This is a submission for the Agent.ai Challenge: Full-Stack Agent (See Details)
What I Built
I built a "Article Visual Carousel Pro Agent". This agent is designed to take raw webpage data (specifically, the output of a web crawler) and intelligently transform it into engaging, multi-slide LinkedIn carousels. I built this agent because I saw a need to streamline the process of creating social media content from existing web content. It automates the extraction of key information, structures it into a visually appealing format, and adds a call to action, saving marketers valuable time and effort. I envision it being used by marketing teams, content creators, and social media managers to quickly generate engaging content for social media platforms, specifically LinkedIn. This can help them repurpose existing content, reach new audiences, and drive engagement.
Here's how the agent works:
-
User Input: The agent starts by collecting user input:
- (1/5) Blog Post URL: The URL of the webpage to be converted into a carousel.
- (2/5) Format: The desired output format (defaults to 'Carousel - PDF - 1080 x 1080').
- (3/5) Theme: The visual theme of the carousel (defaults to 'Light 1').
- (4/5) Brand Name / Handle: The brand's name or social media handle.
- (5/5) Email Address: An email address to contact for more information.
Webpage Scraping: The agent then scrapes the content of the provided
page_url
and saves it tocrawled_page
.AI-Powered Content Structuring: Using the
crawled_page
, the agent invokes a Gemini AI model (gemini-2.0-flash-exp
) to analyze the scraped content. This AI acts as a marketing expert, extracting key information and structuring it into a JSON format (llm_json_output
) suitable for carousel generation.-
Data Validation, Mapping, and Typevis API Integration (Lambda Function): The
llm_json_output
is then passed to an AWS Lambda function (step 8, Language: node). This is where the core logic for preparing the data for the Typevis API is handled. Specifically, the Lambda function:- Validates the AI Output: Checks the
llm_json_output
for data integrity and required fields. - Maps to Typevis API: Transforms the
llm_json_output
into the specific JSON structure required by the Typevis API, ensuring compatibility. - Sends POST Request: Makes a POST request to the Typevis API using the mapped data to generate the carousel. This step also handles any necessary API authentication.
- Validates the AI Output: Checks the
Conditional Output (HTML): The agent uses conditional logic (If/Else) to determine the output styling. Based on the conditions, it outputs the data in HTML format with specific background colors.
Demo
You can interact with my Article Visual Carousel Pro Agent here: https://agent.ai/profile/10br0wobj8t9ns14
Youtube Video:
Agent.ai Experience
My experience with the Agent.ai Builder was mostly positive. The Builder's interface was intuitive and allowed me to quickly prototype and iterate on my agent's logic. The ability to define custom functions and integrate with external APIs was a significant highlight, enabling me to implement the complex logic required for content extraction and structuring.
One of the most delightful moments was seeing the agent successfully parse complex webpage data and generate a structured JSON output, ready to be used with the Typevis API, which is part of an app I'm currently developing. It was rewarding to see the agent transform raw text into a well-formatted, engaging carousel structure. Typevis is still in its alpha stage, and this agent is designed to integrate with it to fully automate the carousel creation process. This integration is a key part of my vision for Typevis.
However, there were also some challenging moments. Debugging complex logic and ensuring the agent handled various edge cases (e.g., missing data, overly long text) required careful planning and iterative refinement. It would be very helpful to have more robust debugging tools and real-time feedback within the builder.
Despite the challenges, the Agent.ai platform is a powerful tool for building intelligent agents, and I'm excited to see its future development and potential applications.
Conclusion
I believe this Content Carousel Creator Agent has the potential to significantly streamline content creation workflows. I encourage you to explore its capabilities and share your thoughts in the comments below. Let me know how you think this agent could be improved or used in new and innovative ways!
Top comments (0)