AetherLang Ω

div align="center"

A powerful Domain-Specific Language for AI Workflow Orchestration
Documentation • Examples • Getting Started • Live Demo

/div

What is AetherLang?

AetherLang (Ω) is a production-ready Domain-Specific Language (DSL) meticulously crafted for constructing, visualizing, and executing intricate AI workflows. It combines the capabilities of platforms like Airflow and Prefect with a clean, declarative syntax tailored specifically for orchestrating Large Language Models (LLMs).

End-to-end AI workflow orchestration with visual debugging

Key Features

• 🎯 28 Specialized Node Types including Guards, LLMs, Retrieval-Augmented Generation (RAG), caching, validation, and more.
• 🔄 Asynchronous Execution Engine featuring built-in async/await capabilities and OpenAI integration.
• 🎨 Visual Flow Designer that allows interactive diagrams with real-time execution visualization.
• 🌐 Bilingual Support - Comprehensive documentation and syntax available in both Greek and English.
• ⚡ Live Execution Streaming powered by WebSocket, providing real-time node state updates.
• 🧪 Extensive Validation mechanisms including type checking, cycle detection, and semantic analysis.
• 🚀 Zero External Dependencies - A pure Python implementation, apart from the OpenAI SDK.
• 📊 Physics-Based Layout - Enables force-directed graph visualization with drag-and-drop functionality.
• 💾 Export & Import features for sharing and version control of workflows.
• 🔧 Monaco Editor Integration for an enhanced coding experience with IntelliSense support.
• 📈 Performance Profiling tools to identify bottlenecks and optimize workflow execution.
• 🧠 AI-Powered Optimization where GPT-4o analyzes workflows and recommends enhancements.

Visual Interface

Visual Debugging

AetherLang provides a professional-grade visual debugging interface that allows users to monitor workflows in real-time, examining performance metrics and obtaining insights throughout the execution process.

Example Workflows

Sample RAG Pipeline

flow RAGPipeline {
  using target "neuroaether" version ">=0.2";

  input text question;

  node Cache: cache ttl=3600;
  node Retriever: rag sources=["docs", "web"], top_k=5;
  node LLM: llm model="gpt-4o", temp=0.3;
  node Validator: validate schema="answer";

  Cache -> Retriever -> LLM -> Validator;

  output text answer from Validator;
}

Learning More

• Getting Started Documentation - A comprehensive introduction, installation guidelines, and basic usage.
• Syntax Reference - Complete reference for language syntax.
• Node Documentation - Detailed descriptions of all 28 supported node types.
• API Documentation - Information on the Python API for parser, runtime, and validation.

Explore the potential of AetherLang for streamlining AI workflows and enhancing productivity through its innovative approach to workflow orchestration.