Skip to main content
Multi-agent blog writing system using LangGraph workflow.

​
Code

from bindu.penguin.bindufy import bindufy
from graph import build_graph
from schemas import AgentResponse

graph = build_graph()

def handler(messages):
    try:
        # Handle possible dict wrapper
        if isinstance(messages, dict) and "messages" in messages:
            messages = messages["messages"]

        if not messages:
            raise ValueError("No messages received")

        last_message = messages[-1]

        # Support both formats:
        # 1) [{"role": "user", "content": "..."}]
        # 2) ["plain string"]
        if isinstance(last_message, dict):
            query = last_message.get("content", "")
        else:
            query = str(last_message)

        result = graph.invoke({
            "topic": query,
            "plan": None,
            "sections": [],
            "final": None
        })

        return result["final"]

    except Exception as e:
        return AgentResponse(
            answer="Agent execution failed.",
        )

config = {
    "author": "amritanshu9973@gmail.com",
    "name": "langgraph_blog_writing_agent",
    "deployment": {
        "url": "http://localhost:3773",
        "expose": True,
        "cors_origins": ["*"],
    },
    "skills": ["skills/blog_writing_agent"],
}

bindufy(config, handler)

​
How It Works

Agent Roles
  • Orchestrator: Breaks topic into structured plan with sections and word counts
  • Workers: Write individual sections simultaneously with specific technical depth
  • Reducer: Aggregates sections into final cohesive markdown article
Map-Reduce Pattern
  • Orchestrator creates detailed plan with specific tasks
  • Fanout distributes tasks to parallel workers
  • Workers write sections simultaneously ensuring constraints
  • Reducer combines sections into final article
State Management
  • topic: User input for blog topic
  • plan: Structured outline with sections and requirements
  • sections: Individual written sections from workers
  • final: Completed markdown article
Execution Flow
  1. Orchestrator breaks topic into detailed plan
  2. Workers write sections in parallel
  3. Reducer aggregates and formats final article
  4. Returns cohesive blog post

​
Run

# Set API key
echo "OPENAI_API_KEY=your-key-here" > examples/langgraph_blog_writing_agent/.env

# Run blog writing agent
uv run examples/langgraph_blog_writing_agent/main.py
Try: “Write a blog post about the future of artificial intelligence in healthcare” Go to frontend and run npm run dev Open http://localhost:5173 and try to chat with the LangGraph blog writing agent

​
Graph Structure

The agent uses a LangGraph workflow with:
  • orchestrator node for planning
  • worker nodes for parallel section writing
  • reducer node for final aggregation
  • State management across workflow steps