- Redesign agents page with architectural-grid (8) design system:
rounded-2xl cards, serif headings, motion tabs, dashed templates section
- Replace agent form popup with full-page detail view (SettingsView style)
with dark planning card, section tooltips, and help button
- Hide advanced mode for slide/excalidraw generators
- Add 'describe images' action to contextual AI assistant
- Add copy button to action/resource preview with HTTP fallback
- Add delete history button to agent run log panel
- Increase AI word limit from 2000 to 5000 (reformulate + transform-markdown)
- Increase max steps slider from 25 to 50
- Fix image description error with clear model compatibility message
- Fix doubled execution count display in agent detail view
- Remove dead files: notes-list-view.tsx, notes-view-toggle.tsx
- Remove 'list' view mode from NotesViewMode type
- Add missing i18n keys (back, configuration, options, copy, cleared)
- Toolbar: remove text labels from all icon buttons (AI, Save, Preview, Convert)
all buttons now icon-only with title tooltip for accessibility
- Toolbar: reposition PanelRight (info panel toggle) to far right after three-dot menu
- Versioning: decouple getNoteHistory/restoreNoteVersion from global userAISettings.noteHistory
now checks note.historyEnabled directly — unblocks manual per-note history
- Versioning: add 'Sauvegarder cette version' button in Versions tab of info panel
calls commitNoteHistory with visual feedback (spinner → success state)
- note-document-info-panel: import commitNoteHistory, add isSavingVersion state
- notes.ts: fix double guard that silently blocked all history operations
Corrige le typecheck Next Docker: module dagre sans .d.ts, et annotations
Presentation vs constructeur pptxgenjs.
Co-authored-by: Cursor <cursoragent@cursor.com>
pptxgenjs sur Node.js ne peut pas charger les URLs qui nécessitent une auth
ou des cookies. On fetch maintenant chaque imageUrl côté serveur et on la
convertit en data URI base64 avant d'appeler buildPresentation.
Si le fetch échoue (timeout 8s, 4xx/5xx), l'imageUrl est supprimée et
le placeholder texte s'affiche à la place.
Co-authored-by: Cursor <cursoragent@cursor.com>
Slide generator (generate_pptx):
- Pivot vers génération PowerPoint natif (pptxgenjs) au lieu de Reveal.js HTML
- 4 nouveaux layouts diagramme : timeline, process, comparison, metrics
- 2 nouveaux layouts image : image-content (texte + image), image-full (plein cadre)
- Redesign visuel de tous les layouts (cover split, section full-bleed, header band)
- Palettes corrigées : bg blanc sur toutes les palettes, contrastes réels
- fit:shrink systématique sur tous les textes pour éviter les débordements
- Extraction automatique des images des notes (Markdown/HTML) et injection dans le prompt IA
- Prompt IA renforcé : impose "style" et "theme" explicitement dans le JSON, impose ≥2 layouts diagramme
- Fix overlap timeline : zones de texte calculées sans collision avec les cercles
- Notification agent mise à jour : bouton download .pptx au lieu d'ouvrir HTML
Excalidraw generator:
- Layout Dagre/ELK pour graphes auto-positionnés
- Styles visuels : coloré, austère, sketch-plus (Virgil font)
- Zones/containers pour architecture-cloud
- Sanitisation du graphe et métriques de qualité de layout
Co-authored-by: Cursor <cursoragent@cursor.com>
- Agent-created notes are now type 'markdown' by default (raw markdown content)
- Switching note type from markdown → richtext auto-converts content to HTML
via /api/ai/convert-markdown endpoint using the existing markdownToHtml utility
- Users can freely switch between markdown and richtext in the note type selector
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add DeepSeek, OpenRouter, Mistral, Z.AI, LM Studio as AI providers
with editable model names via Combobox in admin settings
- Fix OpenRouter broken by normalizeProvider bug in config.ts
- Convert agent-created notes from Markdown to HTML (TipTap rich text)
- Add Notification model + in-app notifications for agent results
- Agent notification click opens the created note directly
- Add note count display on notebook and inbox headers
- Fix checklist toggle in card view (persist state via localCheckItems)
- Add checklist creation option in tabs/list view (dropdown on + button)
- Fix image description ENOENT error with HTTP fallback
- Improve UI contrast across all themes (input, border, checkbox visibility)
- Add font family setting (Inter vs System Default) in Appearance settings
- Fix CSS font-sans variable conflict (removed dead Geist references)
- Update README with new features and 8 providers
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Raise MAX_WORDS from 500 to 2000 for text reformulation
- Error messages now use i18n keys (ai.wordCountMin/Max) instead of
hardcoded English strings
- Client translates server errors using the user's UI language
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The admin form saves Ollama URLs as OLLAMA_BASE_URL_TAGS/EMBEDDING/CHAT,
but the factory only read OLLAMA_BASE_URL — causing 500 errors in Docker
where no localhost fallback exists.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
When chatScope is "this note", the AI had access to note_search and note_read
tools which let it pull content from any note. Now these tools are excluded via
a webOnly flag in buildToolsForChat — only web_search/web_scrape remain if
the user toggled web search.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- Add image description service + API route for AI-powered image analysis
- Image title generation returns 3 selectable suggestions via TitleSuggestions component
- Add "Describe images" action in AI assistant (individual + collective)
- Fix pin refresh propagation in card and tabs view
- Fix note creation refresh in tabs mode, pass all notes to tabs view
- Add RTL support (dir="auto") on note content elements
- Pass UI language dynamically to AI endpoints instead of hardcoded 'fr'
- Add 18 missing i18n keys in both en.json and fr.json
- Sparkles button on images for AI title generation (bottom-right, pulse animation)
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Next.js bakes public/ at build time — dynamically uploaded files were
never served in Docker standalone mode. Store uploads in data/uploads/
and serve via /api/uploads/ with a rewrite for backward compatibility.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- Add stop button to all chat interfaces (floating, contextual, full-page)
- Add conversation sliding window (50 messages) to prevent context overflow
- Add chat timeout warning (30s toast)
- Force response language in chat system prompt (mandatory per-locale)
- Add image paste from clipboard in all note editors (card, list, input)
- Fix upload API to infer extension from MIME type for clipboard images
- Add image description support in note AI chat (base64 vision)
- Fix search regex crash on special characters (escape user input)
- Fix search case-insensitivity on PostgreSQL (mode: 'insensitive')
- Add try/catch around semantic search in chat route (prevent blocking)
- Add new chat button to floating AI assistant
- Fix empty thinking bubbles for reasoning models (filter non-text parts)
- Remove duplicate AI assistant toggle from note editor header
- Improve link metadata scraping (timeout, content-type check, relative URLs)
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Same pattern as TitleSuggestionService: getModel() + generateText
with system/user prompts. LanguageDetectionService (tinyld) auto-detects
note content language. Labels now match note language (Persian note → Persian labels).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Use LanguageDetectionService (tinyld) to detect the language of note
content, same pattern as TitleSuggestionService. No more hardcoded
per-language prompts — single prompt with detected language injected.
Labels now match the note content language (e.g. Persian note → Persian labels).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Rewrote prompts to be strict about content relevance — only suggest
labels directly related to THIS note's content. Raised confidence
threshold to 0.5 in code, 0.7 in prompt. Limited to max 2 suggestions
from existing labels. Prevents same labels being suggested for every note.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Log key points in /api/ai/tags and ContextualAutoTagService to trace
where label suggestions fail: notebook lookup, AI raw response,
JSON parsing, filtered results.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- Use prisma.$transaction in auto-label-creation.service with tx client
- Fix DELETE /api/labels to properly JSON.parse + disconnect labelRelations
- Fix PUT /api/labels rename to handle JSON labels
- Graceful error handling in /api/ai/tags and /api/ai/auto-labels
- Client-side label-deleted event in home-client, notes-tabs-view, label-management-dialog
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- Replace broken upsert (nbScope ?? '') with findFirst+create for null notebookId
- Remove aggressive orphan cleanup that deleted notebook labels after save
- Add syncNoteLabels() to update both Note.labels JSON and labelRelations
- Fix createNote, updateNote, auto-labeling to use syncNoteLabels
- Add fallback: suggestFromExistingLabels → suggestNewLabels if empty
- Lower confidence threshold 0.6→0.3, max suggestions 3→5
- Case-insensitive label matching in suggestions
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- Add scheduledTime, scheduledDay, timezone fields to Agent schema
- Create calculateNextRun() helper with timezone-aware scheduling
- Add POST /api/cron/agents endpoint for external scheduler
- Calculate nextRun on agent create, update, and after execution
- Add time/day picker in agent form (daily/weekly/monthly)
- Show "Next run" countdown in agent card
- Add i18n keys for schedule UI (FR + EN)
External scheduler (N8N, Vercel Cron) should call /api/cron/agents
every 5-15 min. Requires `prisma db push` to apply schema changes.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
MCP server:
- Fix Prisma imports from stale client-generated path to @prisma/client
- Switch schema from SQLite to PostgreSQL for Docker compatibility
- Add prisma generate step to Dockerfile with proper binaryTargets
- Include index-sse.js in Docker build (was excluded by .dockerignore)
- Install openssl and libc6-compat in Alpine image for Prisma runtime
Docker:
- Fix memento-note healthcheck (wget unavailable in bullseye-slim)
Minor fixes:
- scrape.service SSRF protection, middleware route coverage
- canvas-board and note-input type fixes
- next.config turbopack and devIndicators adjustments
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Security:
- Add auth + file type/size validation to upload API
- Add admin auth to /api/admin/ endpoints
- Add SSRF protection to scrape action
- Whitelist fields in PUT /api/notes/[id] to prevent mass assignment
- Protect /lab, /agents, /chat, /canvas, /notebooks routes in middleware
AI provider fixes:
- Add deepseek/openrouter to factory ProviderType (was silently falling back to ollama)
- Fix title-suggestion.service.ts to use factory instead of hardcoded OpenAI
- Fix getAIProvider→getChatProvider in memory-echo, notebook-summary, agent-executor
- Fix getAIProvider→getTagsProvider in notebook-suggestion, title-suggestions, transform-markdown
Functional bugs:
- Fix ALLOW_REGISTRATION AND→OR logic
- Fix note-editor.tsx passing stale props to useAutoTagging instead of local state
- Fix stale Note.embedding type (migrated to NoteEmbedding table)
- Remove hardcoded SQLite path from prisma.ts
Frontend:
- Add AbortController to useAutoTagging and useTitleSuggestions hooks
- Add error rollback to optimistic UI in note-inline-editor
- Remove stale closure over notebookId/language in useAutoTagging
Cleanup:
- Rename docker-compose from keepnotes→memento
- Remove unused unstable_cache import from config.ts
- Remove dead useUndoRedo hook
- Fix TagSuggestion type (add isNewLabel, reasoning)
- Remove dead AIConfig/AIProviderType types
- Fix ghost-tags unused isEmpty var and as any cast
- Fix note-editor titleSuggestions typed as any[]
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- app/api/ai/tags/route.ts: getAIProvider → getTagsProvider (was using
embedding model instead of tags model for legacy tag generation)
- lib/ai/services/auto-label-creation.service.ts: same fix
- hooks/use-auto-tagging.ts: remove notebookId gate that blocked all
suggestions in "General Notes" — the API has a legacy path for this
- app/actions/admin-settings.ts: revalidateTag → updateTag (Next.js 16
requires 2 args for revalidateTag)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Auto-tagging:
- CRITICAL FIX: contextual-auto-tag.service.ts was calling getAIProvider()
(alias for getEmbeddingsProvider) instead of getTagsProvider(). This meant
auto-tagging used the embeddings provider/model instead of the tags one.
Now correctly uses getTagsProvider() in both suggestFromExistingLabels and
suggestNewLabels methods.
- Pass user's detected language to suggestLabels() for localized prompts
(was hardcoded to 'en')
Email:
- Fix Resend "from" field: pass DB config to sendViaResend() instead of
re-fetching from DB. Uses SMTP_FROM from config, with localhost-aware fallback.
- Add "Sender email" field in admin Resend section so users can set SMTP_FROM
- Save SMTP_FROM when Resend is selected (was only saved for SMTP mode)
- Test email button now saves config to DB BEFORE testing, so unsaved form
values are used (was reading stale DB values)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Remove BMAD framework, IDE configs, dev screenshots, test files,
internal docs, and backup files
- Rename keep-notes/ to memento-note/
- Update all references from keep-notes to memento-note
- Add Apache 2.0 license with Commons Clause (non-commercial restriction)
- Add clean .gitignore and .env.docker.example
