Content Management System - Agentic Pipeline

Trying out Langgraph's Orchestrator-worker pattern with a mini project where the agentic system generates product content pages given a product data JSON object. This agent uses a multi agent Orchestrator-Worker pattern provided by Langgraph to coordinate specialized workers that handle question generation, page assembly, and content synthesis.

Features

Parse & Understand Product Data - Converts raw product data into structured format
Auto-Generate 15+ Categorized Questions - Creates informational, safety, usage, purchase, and comparison questions
Custom Templates - FAQ Page, Product Description Page, Comparison Page
Content Logic Blocks - Reusable tools that transform data into compelling copy
Agent Orchestration - LangGraph-based workflow with orchestrator and worker agents
JSON Output - Clean, machine-readable output for all pages
Claude AI Integration - Uses Claude Sonnet 4.5 for intelligent content generation

Architecture

┌─────────────────┐
│  START          │
└────────┬────────┘
         │
         ▼
┌─────────────────┐
│ Parse Product   │ ← Converts raw data to structured model
└────────┬────────┘
         │
         ▼
┌─────────────────┐
│ Orchestrator    │ ← Plans what content to generate
└────────┬────────┘
         │
         ├──────────────┬─────────────┐
         ▼              ▼             ▼
┌──────────────┐ ┌──────────────┐ ┌──────────────┐
│ Question Gen │ │ Page Gen     │ │ ...          │
│ Worker       │ │ Worker       │ │              │
└──────┬───────┘ └──────┬───────┘ └──────┬───────┘
       │                │                │
       └────────────────┼────────────────┘
                        ▼
                ┌──────────────┐
                │ Synthesizer  │ ← Combines all outputs
                └──────┬───────┘
                       │
                       ▼
                   ┌───────┐
                   │  END  │
                   └───────┘

Setup

1. Install Dependencies

pip install -r requirements.txt

2. Configure API Key

Create a .env file in the project root:

ANTHROPIC_API_KEY=your_claude_api_key_here

Get your API key from: https://console.anthropic.com/

3. Run the System

python -m main

Output

The system generates three JSON files in the output/ directory:

• faq.json - FAQ page with Q&As
• product_page.json - Product description page
• comparison_page.json - Product comparison page

Project Structure

content_management_system/
├── src/
│   ├── services/
│   │   ├── agents/           # Agent implementations
│   │   │   ├── base.py
│   │   │   ├── parse_product.py
│   │   │   ├── question_generator.py
│   │   │   ├── page_generation.py
│   │   │   └── synthesizer.py
│   │   ├── graph/            # LangGraph workflow
│   │   │   ├── workflow.py
│   │   │   └── orchestrator.py
│   │   ├── tools/            # Content logic blocks
│   │   │   ├── benefits_tool.py
│   │   │   ├── usage_tool.py
│   │   │   └── comparision_tool.py
│   │   └── llm_service.py    # Claude API wrapper
│   └── utils/
│       ├── state.py          # State management
│       └── input_product.py  # Sample product data
├── output/                   # Generated JSON files
├── main.py                   # Entry point
├── requirements.txt
└── .env                      # API keys (not in git)

How It Works

1. Product Parsing

Converts raw product data into a clean internal model.

2. Question Generation (LLM)

Uses Claude to generate 15+ categorized questions across:

• Informational
• Safety
• Usage
• Purchase
• Comparison

3. Content Logic Blocks (Tools)

Reusable functions that transform data:

• BenefitsTool: Enhances benefits into compelling copy
• UsageTool: Creates detailed step-by-step instructions
• ComparisonTool: Generates product comparisons with analysis

4. Page Assembly (LLM)

Agents autonomously produce:

• FAQ Page: 5+ Q&As with detailed answers
• Product Page: Enhanced with marketing copy
• Comparison Page: Side-by-side with competitor

Technologies

• LangGraph: Agent orchestration framework
• Claude Sonnet 4.5: LLM for content generation
• Python 3.12+: Core language
• Anthropic API: Claude API integration