🎼 Workflow: Orchestration & Global Routing

The system does not use fixed n8n workflows for the entire process. Instead, it employs data-driven Dynamic Orchestration.

Global Routing Table (GRT)

The GRT is the core concept. It is a map (JSON Object) that defines the entire structure of the process. The GRT is loaded into the task at the very first step (Root Task) and is transmitted intact to all descendant tasks.

GRT Structure

{
  "step_1_syllabus": {
    "template_id": "tpl_syllabus_gen",
    "next_ids": ["step_2_lesson_plan"]
  },
  "step_2_lesson_plan": {
    "template_id": "tpl_lesson_plan_gen",
    "cardinality": "one_to_many",
    "split_path": "result.lessons",
    "next_ids": ["step_3_slide", "step_3_quiz"]
  },
  "step_3_slide": {
    "template_id": "tpl_slide_gen",
    "next_ids": []
  },
  "step_3_quiz": {
    "template_id": "tpl_quiz_gen",
    "next_ids": []
  }
}

Operational Mechanism

1. Hydration (In Phase 5):

   • When a task completes, Phase 5 examines its current_stage_config.
   • It retrieves the list of next_ids (e.g., ["step_2_lesson_plan"]).
   • For each next ID, it lookups the grt to obtain the detailed configuration for that step (template, split mode, etc.).
   • It injects this configuration into the current_stage_config of the newly created child task.

2. Propagation:

   • Child tasks receive the grt from their parent tasks.
   • This process repeats until next_ids is empty, marking the end of the chain.

Forking Patterns

1. Linear

• Config: cardinality: "one_to_one"
• Description: Task A completes -> Task B is created. The result data from A becomes the input_data for B.

2. Fan-out

• Config: cardinality: "one_to_many"
• Description: Task A returns an array (e.g., a list of 5 lessons). The system creates 5 B Tasks running in parallel.
• Split Path: Defines the path to the array within the JSON (e.g., result.modules).
• Split Mode:
  • per_item: Each array element generates one task.
  • per_batch: Groups elements together (e.g., 10 items per task).

3. Parallel Branches

• Config: next_ids: ["branch_A", "branch_B"]
• Description: Task A completes -> Simultaneous creation of Task B1 (Slides) and Task B2 (Quiz). These two branches run independently.

Advantages

• Flexibility: To modify a process (add or reorder steps), simply update the JSON config (GRT) in the Database/Preset, without needing to edit n8n workflows.
• Stateless: Each task contains sufficient information to determine the next step, eliminating the need for a massive stateful “Master Workflow” for coordination.