Fabrice AI is a TypeScript framework for building composable AI agents designed for serverless environments. It targets developers seeking a functional, infrastructure-agnostic approach to orchestrating multiple AI agents for complex tasks, simplifying AI agent development without requiring deep AI expertise.

How It Works

Fabrice AI utilizes a functional programming paradigm, emphasizing statelessness and composability. Agents are defined as specialized functions with roles and descriptions, capable of utilizing tools. Workflows orchestrate these agents, defining team members, task descriptions, and expected outputs. The framework manages workflow states as a tree, allowing for delegation and handoffs between agents. Providers abstract LLM interactions, with built-in support for OpenAI and Groq, and custom providers are easily implementable.

Quick Start & Requirements

• Install via npx create-fabrice-ai or npm install fabrice-ai.
• Requires Node.js and TypeScript.
• Example usage: bun your_file.ts or node --import=tsx your_file.ts.
• Official creator tool: https://github.com/callstackincubator/fabrice-ai/tree/main/packages/create-fabrice-ai

Highlighted Details

• Functional and stateless design principles.
• Infrastructure-agnostic, supporting serverless and various LLM providers.
• Built-in agents (Supervisor, Resource Planner, Final Boss) and tools.
• Extensible with custom agents, workflows, providers, and tools using Zod for schema validation.
• BDD testing utilities available via @fabrice-ai/bdd.

Maintenance & Community

• Developed by Callstack.
• Project is open source and free to use.
• Community engagement encouraged via GitHub stars and contact at hello@callstack.com.

Licensing & Compatibility

• MIT License.
• Permissive license suitable for commercial use and integration into closed-source projects.

Limitations & Caveats

The framework is relatively new, and while it aims for simplicity, the full scope of managing complex, long-running, or stateful workflows might require careful design and testing. The "batteries included" approach means relying on the provided abstractions, which might limit deep customization of core execution logic without understanding the internal teamwork and iterate functions.