Skip to main content
The Supervisor worker orchestrates agent execution, manages project state, and handles human-in-the-loop interrupts.

​
Architecture

​
Environment Bindings

interface Env {
  PRD_AGENT: Fetcher;
  DATA_AGENT: Fetcher;
  DESIGN_AGENT: Fetcher;
  LOGIC_AGENT: Fetcher;
  API_AGENT: Fetcher;
  FRONTEND_AGENT: Fetcher;
  DEPLOYMENT_AGENT: Fetcher;
  DB: D1Database;
  STORAGE: R2Bucket;
}

​
Key Functions

​
Project Management

List Projects 
GET /api/projects
// Returns all projects ordered by updated_at
Create Project 
POST /api/projects
// Creates new project with product seed
Hydrate Project 
GET /api/hydrate/:projectId
// Returns complete project state with history

​
Agent Execution

executeParallelNodes Executes agents in parallel (fan-out/fan-in): 
async function executeParallelNodes(
  state: ArchitectureState,
  env: Env
): Promise<ArchitectureState> {
  // 1. Map agents for parallel execution
  // 2. Execute in parallel (Promise.all)
  // 3. Merge results (fan-in)
  // 4. Return updated state
}
executeSequentialNodes Executes agents sequentially: 
async function executeSequentialNodes(
  state: ArchitectureState,
  env: Env
): Promise<ArchitectureState> {
  // Execute agents one by one
  // Wait for each to complete
  // Pass context to next agent
}

​
Streaming

handleStreamRequest Handles Server-Sent Events for real-time updates: 
async function handleStreamRequest(
  request: Request,
  env: Env
): Promise<Response> {
  // Create TransformStream for SSE
  // Stream agent progress events
  // Stream completion events
}

​
Context Assembly

assembleContext Assembles context from previous artifacts: 
async function assembleContext(
  env: Env,
  projectId: string
): Promise<string> {
  // Fetch previous artifacts
  // Combine into context string
  // Return formatted context
}

​
Database Schema

​
projects Table

CREATE TABLE projects (
  id TEXT PRIMARY KEY,
  name TEXT NOT NULL,
  product_seed TEXT,
  current_phase TEXT,
  status TEXT,
  created_at INTEGER,
  updated_at INTEGER
);

​
artifacts Table

CREATE TABLE artifacts (
  id TEXT PRIMARY KEY,
  project_id TEXT,
  phase TEXT,
  type TEXT,
  content TEXT,
  created_at INTEGER
);

​
execution_traces Table

CREATE TABLE execution_traces (
  id TEXT PRIMARY KEY,
  project_id TEXT,
  phase TEXT,
  agent TEXT,
  status TEXT,
  created_at INTEGER
);

​
Storage Strategy

  • D1 Database: Projects, artifacts, execution traces
  • R2 Storage: Large artifacts (if content > threshold)
  • Artifact Keys: artifacts/{projectId}/{phase}/{artifactId}

​
Human-in-the-Loop

The supervisor checks for interrupt signals: 
if (requiresInterrupt(agentResponse)) {
  // Set status to 'interrupted'
  // Store checkpoint
  // Wait for human approval
}
Interrupt points:
  • After Logic Agent (before API design)
  • After API Agent (before Frontend design)
  • After critical phases