The latest articles on Forem by Mohit (@hack_rnd). https://forem.com/hack_rnd https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F2742063%2Febdbe283-4ec9-49d2-80eb-eebe16da0d4a.png https://forem.com/hack_rnd en Mohit Tue, 21 Jan 2025 13:21:46 +0000 https://forem.com/hack_rnd/movie-genie-1ekh https://forem.com/hack_rnd/movie-genie-1ekh <p><em>This is a submission for the <a href="https://srv.buysellads.com/ads/long/x/T6EK3TDFTTTTTT6WWB6C5TTTTTTGBRAPKATTTTTTWTFVT7YTTTTTTKPPKJFH4LJNPYYNNSZL2QLCE2DPPQVCEI45GHBT" rel="noopener noreferrer">Agent.ai</a> Challenge: Full-Stack Agent (<a href="https://dev.to/challenges/agentai">See Details</a>)</em></p> <h2> What I Built </h2> <p>Movie Genie is an AI agent that recommends popular and top rated movies based on user preferences! </p> <p>How it works: </p> <ol> <li>Ask user their preferences for movie recommendation </li> <li>Create structured output based on the user query</li> <li>Search movies via TMDB API using parameters extracted from the user query</li> <li>Generate HTML card layout using the API response </li> <li>Return HTML response to the user </li> </ol> <h2> Demo </h2> <p><a href="https://agent.ai/agent/movie-genie" rel="noopener noreferrer">https://agent.ai/agent/movie-genie</a></p> <h2> Agent.ai Experience </h2> <p>Agent.ai builder is a powerful platform to create AI agents quickly and share with the world. By allowing to assemble multiple agents, agent.ai is truly creating a professional network for agents. </p> <p>Below are some observations while using the builder: </p> <ul> <li>It provides a comprehensive set of "Actions" to implement the workflow of the agent. However, the actions can only be executed sequentially. Allowing parallel actions can be a huge improvement. </li> <li>"Get user input" action allows only one input element at a time (for eg. textbox, text-area, dropdown etc.). Multiple elements can make it better (for eg. form input). </li> <li>Latency is high for LLM calls. I initially used GPT-4o for generating the final HTML response and it was taking a lot of time. Replacing it with llama-3.3-70b made it a bit better. GPT-4o was however great for generating structured output. </li> </ul> devchallenge agentaichallenge ai machinelearning