Inspiration
There's a food truck back at my hometown called FlavorHive that recently gained a ton of traffic through social media marketing. Their videos spread online and word of their business spread in my community, helping their small business grow. I thought that other small businesses could benefit from this kind of growth, so we decided to make an application for small businesses to market on social media easier.
What it does
TrendHive analyzes the market trends for viral youtube videos associated with a sector. Key aspects for presentation such as audio, visual frames, and camera angles are extracted to then generate a video.
How we built it
TrendHive is powered by Google's ADK platform using python and React front-end. We use a sequential agent workflow, with a combination of LLM agents.
Challenges we ran into
The prompt engineering was perhaps the most difficult part of the process; we needed to make sure the prompts weren't too specific relevant information is fetched.
Accomplishments that we're proud of
I think passing state and integrating the agents together was our biggest accomplishment as we used custom logic and APIs within each module.
What we learned
The underlying logic and information passed between the different agents is what controls the whole process. Without proper prompt engineering and transfer of information, the agentic workflow holds no value.
What's next for TrendHive
TrendHive will continue to develop new features by adding new platforms to scrape information from. We are also planning to create a local sponsor finding agent.
Log in or sign up for Devpost to join the conversation.