Stop guessing. Start screening.

AI agent that queries live ClickHouse box office, sentiment & streaming data to tell studios: EXTEND, HOLD or PULL — in plain English. Ask the question. Get the answer. In 30 seconds.

Comment

Note for judges: The backend runs on Google Cloud Run and may take 20-30 seconds to warm up on first request. The frontend shows API Status: Online (top right) when the backend is ready. Please wait for the green indicator before testing the AI Agent Chat.

Inspiration

Every Friday, film studio executives ask the same question: should we keep this film in theaters, or pull it? The decision involves millions of dollars — yet most studios still make it based on gut instinct and last week's spreadsheet.

We wanted to change that. What if a studio exec could just ask the question and get a data-driven answer in 30 seconds?

What It Does

CineMetrics is an AI agent that connects directly to a ClickHouse Cloud data warehouse containing real-time box office receipts, hourly audience sentiment scores, and streaming performance data.

You ask a plain English question:

"Should we pull Iron Colossus 4 from theaters?"

The agent autonomously generates and executes SQL queries against ClickHouse, analyzes week-on-week revenue trends, sentiment trajectory, and streaming velocity, then returns a structured EXTEND / HOLD / PULL recommendation with confidence score and full reasoning.

The confidence score is calculated as:

\( \text{Confidence} = \frac{\text{Signal Strength} \times \text{Data Coverage}}{\text{Uncertainty Factor}} \)

No dashboards to build. No analysts to brief. Just ask and get the answer.

How We Built It

Backend — Java Spring Boot on Google Cloud Run

  • Gemini 2.5 Flash via Vertex AI with a tool-use agent loop
  • ClickHouse Cloud (MergeTree engine) with four time-series tables:

Built With

Share this project:

Updates

Submission history