M
Your Private AI Brain — Remember Everything. Locally.
Quick Start • Features • Screenshots • Tech Stack • Contributing • Support
Beta Release — Mnemosyne v1.1.4-beta is feature-rich and actively developed. Perfect for early adopters who want a private, AI-powered knowledge management system. See Roadmap
Your thoughts deserve better than scattered notes and forgotten bookmarks.
Mnemosyne is an AI-powered personal knowledge management system that runs 100% locally on your machine. No cloud. No subscriptions. No data harvesting. Just you and your AI-enhanced second brain.
New: Want more power? Enable experimental Cloud AI to connect Claude, ChatGPT, or custom endpoints — for those brave souls who trust Big Brother with their diary entries. Your keys are encrypted, usage is tracked, and you can always retreat back to local-only mode when the paranoia kicks in.
| What Makes It Different | |
|---|---|
| Truly Private | Your data never leaves your machine. Period. |
| AI That Understands | Local AI analyzes your notes, images, and documents — finding connections you missed |
| Living Knowledge Graph | Watch your ideas connect and evolve visually |
| Visual Intelligence | Upload images and let AI extract text, describe content, and auto-tag |
| Document Management | Upload PDFs for AI enrichment, review, and semantic search |
| Dashboard Home | See your entire knowledge base at a glance — metrics, quick capture, calendar, and system health |
| Ask Your Brain | Chat with your knowledge base in two modes — retrieval-augmented or brain AI companion |
Dashboard — Your Knowledge Base at a Glance Metrics, Brain Focus, Quick Capture, Calendar, Favorite Images, Tasks, Knowledge Graph stats, and System Status — all in one view |
|
Smart Gallery 3-column layout with albums, tags, and smart filters |
Notes Management Collections, wikilinks, hashtags, and rich text editing |
PDF Document Management Upload, AI enrichment, review workflow, and collections |
Daily Journal Calendar view with mood tracking, insights, and daily notes |
Knowledge Graph — Map Full knowledge map with communities and semantic clusters |
Knowledge Graph — Explore Focus on any node to see connections, backlinks, and details |
AI Chat — Dual Mode NEXUS RAG retrieval and Brain AI companion with topic memory |
Upload Multi-file upload with real-time AI analysis and customizable settings |
- Knowledge Base Overview — See total notes, images, documents, and tags at a glance
- Brain Focus — Visualize your current focus node from the Knowledge Graph
- Quick Capture — Jot down thoughts instantly without leaving the dashboard
- Calendar Widget — Navigate to journal entries with an interactive monthly calendar
- Favorite Images — Browse your starred photos right from home
- Tasks & System Status — Track your to-dos and monitor Ollama, database, Redis, and embedding health
- Customizable Layout — Rearrange and toggle widgets to fit your workflow
- Rich Text Editor — Full-featured Tiptap editor with markdown support
- [[Wikilinks]] — Connect ideas with bidirectional links
- #Hashtags — Organize with auto-extracted tags
- Backlinks — See every note that references the current one
- AI Image Analysis — Upload photos, get automatic descriptions and tags via Ollama
- Text Recognition — Extract text from images, screenshots, and documents
- Smart Gallery — 3-column layout with albums and tag filtering
- Auto-Generated Notes — Every image becomes a searchable, linked note
- Upload & AI Enrichment — Upload PDFs and let AI generate summaries, key points, and tags
- Review Workflow — Approve, reject, or edit AI-generated metadata before publishing
- Unified Collections — Organize PDFs and notes together in shared collections
- Full-Text Search — Semantic search across all document content
- Calendar View — Navigate entries with an interactive calendar
- Mood Tracking — Record and visualize your emotional patterns
- Daily Insights — AI-generated reflections on your journal entries
- Quick Capture — Fast entry with templates and prompts
- Interactive D3.js Visualization — See how your ideas link together
- Community Detection — AI discovers clusters and themes in your knowledge
- Semantic Edges — Connections weighted by meaning, not just links
- Path Finding — Discover hidden connections between distant ideas
- NEXUS RAG — Graph-native adaptive retrieval with citation-backed answers from your notes
- Brain AI Companion — Persistent AI that remembers topics and builds understanding over time
- Streaming Responses — Real-time answers as AI thinks
- Topic Management — Organize conversations by subject
For those who trust Big Brother with their notes — and get 200K token context windows in return.
- Claude, ChatGPT & Custom Endpoints — Connect Anthropic, OpenAI, or any OpenAI-compatible API (Groq, Together AI, etc.)
- Encrypted API Keys — Your keys are Fernet-encrypted at rest, never stored in plaintext
- Per-Feature Model Selection — Pick different cloud models for RAG chat vs Brain companion
- Token Usage Tracking — Monitor your cloud API costs with per-query logging
- Graceful Fallback — If the cloud fails, Mnemosyne seamlessly falls back to local Ollama
- Your Choice — Local-only by default. Cloud AI is opt-in with clear privacy warnings
- AI Clusters — Automatically grouped similar notes
- Inbox — Recent captures in one place
- Orphans — Notes waiting for connections
- Search by Meaning — Find notes by concept, not just keywords (pgvector)
- Full-Text Search — Traditional search when you need it
- Cross-Content — Search across notes, images, and documents simultaneously
Get your private AI brain running in 3 steps:
- Docker and Docker Compose
- 8GB RAM minimum (16GB recommended for AI features)
- NVIDIA GPU + NVIDIA Container Toolkit (recommended, not required)
No NVIDIA GPU? Remove the
deploy:blocks fromollamaandcelery_workerindocker-compose.yml. AI will run on CPU (slower but works).
# 1. Clone the repository
git clone https://github.com/Simmak7/project-Mnemosyne.git
cd project-Mnemosyne
# 2. Configure environment
cp .env.example .env
# Edit .env with your settings (defaults work for most users)
# 3. Launch Mnemosyne
docker-compose up -d --build# Pull AI models (one-time, ~8GB total download)
docker exec -it ollama ollama pull qwen2.5vl:7b-q4_K_M # Vision & image analysis
docker exec -it ollama ollama pull qwen3:8b # RAG text generation
docker exec -it ollama ollama pull llama3.2:3b # Brain AI & NEXUS navigation
docker exec -it ollama ollama pull nomic-embed-text # Semantic embeddingsAbout AI Models: These are the recommended defaults. You can swap models per capability (vision, RAG, Brain) anytime from Settings > AI Models — including pulling new models directly from the UI.
| Service | URL |
|---|---|
| App | http://localhost |
| API Docs | http://localhost/docs |
Custom port? Change
WEB_PORTin.env(default: 80). Access athttp://localhost:WEB_PORT.
Don't feel like reading instructions? Just paste this into Claude Code, Cursor, or your favorite AI coding assistant:
Clone https://github.com/Simmak7/project-Mnemosyne.git, set it up with
Docker, pull all the required Ollama models, and get it running on
localhost. Surprise me when it's done.
Sit back, grab a coffee, and let the machines do the work.
| Layer | Technology |
|---|---|
| Frontend | React 18, TanStack Query, Tiptap Editor, D3.js |
| Backend | FastAPI, SQLAlchemy, Celery, Pydantic |
| Database | PostgreSQL + pgvector |
| Cache/Queue | Redis |
| AI/ML | Ollama (local) + Anthropic, OpenAI (optional cloud) |
| Infrastructure | Docker Compose, Nginx reverse proxy |
- Performance Optimization — Multi-stage Docker build serves production React via nginx; gzip compression and immutable cache headers for static assets
- Backend Workers — Uvicorn multi-worker mode (2 workers), removed dev
--reloadflag - Graph Query Optimization — Batch wikilink/backlink resolution eliminates N+1 queries (~1,100 → 5 queries for full graph)
- Bundle Cleanup — Removed unused three.js, d3, react-force-graph (~290 KB gzipped saved); direct-path lazy imports for all routes
- Brain Graph Fixes — Depth filter now works correctly in Explore view; improved node spacing and smaller labels prevent overlapping when zoomed in
- Image Caching — ETag + Cache-Control headers on gallery images (24h browser cache)
- Celery Throughput — Worker concurrency increased from 2 to 4
- Clearer Navigation — Renamed confusing labels ("Studio" → "Upload", "Brain" → "Knowledge Graph"), added Home button to sidebar with visual divider
- Browser Back/Forward — Hash-based URL routing (
/#/notes,/#/gallery) enables browser history navigation and bookmarkable tabs - Streamlined Onboarding — Reduced from 11 steps to 5 focused steps that teach the core workflow
- Pipeline Visibility — Toast notifications after uploads with direct navigation to the created note or document
- Customizable Sidebar — Hide/show nav items via a Customize button; persists to localStorage; works on mobile too
- Unified Collections — Notes and Documents now share one collection system; add both to the same collection with cross-counts
- Mobile Navigation — Bottom tab bar with "More" drawer for secondary items, swipe navigation between panels
- Removed Daily Notes from Notes sidebar — Journal is now the single entry point for daily notes
- Dashboard Home Page — Knowledge base overview with customizable widgets (metrics, Brain Focus, Quick Capture, Calendar, Favorite Images, Tasks, Knowledge Graph stats, System Status)
- Custom Model Selection — Pick your own Ollama models per capability (vision, RAG, Brain) from Settings
- Data Export — Export your notes (Markdown), images, tags, and activity as a downloadable ZIP archive
- Nginx Reverse Proxy — Production-ready deployment with proper routing for 2FA, login, and API
- Self-Hosted Docker Setup — Single
docker-compose updeploys everything including Nginx on port 80
- Cloud AI Providers — Connect Claude, ChatGPT, or any OpenAI-compatible endpoint
- Encrypted API Key Management — Fernet-encrypted per-user key storage
- Token Usage & Cost Tracking — Monitor your cloud API spend
- Provider-Aware Model Registry — Cloud models alongside local Ollama
- Auto-expanding Chat Input — Textarea grows with your message
- Smart note-taking with wikilinks, hashtags, and backlinks
- AI image analysis with Qwen 2.5 VL vision model
- PDF document management with AI enrichment and review workflow
- Document collections for organizing PDFs
- Daily journal with calendar, mood tracking, and AI insights
- Knowledge graph with community detection, semantic edges, and path finding
- Dual-mode AI chat — NEXUS RAG retrieval + Brain AI companion
- Smart gallery with albums and tag filtering
- Smart buckets with AI clustering and orphan detection
- Semantic search across notes, images, and documents (pgvector)
- Neural Glass design system
- User onboarding flow
- Two-factor authentication (TOTP) and session management
- Import from Obsidian, Notion, Evernote
- Mobile-responsive design improvements
- Browser extension for web clipping
- Audio note transcription
- Mobile App — Your brain in your pocket
- Collaborative Spaces — Shared knowledge bases
- Plugin System — Extend with community add-ons
Have ideas? Open an issue — we're building this together!
Mnemosyne is open source and we love contributions! Whether it's:
- Bug reports
- Feature suggestions
- Documentation improvements
- Code contributions
Check out our Contributing Guide to get started.
Mnemosyne is built with love by a solo developer. If it helps you build your second brain, consider supporting continued development:
Other ways to help:
- Star this repo — It helps others discover Mnemosyne
- Report bugs — Help us improve
- Spread the word — Tell your friends about private AI
Mnemosyne is open source software licensed under the GNU Affero General Public License v3.0.
Mnemosyne — Named after the Greek goddess of memory
Your thoughts. Your images. Your documents. Your knowledge. Your brain.
Built with care by @Simmak7
