Thanks to visit codestin.com
Credit goes to devpost.com

Inspiration

What it does

Inspiration- Creating viral social media content is one of the biggest challenges marketers face today. I was inspired by the endless hours teams spend researching trends, brainstorming content, and manually coordinating campaigns across platforms. The idea struck me: What if AI agents could handle each part of this process automatically? I envisioned a system where multiple specialized AI agents work together like a marketing team - one analyzing trends, another writing content, a third selecting visuals, and a fourth optimizing performance. This would transform campaign creation from days of manual work into minutes of AI-powered automation. What I learned- This project taught me the power of multi-agent AI systems and how different AI models can collaborate seamlessly. I discovered how to orchestrate multiple APIs (Google Gemini, Unsplash, Google Trends) to work together through a FastAPI backend. The biggest learning was understanding how to structure AI agents with specific roles and getting them to produce consistent, high-quality output. I also learned about the importance of starting with backend functionality first - the agents are the core value proposition, so getting them working was critical before building the frontend. How I built it.- I built VyralFlow AI as a monorepo with backend-first development: Backend: Used Claude Code to generate a complete FastAPI system with 4 specialized AI agents:

Trend Analyzer - scrapes trending topics and hashtags Content Writer - generates engaging posts using Google Gemini Visual Designer - selects relevant images from Unsplash Performance Analytics - tracks viral probability and engagement

Frontend: React + Vite with TypeScript, shadcn/ui components, and Acernity animations for a modern, responsive interface. Integration: Connected Google Gemini for content generation, Unsplash for high-quality visuals, and Google Trends for real-time trend analysis. Architecture: Designed as a scalable multi-agent system where each agent runs independently and stores results in Firestore. Challenges I faced... The biggest challenge was orchestrating multiple AI agents to work together seamlessly. Getting the agents to communicate and produce consistent output required careful prompt engineering and error handling. API integration was another hurdle - ensuring the Unsplash API returned diverse, relevant images instead of duplicates, and making sure Google Gemini generated genuinely engaging content rather than template responses. Real-time processing was tricky - building a system that could handle campaign creation requests, process them through multiple agents, and return results efficiently while maintaining data consistency in Firestore. The final challenge was keeping everything cost-effective for a hackathon - I solved this by using all free tiers (Google Gemini, Firestore, Netlify, Google Cloud Run) to achieve a $0 total cost while maintaining production-quality functionality.

Built With

Share this project:

Updates