Smart Social media AI

Inspiration

Managing and growing a social media presence is time-consuming and requires constant creativity. From researching trends to crafting engaging posts and visual ideas, it’s a repetitive process. We were inspired to build an AI-driven assistant that acts like a smart content strategist—capable of automating these tasks intelligently using the power of LLMs and Google's Agent Development Kit (ADK).

What it does

Smart Social Media AI is a multi-agent system that:

1. Identifies trending hashtags for any topic
2. Generates concise, engaging social media posts
3. Suggests creative visual concepts to go along with the post

All of this is done in real-time through a CLI-based interface powered by Gemini 2.5 Pro models, offering a complete content creation package with a single prompt.

How we built it

1. We used the Google Agent Development Kit (ADK) to define modular sub-agents.
2. Each agent is powered by Gemini 2.5 Pro via LlmAgent class. The system is written in Python, using a hierarchical agent structure:
3. trend_finder_agent
4. content_writer_agent
5. visual_concept_agent All orchestrated by a social_media_agent (parent). Execution happens in the terminal using the ADK's agent runner CLI.

Challenges We ran into

1. Understanding the agent configuration and running structure in ADK (agent.py vs agent.yaml).
2. Initial API integration hurdles and resolving GoogleLLM/GeminiLLM imports.
3. Creating realistic, non-repetitive outputs through prompt engineering for sub-agents. 4.Managing agent communication in a seamless flow under one parent agent.

Accomplishments that we're proud of

1. Successfully implemented a fully functioning AI assistant that can brainstorm, write, and plan content autonomously.
2. Created a clean and reusable architecture with modular agents.
3. Learned and implemented Google’s Agent Development Kit within a short timeframe.
4. Developed a real-world use case that can scale into a product.

What we learned

1. Hands-on experience with Google ADK and multi-agent coordination.
2. How to design task-specific agents that work together in hierarchy.
3. Working with LLMs for content generation and prompt tuning.
4. Importance of flow design and result structuring in agent-based systems.

What's next for Smart Social media AI

1. Integrate actual API-based trend data (e.g., from X/Twitter, Instagram).
2. Add a UI/UX layer (React/Next.js) for real-time content previews.
3. Schedule auto-posting and maintain a content calendar via Google Calendar or Trello API.
4. Support multilingual post creation and image generation with generative AI tools like DALL·E or Imagen.

Built With

• cli
• gemini-2.5-pro
• google-agent-development-kit-(adk)
• google-cloud
• python
• virtual-environment
• windows-11
• yaml