CineScope AI

Inspiration

Film crews lose time and money when permit, weather, and location risks surface too late. CineScope AI turns a screenplay excerpt and production context into an auditable pre-production risk board.

What it does

CineScope AI creates a greenlight score, structured risks, an audit trail, and a Recheck loop that can reopen or resolve risks as evidence changes.

How we built it

The web app is Python with a lightweight browser UI. The AI runtime path is implemented with Google Agent Development Kit / Gemini and Parallel Search for evidence retrieval. AWS is used only as non-AI web hosting, which is permitted by the contest rules. The repository contains the runtime integration code, tests, architecture notes, and setup instructions.

Verified demo state

The public demo video shows the deterministic MOCK mode running on the hosted app: an audit returns a 90/100 greenlight score, creates an audit trail, renders risks, and demonstrates the Recheck loop changing a risk from OPEN to RESOLVED. MOCK mode makes zero third-party network calls. Live Google/Parallel execution requires the corresponding API credentials and the app fails closed if they are absent.

Challenges

The main challenge was preserving an honest no-spend path while keeping the same agent workflow and data contract for live Google/Parallel execution.

What we learned

Agentic production tooling is more useful when every risk has a state, evidence trail, and repeatable recheck action instead of one opaque recommendation.