llml CLAUDE.md

LLML Project - Claude AI Documentation

Core LLML Specification

Read .cursor/rules/spec.mdc

Coding Rules

Reference .cursor/rules/{language = ["py", "ts", "go", "rs"]}.mdc

Project Overview

LLML (Lightweight Language Markup Language) is React for Prompts - a multi-language compositional primitive that revolutionizes AI context engineering. Just as React transformed web development by making complex UIs composable and maintainable, LLML transforms AI development by making complex contexts composable and maintainable.

The project provides identical functionality across four programming languages: Python, TypeScript/JavaScript, Rust, and Go, converting nested data structures into human-readable, XML-like markup optimized for AI model attention.

Core Purpose

• Compositional Context Engineering: Build complex AI contexts from simple, reusable components
• Declarative AI Interactions: Describe what your prompt should contain, not how to format it
• Maintainable AI Systems: Replace brittle string concatenation with robust data composition
• Structured Document Creation: Generate documents with clear hierarchy from data structures
• Data Serialization: Convert data into more readable format than JSON/YAML for AI applications

Key Benefits

• Component-Like Composition: Build complex prompts from simple, reusable pieces
• Declarative Approach: Focus on what you want, not how to format it
• Maintainable & Robust: Changes to data automatically propagate without breaking
• Zero Configuration: Works out-of-the-box with sensible defaults
• Consistent Output: Identical results across all language implementations
• LLM-Optimized: Structured format reduces AI model cognitive load and improves performance
• Developer Experience: Type-safe, predictable, debuggable context engineering
• Extensible Formatters: Customizable formatter system like React's component system

Project Structure

/Users/knrz/Git/zenbase-ai/llml/
├── README.md                 # Main project documentation
├── justfile                  # Cross-language task runner
├── py/                       # Python implementation
│   ├── README.md
│   ├── pyproject.toml
│   ├── src/
│   │   └── zenbase_llml/
│   │       ├── llml.py      # Main implementation
│   │       └── formatters/  # Formatter system
│   │           ├── base/    # Base type formatters
│   │           └── vibe_xml/ # VibeXML formatters
│   └── tests/
├── ts/                       # TypeScript implementation
│   ├── README.md
│   ├── package.json
│   ├── src/
│   │   ├── index.ts         # Main implementation
│   │   └── formatters/      # Formatter system
│   │       ├── base/        # Base type formatters
│   │       └── vibe-xml/    # VibeXML formatters
│   └── tests/
├── rs/                       # Rust implementation
│   ├── README.md
│   ├── Cargo.toml
│   ├── src/
│   │   ├── lib.rs           # Main implementation
│   │   └── formatters.rs    # Core formatting logic
│   └── tests/
└── go/                       # Go implementation
    ├── README.md
    ├── go.mod
    ├── pkg/llml/
    │   └── llml.go          # Main implementation
    └── tests/