ws-lang

README

ws-lang — Universal Capability Orchestration Language

ws-lang is a domain-specific orchestration language for composing capabilities (voice processing, image processing, search, notification, etc.) into executable instruction sets. It is designed for LLM-native generation — the language is optimized so that LLMs can output correct syntax with near-zero error rates.

🌍 Multi-Language by Design

ws-lang is not an English-only language. Keywords can be written in any human language, and a deterministic Translation Layer maps them to a canonical internal form.

10 languages currently supported:

Language	Code	Example
🇬🇧 English	en (default)	task, step, cap, retry, merge
🇨🇳 Chinese	zh	任务, 步骤, 能力, 重试, 合并
🇯🇵 Japanese	ja	タスク, ステップ, 呼出, リトライ, 結合
🇪🇸 Spanish	es	tarea, paso, capacidad, reintentar, combinar
🇫🇷 French	fr	tâche, étape, capacité, réessayer, fusionner
🇵🇹 Portuguese	pt	tarefa, etapa, capacidade, retentar, mesclar
🇷🇺 Russian	ru	задача, шаг, возможность, повторить, объединить
🇮🇳 Hindi	hi	कार्य, चरण, क्षमता, पुनःप्रयास, विलय
🇧🇩 Bengali	bn	কাজ, ধাপ, ক্ষমতা, পুনরায়_চেষ্টা, একত্রিত
🇸🇦 Arabic	ar	مهمة, خطوة, قدرة, إعادة_محاولة, دمج

# English
task "image-pipeline" {
    step download { cap: ws-storage }
}

# 中文
任务 "图像处理" {
    步骤 下载 { 能力: ws-storage }
}

# 日本語
タスク "画像処理" {
    ステップ ダウンロード { 呼出: ws-storage }
}

# Español
tarea "procesar-imágenes" {
    paso descargar { capacidad: ws-storage }
}

# Français
tâche "traitement-images" {
    étape télécharger { capacité: ws-storage }
}

# العربية
مهمة "معالجة-الصور" {
    خطوة تحميل { قدرة: ws-storage }
}

🧩 Architecture

User Intent
  → Frontend Program (LLM)
      → KV Line Format (LLM-native)
          → Parser + Validator
              → YAML Instruction Set (canonical)
                  → ws-tasks Engine (execution)

✨ Key Features

• KV Line Format — LLM-optimized syntax: no brackets, no commas, no quotes to mismatch
• Deterministic Translation — Multi-language support via lookup tables, not AI (zero ambiguity)
• DAG-native — Steps automatically form dependency graphs; parallel branches execute concurrently
• Capability Abstraction — Each capability (ws-voice, ws-image, etc.) is a self-describing binary with a .cap.yaml manifest
• Self-documenting — YAML internal format supports comments for execution audit trails

📦 Repository Structure

ws-lang/
├── parser/          # KV line format parser (Go)
├── compiler/        # KV → YAML compiler
├── schema/          # JSON Schema for instruction sets
├── translations/    # Multi-language keyword mappings (*.json + generated translations.go)
├── examples/        # Example instruction sets (10 languages)
│   ├── meeting-notes.ws       # English
│   ├── meeting-notes.ws.zh    # Chinese
│   ├── spanish-test.ws        # Spanish
│   ├── arabic-test.ws         # Arabic
│   ├── french-test.ws         # French
│   ├── portuguese-test.ws     # Portuguese
│   ├── russian-test.ws        # Russian
│   ├── hindi-test.ws          # Hindi
│   ├── bengali-test.ws        # Bengali
│   └── japanese-test.ws       # Japanese
└── docs/            # Documentation

🚀 Quick Start

# Parse a KV format instruction set
go run parser/parser.go examples/meeting-notes.ws

# Compile to canonical YAML
go run compiler/compiler.go examples/meeting-notes.ws

# With translation (Chinese keywords → canonical)
go run compiler/compiler.go -lang zh examples/meeting-notes.ws.zh

# With translation (Spanish keywords → canonical)
go run compiler/compiler.go -lang es examples/spanish-test.ws

# With translation (Arabic keywords → canonical)
go run compiler/compiler.go -lang ar examples/arabic-test.ws

# List all available languages
ls translations/*.json

📖 Learn More

• Specification
• KV Line Format
• Translation Layer
• Architecture

🤝 Contributing

Contributions are welcome! Add a new language translation, improve the parser, or suggest capability manifests.

📄 License

MIT