🤖 LangGraph Blog Automation Agent

An autonomous, cloud-based content generation agent built with LangGraph. It reads blog topics from a Google Sheet, drafts SEO-optimized articles using LLMs, evaluates content quality using a hybrid evaluator (Heuristics + LLM-as-a-Judge), publishes the articles to styled Google Docs, and updates the spreadsheet with the document link and execution date.

---

🔄 Pipeline Workflow

Here is how the agent processes your content pipeline:

---

🛠️ Features

• Google Workspace Integration: Seamlessly fetches inputs from Google Sheets and writes content directly to Google Docs.
• Hybrid Content Evaluation: Auto-rates every post on word counts, structure, and qualitative dimensions (SEO, coherence, actionability) using an LLM-as-a-Judge.
• Background Watch Mode: Can run continuously (--watch mode) to poll the Google Sheet and automate new entries instantly.

---

📁 Repository Structure

AI_Agent/
├── main.py                  # CLI entry point (single run or --watch)
├── requirements.txt         # Package dependencies
├── README.md                # Documentation manual
├── agent_flow_diagram.png   # Pipeline flowchart
│
├── agent/                   # LangGraph orchestration
│   ├── state.py             # AgentState schema
│   ├── nodes.py             # Pipeline execution nodes
│   ├── graph.py             # Workflow graph construction
│   └── evaluator.py         # Heuristics & LLM-as-Judge
│
└── utils/                   # Workspace & LLM helpers
    ├── llm_factory.py       # Groq/Ollama wrapper
    ├── gdocs_handler.py     # Doc creation & Markdown parsing
    └── gsheets_handler.py   # Sheet readers & writers

---

🚀 Setup & Installation

1. Install Dependencies

Clone the repository, navigate to the folder, and install the required Python packages:

pip install -r requirements.txt

2. Google Cloud Console Configuration

The agent requires access to Docs, Drive, and Sheets APIs.

1. Go to the Google Cloud Console.
2. Create a project and enable these three APIs:
   • Google Sheets API
   • Google Docs API
   • Google Drive API
3. Configure the OAuth Consent Screen:
   • Set the User Type to External.
   • Under Test Users, add your Gmail address (required while the app is in testing mode).
4. Create Credentials:
   • Go to Credentials -> Create Credentials -> OAuth client ID.
   • Select Desktop app as the Application type.
   • Download the generated JSON credentials file.
   • Rename this file to google_creds.json and place it in your project's root folder.

3. Google Sheet Setup

1. Create a new Google Sheet with the following headers in Row 1:

Category	Topic	Updated Date	Link

2. Add a few blog ideas in the Category and Topic columns. Leave the Updated Date and Link columns blank.
3. Extract the Spreadsheet ID from the sheet's URL: https://docs.google.com/spreadsheets/d/ YOUR_SPREADSHEET_ID /edit

4. Configure Environment Variables

Create a file named .env in the root directory:

LLM_PROVIDER=groq
GROQ_API_KEY=your_groq_api_key_here
GROQ_MODEL=llama-3.1-8b-instant

# Google Sheets Configuration
GOOGLE_SHEET_ID=your_spreadsheet_id_here
GOOGLE_SHEET_RANGE=Sheet1!A:D

(Optional: Set LLM_PROVIDER=ollama to run models locally using Ollama).

---

💻 Running the Agent

Single Execution

Processes all currently pending rows in the sheet, updates the links/dates, and exits:

python main.py

Live Watch Mode

Listens for new entries in the sheet, checking for new rows every 60 seconds:

python main.py --watch

Press Ctrl+C to terminate the loop.

---

📊 Evaluation Mechanics

The evaluator (agent/evaluator.py) ensures generated content meets quality standards:

1. Heuristics (Fast): Checks if word count is > 600, verifies ≥ 3 heading levels (##), and matches target topic keywords in the intro.
2. LLM-as-a-Judge: Evaluates readability, coherence, and actionability on a 1-5 scale, returning structured scores.

---