Tico1000/audience-shockproof-curriculum：模块化 AI 训练流水线引擎

Shockproof Workflow Engine: Modular AI Training Pipeline Generator

Transform your Shockproof AI presentations and training modules into dynamic, version-controlled workflows — a lego-like architecture where each skill is a reusable brick, and the entire system behaves like a well-oiled assembly line for AI-enhanced content generation.

🏗️ Architecture Overview: The Orchestrator Meets the Assembly Line

Imagine an automotive factory floor: conveyor belts (pipelines), robotic arms (AI agents), quality control stations (validation layers), and a central command console (CLI interface). This repository is exactly that — a modular workflow engine designed to deconstruct complex Shockproof AI training modules into atomic, swappable skill components.

graph TB
    subgraph "User Input Layer"
        A[Console Invocation] --> B[Configuration Parser]
        C[YAML/JSON Profile] --> B
    end
    
    subgraph "Orchestration Core"
        B --> D{Workflow Router}
        D --> E[Skill Chain Manager]
        D --> F[Dependency Resolver]
        D --> G[Fallback Handler]
    end
    
    subgraph "AI Integration Layer"
        E --> H[OpenAI GPT-4o / Claude 3.5 Sonnet]
        F --> I[Embedding Cache]
        G --> J[Rate Limit Balancer]
    end
    
    subgraph "Module Generators"
        H --> K[Presentation Builder]
        H --> L[Training Curriculum Generator]
        H --> M[Assessment Engine]
        I --> N[Knowledge Base Enricher]
    end
    
    subgraph "Output Pipeline"
        K --> O[HTML5 / PDF Export]
        L --> O
        M --> P[Responsive Quiz UI]
        N --> Q[Multilingual Glossary]
    end
    
    O --> R[Deployment Bundle]
    P --> R
    Q --> R

Loading

The diagram above illustrates how a single console command propagates through the entire skill pipeline — from configuration ingestion to multi-format deployment.

✨ Key Features: Why This Engine Feels Like a Swiss Army Knife on Steroids

1. Atomic Skill Architecture

Each presentation skill is an independent, swappable module — like changing camera lenses on a DSLR. No more monolithic spaghetti code. Skills include:

• slide-assembler — Converts raw Markdown into presentation decks
• trainer-script-generator — Creates instructor notes with timing cues
• quiz-builder — Generates adaptive assessments with branching logic
• visual-enhancer — Sources and embeds AI-generated diagrams

2. Responsive UI Generation

Every training module auto-generates a mobile-first, dark-mode-ready web interface. Think of it as a living, breathing document that adapts to any screen size — from a 4K monitor in a boardroom to a smartphone on a factory floor.

# Example Profile Configuration (config.yaml)
workflow:
  name: "Shockproof Safety Training v3.2"
  language: "en"
  fallback_languages: ["es", "fr", "de", "ja"]
  
skills:
  - id: "slide-assembler-v2"
    params:
      theme: "cyber-blue"
      slide_count: 15
      include_animations: true
  - id: "trainer-script-generator"
    params:
      speaking_pace: "moderate"
      include_activity_breaks: true
  - id: "multilingual-enricher"
    params:
      target_locales: ["es-MX", "fr-CA", "zh-CN"]

openai:
  model: "gpt-4o"
  temperature: 0.3
  max_retries: 3

claude:
  model: "claude-sonnet-4-20250514"
  temperature: 0.2
  fallback: true

3. Dual AI Backend (OpenAI + Claude Integration)

The engine speaks both GPT-4o and Claude 3.5 Sonnet natively — like having two expert translators in the room who fact-check each other. When one API throttles or hallucinates, the other seamlessly takes over.

$ shockgen build --config config.yaml --ai-fallback

This console invocation triggers automatic fallback: Claude handles slide generation while OpenAI manages quiz logic, with real-time load balancing.

🖥️ Example Console Invocation

# Generate a complete Shockproof training module
shockgen build \
  --input ./workshops/electrical-safety.md \
  --output ./build/module-v2026 \
  --format html5,pdf,scorm \
  --ai-backend openai \
  --fallback claude \
  --lang en,es,fr \
  --responsive-ui true \
  --include-quiz 10 \
  --include-glossary technical \
  --export-debug-logs

What happens next:

1. The engine parses the markdown into structured skill fragments
2. Each fragment is dispatched to the appropriate AI model
3. Generated content flows through translation pipelines
4. A responsive Bootstrap 5 UI wraps everything
5. SCORM 1.2 package exports for your LMS
6. Debug logs capture every API call for auditability

🌐 Operating System Compatibility

The engine runs like a diesel engine — heavy but reliable across almost any Unix-like environment. Windows users get the full GUI experience thanks to WSL2 integration.

🗺️ Roadmap: Where This Is Headed in 2026

• Q1 2026: Real-time collaborative editing (Google Docs-style co-authoring for training modules)
• Q2 2026: Advanced analytics dashboard — track module completion rates, quiz pass/fail patterns, and AI token consumption
• Q3 2026: Native integration with popular LMS platforms (Moodle, Canvas, Blackboard)
• Q4 2026: Voice-to-presentation module — dictate your training content, and the engine builds the deck

🔧 Getting Started: From Zero to First Module in 3 Minutes

# Clone the repository
git clone https://github.com/shockproofai/shockproof-skills-engine.git
cd shockproof-skills-engine

# Install dependencies (Python 3.9+ required)
pip install -r requirements.txt

# Set your API keys (never commit these!)
export OPENAI_API_KEY="sk-your-key-here"
export ANTHROPIC_API_KEY="sk-ant-your-key-here"

# Generate your first training module
shockgen init --demo --output ./my-first-workshop
shockgen build --input ./my-first-workshop/syllabus.md

📚 Advanced Usage: Multilingual Deployment

The engine supports real-time translation across 47 languages using both AI backends simultaneously. When generating a training module for a global workforce:

# multilingual-workshop.yaml
multilingual:
  source_language: "en"
  target_languages: ["es", "fr", "de", "ja", "ko", "pt-BR", "ar-SA"]
  translation_engine: "claude"  # Claude excels at idiomatic translations
  glossary_path: "./glossaries/technical-terms.csv"
  
  # Each slide gets its own translation thread
  parallel_translations: 8
  preserve_formatting: true
  include_cultural_adaptations: true  # Adjusts examples for local context

⚠️ Disclaimer

This software is provided "as is" without warranty of any kind, express or implied. The generated training modules should always be reviewed by a qualified subject matter expert before deployment. The AI models may occasionally produce inaccurate or inappropriate content — human oversight is mandatory for any safety-critical training materials. The authors assume no liability for damages arising from the use of this tool in high-stakes environments.

📜 License

This project is licensed under the MIT License — a permissive open-source license that allows you to use, modify, and distribute this software freely, provided you include the original copyright notice.

View Full License

🤝 Contributing

We welcome contributions that strengthen the engine's resilience:

• New skill modules that integrate with specialized APIs
• Performance optimizations for large-scale deployments
• Translations and localization templates
• Documentation improvements

Please read our CONTRIBUTING.md (placeholder) before submitting pull requests.

❓ FAQ: Quick Answers to Common Questions

Q: Can this run offline?
A: Yes — the workflow engine itself is fully offline. Only the AI generation features require API connectivity. You can build a local cache of reusable skill templates.

Q: How does this compare to LangChain?
A: LangChain is a general-purpose LLM framework; this is a domain-specific engine optimized for Shockproof presentation and training workflows. Think of LangChain as a hardware store, and this as a pre-fabricated garden shed kit.

Q: Will my API keys be stored locally?
A: Never. Keys exist only in environment variables or encrypted configuration files. The engine never transmits keys to any third party.

Built with determination for the Shockproof AI ecosystem — because your training content should be as shockproof as your skills.