🖋️ LekhAI

Your Autonomous, AI-Powered Editorial Team

LekhAI (derived from the Sanskrit word Lekh meaning "writing") is an advanced multi-agent system that orchestrates a team of AI personas to create high-quality, technically accurate blog posts.

Unlike standard "text generators," LekhAI doesn't just guess. It researches the live web, plans a strategic outline, drafts with context, and polishes the final output—mimicking a real-world editorial newsroom.

🚀 Key Features

• 🤖 Multi-Agent Architecture: Four specialized agents (Researcher, Strategist, Writer, Editor) working in a coordinated sequence.
• 🌐 Live Web Access: The Research Agent uses DuckDuckGo Search to fetch real-time data, sources, and recent developments (no API key required).
• ⚡ Real-Time Streaming: Uses Server-Sent Events (SSE) to stream progress and content to the UI instantly.
• 🧠 Gemini Powered: Built on Google's Gemini 2.5 Flash (or 1.5 Flash) for high speed and low cost.
• 🎨 Modern UI: A clean, responsive interface built with Tailwind CSS and Glassmorphism design.
• 💾 Session History: Tracks and saves your past generations for easy retrieval.
• 📝 Markdown Support: Full rendering of headings, lists, and code blocks.

🛠️ System Architecture

LekhAI uses a Sequential Coordinator pattern. The main application acts as the orchestrator, passing the state between agents.

graph LR
    A[User Input] --> B(Coordinator)
    B --> C{Research Agent}
    C -- Tools: Web Search --> D[Live Internet]
    D --> C
    C -- Key Points --> E{Outline Agent}
    E -- Structure --> F{Writing Agent}
    F -- Draft --> G{Editing Agent}
    G -- Final Blog --> H[User UI]

Loading

The Agent Roster

• 🔍 Researcher: Scours the web for 5-10 key facts and sources.
• 📋 Strategist: Converts raw facts into a logical, SEO-friendly outline.
• ✍️ Writer: Expands the outline into a full draft with code snippets and examples.
• ✨ Editor: Fixes grammar, adjusts tone, and ensures professional formatting.

📦 Installation

Prerequisites

• Python 3.10 or higher
• Google Gemini API Key

1. Clone the Repository

git clone https://github.com/Vrajesh-Sharma/Blog-Agent.git
cd LekhAI

2. Create a Virtual Environment

python -m venv venv
# Windows
venv\Scripts\activate
# Mac/Linux
source venv/bin/activate

3. Install Dependencies

pip install -r requirements.txt

4. Configure Environment Create a .env file in the root directory:

# Code snippet
GOOGLE_API_KEY=your_gemini_api_key_here
GEMINI_MODEL=gemini-2.5-flash  # Recommended for speed

5. Run the Application

python app.py
Access the application at: http://127.0.0.1:5000

📂 Project Structure

LekhAI/
├── agents/                 # The brains of the operation
│   ├── agent_base.py       # Base class handling Gemini connections
│   ├── research_agent.py   # Web searching logic
│   ├── outline_agent.py    # Structuring logic
│   ├── writing_agent.py    # Drafting logic
│   └── editing_agent.py    # Polishing logic
├── tools/                  # Capabilities given to agents
│   ├── google_search_tool.py # DuckDuckGo integration
│   └── create_outline_tool.py
├── templates/              # Frontend UI
│   ├── base.html           # Layout and Tailwind setup
│   ├── index.html          # Landing page
│   ├── generate.html       # Streaming generation interface
│   └── history.html        # Past blogs view
├── services/               # State management
│   ├── session_service.py
│   └── memory_service.py
├── app.py                  # Main Flask application & Coordinator
├── requirements.txt        # Python dependencies
└── .env                    # API Keys (Hidden)

🔧 Configuration & Customization

Changing Models

You can switch between Gemini versions in your .env file.

• Speed/Free Tier: gemini-2.5-flash or gemini-1.5-flash

• High Quality: gemini-1.5-pro (Note: Lower rate limits may apply)

Modifying Agent Prompts

To change how an agent behaves (e.g., making the Writer more humorous), edit the system_instruction in the respective file under agents/.

Example (agents/writing_agent.py):

system_instruction=(
    "You are a witty technical writer. Use analogies and humor..."
)

❓ Troubleshooting

Error: 429 Resource Exhausted

• Cause: You are hitting the Gemini Free Tier rate limits (approx 15 requests/min).
• Fix: The system has built-in time.sleep(5) delays in app.py. If you still see this, increase the sleep time to 10 seconds.

Error: "Could not convert part.function_call"

• Cause: The model tried to use a tool but the code didn't handle the callback.
• Fix: Ensure agents/agent_base.py has enable_automatic_function_calling=True inside the chat session initiation.

Search Results are Empty

• Cause: DuckDuckGo rate limiting.
• Fix: The tool catches this error gracefully. Try again in a few minutes or ensure your IP isn't blocked.

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

1. Fork the project
2. Create your Feature Branch (git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add some AmazingFeature')
4. Push to the Branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

👨‍💻 Author

Vrajesh Sharma

• LinkedIn
• Portfolio
• GitHub

_{Built with ❤️}