TrendTracker - A Multi-Agent System for Analyzing and Generating New Ideas

Filipe Arruda — Sun, 26 Jan 2025 19:19:22 +0000

This is a submission for the Agent.ai Challenge: Assembly of Agents (See Details)

What I Built

TrendTracker is a two-agent system designed to identify trending ideas from challenge submissions and generate new concepts or improvements based on those trends.

ScraperAgent:
- Gathers data (titles, descriptions, reaction counts) from DEV.to challenge submissions.
- Stores the scraped results for further analysis.
TrendTracker Agent:
- Invokes the ScraperAgent to collect the latest data.
- Analyzes the scraped results, ranking submissions by popularity.
- Generates creative ideas or improvements inspired by the top-performing submissions.

I built this system to streamline the discovery of trending topics and to spark innovation for anyone looking to create an AI-driven solution. Imagine a scenario where you want to see what’s resonating with users, then quickly brainstorm new features, side projects, or product enhancements. TrendTracker makes that process seamless.

Demo

https://agent.ai/agent/trendtracker

Below is a quick walkthrough of how the system works:

Step 1: ScraperAgent scrapes the DEV.to challenge page for new submissions.
Step 2: TrendTracker calls the scraper’s output, runs an LLM prompt to analyze the data, and generates suggestions.
Step 3: You receive a ranked list of top ideas, along with improvement proposals or entirely new agent concepts.

Agent.ai Experience

Building with the Agent.ai Builder was straightforward and fun. The “Assembly of Agents” concept allowed me to chain multiple agents together easily. Here’s what stood out:

Delightful Moments:
- The drag-and-drop SmartFlow interface for chaining the scraper and idea-generator.
- Being able to store intermediate results and pass them along to the next agent.
- Quickly iterating on prompts without leaving the browser interface.
Challenges:
- Making sure the scraper returns clean, structured data so the TrendTracker agent can interpret it accurately.
- Refining the LLM prompt to focus on creative suggestions rather than just summarizing.

Overall, the combination of multi-agent orchestration and prompt-driven AI made it easy to build a powerful system in a short amount of time.

CheckMate - quick and easy-to-use fact-checking assistant that combats misinformation

Filipe Arruda — Sat, 25 Jan 2025 03:39:55 +0000

This is a submission for the Agent.ai Challenge: Full-Stack Agent (See Details)

What I Built

CheckMate is a fact-checking assistant built on Agent.ai. It helps users quickly verify suspicious headlines, viral claims, or rumors by querying reliable fact-check sources (e.g., Google Fact Check Tools). With a simple chatbot interface, CheckMate encourages critical thinking and combats misinformation. It’s envisioned as a handy tool for anyone who wants to quickly confirm or debunk circulating claims, especially in fast-paced social media environments.

Demo

You can test the CheckMate agent at:
https://agent.ai/agent/checkmate

Here’s how it works in action:

User inputs a suspicious claim or headline.
CheckMate queries the Google Fact Check API via a serverless function.
The Agent returns summarized fact-check results or indicates no matches.

Agent.ai Experience

Building CheckMate with Agent.ai was an exciting journey. The Flow Builder made it straightforward to create the conversation structure. I utilized the Call Serverless Function action to connect with AWS Lambda function (written in Python) that queries the Google Fact Check Tools API.

Delightful Points

The visual Flow Builder helped us quickly link user input to our serverless logic.
Testing the agent in real-time allowed for rapid iteration.

Challenges

Handling the prefix in Google’s JSON response ()]}'\n) required custom parsing.
Ensuring the Lambda environment had the right libraries or using only standard Python libraries was a bit tricky at first.

In the end, CheckMate can streamline fact-checking into a simple chatbot flow, demonstrating how AI Agents can help combat misinformation and empower users to confirm what they see online.

Forem: Filipe Arruda

TrendTracker - A Multi-Agent System for Analyzing and Generating New Ideas

What I Built

Demo

Agent.ai Experience

CheckMate - quick and easy-to-use fact-checking assistant that combats misinformation

What I Built

Demo

Agent.ai Experience

Delightful Points

Challenges