Nexus - Agentic Command Center

Transform Tableau from a read-only dashboard into an Agentic Command Center. Analyze root causes, simulate outcomes, and execute business actions—all in one click.

Comment

Inspiration

We noticed a fundamental flaw in modern business analytics: The "Insight-to-Action" Gap. Supply chain managers spend hours staring at Tableau dashboards to identify problems (e.g., "Stock is low"), but to fix them, they have to leave the dashboard, open ERP software, or draft emails. This context switching kills productivity. We asked: "What if the dashboard didn't just show us the problem, but also helped us fix it?"

What it does

Nexus is a next-generation wrapper for Tableau that adds an "Agentic Layer" on top of your data. It operates in three stages:

  1. Proactive Intelligence: Click a data point, and our AI Agent (Gemini 1.5 Flash) analyzes the root cause instantly.
  2. Live Simulation: Users can adjust variables via the Nexus UI, and the Tableau dashboard updates in real-time to forecast the impact.
  3. Autonomous Execution: With a single click, the system triggers real-world actions (emails, DB updates) via secure API calls.

How we built it

  • Frontend: Built with React 18 (Vite) and Tailwind CSS. We designed a "Glassmorphism" interface to make data analysis feel immersive.
  • Integration: We utilized the Tableau Embedding API v3. We implemented deep bi-directional communication, using event listeners (markSelectionChanged) to read data from charts and programmatically updating Parameters to drive simulations.
  • The AI Agent: We integrated Google Gemini 1.5 Flash for context-aware reasoning and structured JSON outputs.
  • Backend: Used Node.js/Express and MongoDB to create an audited "Action Log" ensuring compliance.

Challenges we ran into

  • Bi-directional Latency: Ensuring that the communication between the Tableau Embedding API (frontend), the AI model (backend), and the database happened instantly was difficult. We optimized our API routes to ensure the "Agent" feels responsive (sub-3-second latency).
  • Context Management: Teaching the AI to understand the specific context of a selected Tableau mark required careful data structuring.

Accomplishments that we're proud of

  • Successfully implementing "Write-back" capabilities in a Tableau environment without using expensive external plugins.
  • Creating a Simulation Mode that allows the external UI to control Tableau parameters seamlessly.
  • Designing a UI that looks like a sci-fi movie interface but solves boring enterprise problems effectively.

Built With

+ 5 more
Share this project:

Updates