ws-workflow-engine-bridge

Author Workflow Engine Bridge

Import prompt sequence JSON files from the Author Workflow Engine and execute them as AuthorClaw automations.

What This Does

The Author Workflow Engine (AWE) lets you build multi-step prompt chains — like "research → premise → outline → write chapter 1." This bridge imports those chains and runs them inside AuthorClaw with full access to your book bible, voice profile, and memory.

JSON Schema Expected

{
  "name": "Novel Outline Generator",
  "description": "Takes a premise and generates a complete outline",
  "version": "1.0",
  "steps": [
    {
      "id": "step-1",
      "name": "Premise Refinement",
      "prompt": "Take this premise and refine it: {{INPUT}}. Make the stakes personal and the conflict built-in.",
      "model": "auto",
      "temperature": 0.7,
      "outputVariable": "refined_premise"
    },
    {
      "id": "step-2",
      "name": "Character Development",
      "prompt": "Based on this premise: {{refined_premise}}\n\nCreate 3-5 main characters with names, roles, motivations, and flaws.",
      "model": "auto",
      "temperature": 0.8,
      "outputVariable": "characters"
    },
    {
      "id": "step-3",
      "name": "Three-Act Outline",
      "prompt": "Premise: {{refined_premise}}\nCharacters: {{characters}}\n\nCreate a detailed three-act outline with chapter breakdowns.",
      "model": "auto",
      "temperature": 0.7,
      "outputVariable": "outline"
    }
  ],
  "variables": {
    "INPUT": { "type": "user_input", "prompt": "Enter your story premise:" }
  }
}

Execution Process

1. Import: Load AWE JSON into AuthorClaw's automation engine
2. Enhance: Automatically inject AuthorClaw context into each step:
   • Book Bible data (characters, locations, timeline)
   • Voice Profile (so output matches author's style)
   • Style Guide rules
   • Active project context
3. Execute: Run steps sequentially, passing output variables between steps
4. Route: Use AuthorClaw's smart AI routing per step:
   • "model": "auto" → Let AuthorClaw pick the best provider
   • "model": "creative" → Route to Claude or GPT-4o
   • "model": "fast" → Route to Ollama or Gemini
   • "model": "cheap" → Route to DeepSeek
5. Save: Store all outputs in the project workspace
6. Review: Present results with option to re-run any step