AI contract.
Canvas implementation is complete and final validation is in progress. The progress record distinguishes confirmed production-path evidence from the pending final repository checks.