Top10 Anything

AI Agent Powered Ranking Insights with Tableau

Comment

Top10 Anything - Project Story

Inspiration

The inspiration for Top10 Anything came from two key insights during our exploration of AI and BI integration:

  1. TopX Information Seeking Pattern: I noticed how frequently people search for "Top X" rankings across various domains - from technology choices to consumer products. This universal pattern of seeking curated, ranked information revealed an opportunity to create something interesting using AI capabilities.

  2. AI Agents Meet BI Tools: While experimenting with AI agent frameworks, I was inspired by their ability to produce structured outputs. This led to an exciting realization: we could bridge the gap between Large Language Models' analytical capabilities and Tableau's visualization power. By combining these technologies, we could transform unstructured web data into meaningful, interactive visualizations automatically.

What it does

Top10 Anything is an AI-powered system that:

  1. Generates Rankings: Users can input any topic (e.g., "Best Electric Cars 2024", "Top Programming Languages") and get AI-analyzed rankings.

  2. Provides Deep Analysis: For each ranked item, AI agents(with LLM and web search) provide comprehensive analysis:

    • Calculates comprehensive scores
    • Identifies key advantages
    • Extracts quantitative metrics
    • References authoritative sources

  3. Automates Visualization: The system:

    • Converts AI analysis into Tableau-ready data
    • Automatically updates Tableau dashboards
    • Provides interactive visualizations

How we built it

The development process involved several key components:

  1. AI Agent Architecture:

    • Built using Pydantic-AI framework
    • Implemented DuckDuckGo search integration
    • Developed ranking using LLM and web search
    • Created data validation models

  2. Tableau Integration:

    • Used Tableau Hyper API for data source generation
    • Implemented REST API and TSC for cloud publishing
    • Created embedded dashboard interface
    • Built update mechanism

  3. Web Interface:

    • Developed FastAPI backend
    • Created simple frontend web page
    • Integrated Tableau embedding
    • Implemented progress tracking

Challenges we ran into

  1. System Design:

    • Designing a system that can handle virtually any ranking topic
    • design the workflow of the system
    • Balancing the scope of the system with the complexity of the topic
    • Ensuring the system can handle the data from the web search

  2. Data Consistency:

    • Ensuring consistent ranking criteria across different topics
    • Handling varying data availability
    • Maintaining data quality standards

  3. Technical Integration:

    • Coordinating between AI analysis and Tableau updates
    • Managing asynchronous operations
    • Ensuring reliable data pipeline

Accomplishments that we're proud of

  1. Seamless Integration: Successfully bridged the gap between AI analysis and Tableau visualization

  2. Flexibility: Created a system that can handle virtually (almost) any ranking topic

  3. Automation: Built an end-to-end pipeline requiring minimal user intervention

  4. Finishing it in a short period of time

What we learned

  • Best practices for AI agent development
  • Tableau API integration patterns
  • better work with AI powered IDEs (cursor)

What's next for Top10 Anything

  1. Enhanced AI Capabilities:

    • Multiple specialized agents (reasoning, search, scoring)
    • Advanced reasoning with newer models
    • Improved source validation

  2. Tableau Integration:

    • Multi-table data models
    • Enhanced visualization templates

  3. User Features:

    • Human-in-the-loop validation

Built With

Share this project:

Updates