Flowra - AI workflow development tool open-sourced by Magic Hitch and Wooli WULI team

What is Flowra?

Flowra is an open source graph execution engine and node package development tool from the ModelScope and WULI teams, and is a core component of FlowBench, which organizes workflows through a directed acyclic graph (DAG). Flowra organizes workflows through directed acyclic graphs (DAGs) with powerful features such as smart caching, parallel scheduling, distributed support, and error recovery to efficiently handle complex AI workflow tasks. Flowra has a rich set of built-in component types and value types to support a variety of multimedia data, such as images, videos, audios, 3D grids, and other multimedia data, as well as numeric inputs, selectors, and other front-end interactive components to ensure type safety and automatic runtime validation. Secure and automatic runtime validation. Flowra provides a complete development tool chain, from project creation to packaging and publishing, supporting local debugging and development environment configuration, greatly simplifying the development process of node packages.

Features of Flowra

• Well-established type system: Flowra comes with a rich set of built-in component types and value types that support a wide range of multimedia data types such as images, video, audio, 3D grids, and front-end interactive components such as numeric inputs, selectors, placeholders, etc., and are type-safe and automatically validated at runtime.
• DAG Execution Engine: A Directed Acyclic Graph (DAG) is used to organize workflows with core features such as intelligent caching, parallel scheduling, distributed support, and error recovery.
• Complete development tool chain: Provide tools for the whole process from project creation to package release, including creating new node package projects, managing node groups and nodes, building node packages, debugging nodes locally, and configuring the development environment.
• Seamless integration with ModelScopeBuilt-in support for ModelScope, which can download and manage AI models in one line of code, automatically download model files, provide a unified model management interface, and support model caching and version control.
• Flexible storage backend: Supports a variety of object storage services, such as AliCloud OSS, MinIO, local file system, large files and intermediate results can be automatically uploaded to the cloud to save local storage space, while supporting data sharing in a distributed environment.

Flowra's core strengths

• Well-established type systemRich components and value types are built-in, covering multimedia data such as images, videos, audios, 3D grids, and interactive components such as numeric inputs, selectors, etc., which are automatically validated at runtime to ensure type safety.
• Efficient DAG execution engine: Adopting a directed acyclic graph to organize the workflow, with features such as intelligent caching, parallel scheduling, distributed support, and error recovery, it can efficiently handle complex tasks.
• Complete development tool chain: Provide a full range of tools from project creation to packaging and publishing , support for local debugging and development environment configuration , to simplify the node package development .
• Seamless integration with ModelScopeAI models can be downloaded and managed with a single line of code, supporting model caching and version control for easy model integration.
• Flexible storage backendIt supports various storage services such as AliCloud OSS, MinIO, local file system, etc., which facilitates the sharing of large files and intermediate result storage.

What is the official Flowra website?

• GitHub repository:: https://github.com/modelscope/flowra

Who is Flowra for?

• AI developers: Flowra provides a rich system of tools and types that help AI developers rapidly build and deploy AI workflows, simplify model integration and debugging, and improve development efficiency.
• data scientist: Its powerful DAG execution engine and multimedia data support enable data scientists to efficiently process and analyze multiple data types, accelerating data-driven project development.
• research worker: Flowra's distributed support and intelligent caching capabilities provide researchers with powerful computational support for large-scale experiments and modeling studies.
• Corporate Technical Team: Enterprises can use Flowra's distributed and storage back-end capabilities to quickly build AI applications for efficient resource management and team collaboration.