Social Media Design Agent

An intelligent multi-agent system built with Google ADK for creating and iterating on social media content. This agent demonstrates advanced AI orchestration patterns with two distinct operational modes: fast interactive design assistance and autonomous quality assurance through "Deep Think Mode."

Features

• Dual-Mode Operation: Choose between speed (Regular Mode) or quality (Deep Think Mode)
• Intelligent Image Generation: Powered by Gemini's image generation capabilities
• Reference Image Support: Upload images for style and composition guidance
• Versioned Asset Management: Automatic versioning and organization of created content
• Autonomous Quality Loop: Multi-agent review and refinement system
• Context Awareness: Maintains session state for iterative improvements

Architecture

The agent consists of several specialized sub-agents:

• Main Social Media Agent: Handles user interactions and routing
• Content Generation Agent: Creates and edits images with detailed instructions
• Content Review Agent: Evaluates quality across multiple dimensions
• Loop Control Agent: Makes intelligent decisions about iteration continuation
• Deep Think Loop: Orchestrates autonomous creative iteration

Prerequisites

• Python 3.13+
• Google ADK (pip install google-adk)
• Gemini API key with image generation access

Setup

1. Clone the repository:

   git clone <repository-url>
   cd social_media_agent

2. Install dependencies:

   pip install -r requirements.txt

3. Set up environment variables: Create a .env file in the project root:

   GEMINI_API_KEY=your_gemini_api_key_here

4. Verify installation:

   python -c "import google.adk; print('ADK installed successfully')"

Running the Agent

Development UI (Recommended for testing)

Start the ADK development server:

adk web

Then navigate to http://127.0.0.1:8000/dev-ui in your browser and interact with the agent through the web interface.

Usage Examples

Regular Mode (Fast)

Simple, direct interaction for quick iterations:

User: "Create a holiday promotion poster in 16:9 aspect ratio"
Agent: [Generates image quickly using generate_image tool]
User: "Make the text larger"
Agent: [Uses edit_image tool for quick refinement]

Deep Think Mode (Quality)

Autonomous quality assurance with multiple iterations:

User: "Deep think create a professional product launch announcement"
Agent: [Enters autonomous loop]
       1. Generates initial content
       2. Reviews for quality and adherence
       3. Refines based on feedback
       4. Repeats until professional standard
       5. Presents final result

With Reference Images

Upload inspiration images for style guidance:

User: [Uploads reference image] "Create a social media post in this style"
Agent: [Automatically saves reference as reference_image_v1.png]
       [Uses reference for style and composition guidance]

Key Commands

• Regular Generation: "Create a [description]"
• Deep Think Mode: "Deep think create a [description]"
• With Reference: Upload image + "Create something in this style"
• List Assets: "Show me all my assets"
• List References: "What reference images do I have?"
• Load Previous: "Show me [asset_name_v2.png]"

File Structure

social_media_agent/
├── agent.py                 # Main agent configuration
├── deep_think_loop.py       # Deep think mode implementation
├── tools/
│   └── post_creator_tool.py # Image generation and editing tools
├── requirements.txt         # Python dependencies
└── README.md               # This file

Configuration

The agent can be customized by modifying:

• Max iterations: Change max_iterations=5 in deep_think_loop.py
• Models: Update model names in agent configurations
• Instructions: Modify agent instructions for different behaviors
• Tools: Add or remove tools from the agent's toolkit

Troubleshooting

Common Issues

1. "GEMINI_API_KEY environment variable not set"

   • Ensure your .env file contains the API key
   • Verify the key has image generation permissions

2. "Artifact service is not initialized"

   • The agent uses InMemoryArtifactService by default
   • For production, consider using GcsArtifactService

3. "Deep think mode not activating"

   • Ensure you include "deep think" in your prompt
   • Check that sub_agents=[deep_think_loop] is configured

Debug Mode

Enable detailed logging:

export PYTHONPATH=.
python -c "
import logging
logging.basicConfig(level=logging.DEBUG)
from social_media_agent.agent import runner
# Your agent interaction code here
"

Advanced Usage

Custom Asset Names

Provide meaningful names for better organization:

# In your prompts, specify asset names
"Create a holiday_promotion poster"
# Results in: holiday_promotion_v1.png, holiday_promotion_v2.png, etc.

Reference Image Management

# List all reference images
"What reference images do I have?"

# Use specific reference
"Create a design using reference_image_v2.png as inspiration"

# Use latest uploaded reference
"Create something based on the latest reference image"

Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

Support

For issues and questions:

• Check the troubleshooting section above
• Review ADK documentation: https://google.github.io/adk-docs/
• Open an issue in the repository