Project Introduction: Anything AI
Anything AI is an innovative automation framework designed for users seeking an intuitive and comprehensive solution for their automation needs. Developed in the fast and efficient Rust programming language, this tool integrates full-stack capabilities and artificial intelligence to facilitate smooth and powerful automated workflows.
Overview
Anything AI stands apart as a robust platform that could be likened to a collaboration between Posthog's user-friendly analytics and Zapier's seamless automation. The project aims to provide users with a rich feature set and an extensive framework for creating and managing automated workflows without compromising ease of use.
Features
-
Workflow Rest API: Using Axum, Anything AI offers a fully functional REST API for managing workflows. Users can perform CRUD (Create, Read, Update, Delete) operations through Postgrest over the Supabase platform. Additionally, the API supports workflow version management and publishing, ensuring workflows are efficiently tracked and deployed.
-
Durable Workflow Processing: The system is equipped to handle complex workflow traversal and task planning. Users can bundle actions using Tera for configuration. The task queue, also managed via Postgrest and Supabase, processes tasks systematically, facilitating timely execution of multiple operations. It includes a trigger management system with cron triggers, while webhook and polling triggers are under development.
-
Authentication and Authorization: Employing Supabase for user management, Anything AI supports OAuth integration and incorporates a team authorization and billing system. Using the BaseJump template with PostgreSQL's Row Level Security, it ensures user data is securely managed.
-
Workflow Studio: The platform includes a Workflow Studio with a drag-and-drop editor powered by ReactFlow. Users can configure actions using json-schema-form, while the UI is styled with TailwindCSS and ShadCN. Workflow testing management features are present, with some functionalities like single action testing and template management still on the roadmap.
-
Template Marketplace: Users can create profiles, and publish their workflow and action templates, fostering a community-driven environment where creativity and collaboration thrive.
Architecture Goals
Anything AI is designed with flexibility and comprehension in mind. It aspires to be an open tool that supports creativity and extensibility without losing user-friendliness. The architecture encourages the gradual integration of new AI technologies. Self-authoring capabilities are a cornerstone of the platform, aiming to maintain data in human-centric formats that are easily processed by low-cost language models (LLMs).
Extensibility
Anything AI’s system extensively supports plugin usage. Each action type within the framework is defined by plugins, similar to how tasks are managed in VSCode. Users are encouraged to develop and share their custom plugins, which are built using WebAssembly (WASM) and can be crafted in any language compatible with WASM.
User-Friendly Interface
The interface is crafted to be self-descriptive, allowing users to discern the purpose of workflows at a glance rather than focusing on the technical details. Configuration is facilitated through templating arguments, leveraging previous results, environment variables, and system constants.
Core Team
Carl Lippert spearheads the project and brings expertise and vision in crafting a user-focused automation platform.
Through Anything AI, users are empowered with a versatile framework that supports an imaginative and user-centric approach to building automated processes, paving the way for accessible and impactful AI-assisted exploits.