TongFlow is an open-source, multi-modal Generative AI workflow studio designed to simplify the creation of complex AI-driven content. It empowers users, from individuals to enterprises, to orchestrate various AI models and modalities through an intuitive, visual interface, enabling imaginative content generation with a low barrier to entry.

How It Works

TongFlow conceptualizes AI models as "modality transforms" (e.g., text-to-image, text-to-audio). Each capability is exposed as a node within a workflow canvas. The core user interaction revolves around three simple operations: "add" (input data/prompts), "transform" (apply AI models), and "combine" (integrate outputs). This approach, coupled with an open, plugin-based architecture, allows for extensive extensibility and flexibility, enabling users to freely arrange AI models to generate unique creations without needing to manage complex AI parameters directly.

Quick Start & Requirements

• Installation: Ready-to-run desktop applications are available for macOS (Apple Silicon/Intel) and Windows. Alternatively, users can install from source using pnpm install, pnpm plugins:install, pnpm start:prod, or deploy via Docker using docker run or docker compose.
• Prerequisites: Installation from source requires Node.js (with pnpm) and a Python 3.10+ interpreter. Docker deployment abstracts these.
• Dependencies: Plugins require credentials (API keys like OpenAI, Gemini, OpenRouter, or Modal tokens) configured via the in-app Settings or environment variables. Official plugins often leverage Modal for GPU compute.
• Setup: The desktop app requires installation. First-time plugin installations (local or Docker) may be slower due to automatic Python virtual environment provisioning.
• Links: Download installers from the Releases page. Plugin development details are in docs/plugins.md.

Highlighted Details

• Broad Modality Support: Handles text, images, audio, video, 3D models, and documents, offering capabilities like text-to-image/video, image editing, speech synthesis/recognition, lip-sync video generation, and single-image-to-3D model conversion.
• Advanced Workflow Examples: Demonstrates complex pipelines for creating talking-head avatars and music videos by composing multiple AI modalities.
• Extensible Plugin System: An open ecosystem allows any platform to develop and integrate plugins. Official plugins utilize Modal for GPU access (offering up to $30/month free compute) and integrate with major APIs like OpenAI and Google Gemini.
• Customization: The core ABI and SDK facilitate the development of custom plugins to integrate proprietary or novel AI models.

Maintenance & Community

• Community: Active community engagement is encouraged via Discord and a WeChat group.
• Business & Enterprise: TongFlow offers enterprise solutions, including local GPU deployment, custom node/plugin development, API integration, and partnerships for cloud-hosted AI studios. Contact business@tongflow.com for inquiries.
• Contributions: Project contributions are managed via a Contributor License Agreement (CLA).

Licensing & Compatibility

• License: TongFlow employs a dual-licensing model:
  • AGPL-3.0: Free for individuals, research, and open-source projects, but requires source code disclosure for network-accessible services (Section 13).
  • Commercial License: Available for organizations requiring closed-source or SaaS product integration without AGPL obligations, offering warranties and support. Contact business@tongflow.com for terms.
• Compatibility: AGPL-3.0 compliance is critical for SaaS deployments. Commercial licensing is necessary for closed-source integration.

Limitations & Caveats

• macOS Gatekeeper: Users on macOS may encounter Gatekeeper warnings on the first launch of the desktop app, requiring manual steps to bypass ("TongFlow is damaged...").
• Plugin Dependency: The application ships without pre-installed plugins; users must manually select and install desired functionalities via the plugin manager.
• Credential Management: Functionality is dependent on correctly configuring API keys and/or Modal tokens within the application's settings or environment variables.