Skip to content

About

mofa‑studio: All‑in‑one Tauri‑based desktop AI assistant for end‑users. Image, video, writing & visual workflow builder, with unique floating desktop‑pet interaction.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Repository files navigation

mofa-studio

🇬🇧 English | 中文

mofa-studio is a local-first, all-in-one AI creation workstation for creators, built with Tauri and an isomorphic Rust backend. It fuses three product paradigms into one desktop app:

  • Chat as Creation — multimodal conversational entry for Q&A, vision, image/video generation and writing
  • Task as Delivery — one-sentence tasks autonomously planned, executed, reviewed and delivered as documents/decks/media
  • Workflow as Production Line — node-based visual pipelines for batch content production

All model access is BYOK (Bring Your Own Key): keys live in OS keychains, data stays on-device, and there is no subscription paywall.

📋 Product requirements documents (bilingual): docs/prd/

Features

Three-Mode Workbench

  • Assistant — default home: streaming multi-model chat with deep-thinking mode, web search with citations, image understanding, in-chat image/video generation, voice input & TTS readout
  • Projects — kickoff → autonomous planning → single-agent execution → multi-expert review panel → acceptance & delivery, with breakpoint resume
  • Automation — unattended scheduled pipelines (cron triggers) that keep running in the tray after the window closes

Creation Toolbox (16 tools)

  • Image generation — text-to-image / image-to-image / masked inpainting, multi-reference consistency, batch production with per-platform size presets (Xiaohongshu 3:4, Douyin 9:16, Bilibili 16:9)
  • Video generation & media processing — text/image-to-video; video→GIF, batch image compression, transcoding with platform presets (ffmpeg)
  • Documents — AI writing with personal style library (TipTap), PPT generation, AI spreadsheet, deep research with cost estimates and cited reports
  • Audio — music generation, meeting transcription with speaker diarization, full podcast studio (script → multi-voice TTS → waveform editing → MP3 export)

Task Workbench

  • Five-concept system: Connectors (OAuth/MCP), Skills (self-contained mini-agent packs), Experts, Expert Teams, Inspiration marketplace ("make alike" one-click replication)
  • Deliverables center with unified preview (Word/Excel/PPT via Univer), file tree and change diffs
  • Authorized-directory local file operations, long-term memory system, SOP one-click deposit

Creation Workflow (ComfyUI-parity, lowered barrier)

  • Node canvas with generation node library, execution queue with signature-cache incremental execution
  • Model management center (cloud API + local Ollama dual track), template market with missing-dependency detection
  • App Mode: experts compose workflows, everyone else fills a simple form

Platform Foundation

  • Embedded Axum backend (same codebase compiles to a standalone server for future web/mobile)
  • llm-gateway normalizing 130+ model vendors with usage metering, quota alerts and key failover
  • Tracing observability: chat/task/workflow span instrumentation with a usage & cost dashboard

Desktop Companion

  • Draggable floating ball, screenshot-to-ask, text-selection Q&A, global hotkeys, tray residency
  • Legacy B-side admin modules (organization, audit, load testing, monitoring) tucked behind an opt-in Expert Mode

Tech Stack

Category Technology
Framework Tauri 2.x + React 19
Language TypeScript 5.8 / Rust
Build Tool Vite 8
UI Components Ant Design 6
Styling Tailwind CSS 4
State Management Zustand 5
Routing React Router 7
Workflow Graph XYFlow
Office Suite Univer (Sheets/Docs/Slides)
Backend (embedded) Axum · SQLite (+FTS5/sqlite-vec) · ffmpeg sidecar · OS keychain
Internationalization i18next

Project Structure

mofa-studio/
├── src/                    # Frontend source code
│   ├── components/         # Reusable components
│   ├── pages/              # Page components
│   ├── hooks/              # Custom hooks
│   ├── stores/             # Zustand state management
│   ├── services/           # API services (mock/real dual-track)
│   ├── types/              # TypeScript type definitions
│   ├── theme/              # Theme configuration
│   ├── i18n/               # Internationalization config
│   ├── floating/           # Floating window components
│   └── tracing/            # Conversation span tracing
├── src-tauri/              # Tauri shell (tray, floating ball, window mgmt)
│   └── crates/server-core/ # Embedded local-first backend (Axum + SQLite)
├── docs/
│   └── prd/                # Product requirements documents (en/zh)
└── public/                 # Static assets

The embedded Axum backend (src-tauri/crates/server-core) boots with the app on 127.0.0.1:<dynamic port>: local-first SQLite document storage behind the agentos-compatible REST surface, no-login local sessions, and a native WebSocket endpoint. The frontend resolves the server through the get_server_info Tauri command at startup. Deeper PRD capabilities (llm-gateway BYOK proxy, FTS5/sqlite-vec search) are still planned work.

Getting Started

Prerequisites

  • Node.js 18+
  • Deno (recommended) or pnpm/npm
  • Rust 1.90+
  • System dependencies see Tauri official docs

Install Dependencies

deno install

Development Mode

# Start frontend dev server
deno task dev

# Start Tauri dev mode (includes frontend)
deno task tauri-dev

Build & Release

# Build frontend
deno task build

# Build Tauri application
deno task tauri-build

Configuration

Environment Variables

Copy .env.example to .env and configure:

cp .env.example .env

Key configuration items:

  • VITE_APP_SERVER_URL - API base URL of your backend instance (browser dev only; inside the Tauri shell the embedded local backend is used automatically)
  • VITE_APP_ENABLE_MOCK - toggle mock data layer
  • VITE_APP_FLOATING_MODE - floating or window launch mode
  • VITE_APP_TITLE - Application title

License

This project uses a commercial source code license agreement. See the LICENSE file for details.

Important Notice:

  • This software is licensed for source code viewing only, for personal learning or research purposes
  • Copying, modification, distribution, or commercial use is prohibited
  • For commercial licensing, please contact the copyright holder

Acknowledgements

We would like to express our sincere gratitude to:

  • Dr. Wu - For invaluable guidance, support, and inspiration throughout the development of this project.
  • The mofa-org team - For their dedication, collaboration, and contributions that made this project possible.

About

mofa‑studio: All‑in‑one Tauri‑based desktop AI assistant for end‑users. Image, video, writing & visual workflow builder, with unique floating desktop‑pet interaction.

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages