cinemaos

Inspiration

NASA mission control, Actual film production artifacts, Screenplay typography

What it does

Script Breakdown agent, Risk Scout agent, Mission Control, A hard approval boundary

How we built it

Stack: Next.js 14, TypeScript, Tailwind CSS, Google Gemini API, Prisma + PostgreSQL, React Flow, Framer Motion, Zod, Zustand.

Challenges we ran into

The problem it solves:- Before a single scene is shot, a script supervisor manually breaks a screenplay into scenes (sluglines, cast, locations, page counts), and a line producer separately combs the schedule for anything that could blow the budget or put crew at risk — night shoots stacked back to back, repeated locations forcing costly company moves, scenes missing key data. Both passes are slow, manual, and happen too late to cheaply change a schedule.

Accomplishments that we're proud of

We shipped a real approval boundary, The AI output is trustworthy by construction, not by luck, We built a genuine data model, The design has a point of view, It actually runs

What we learned

Enforce structured AI output with schemas, not hope; design the approval boundary early; ground "cinematic" UI in real film craft, not generic dark mode; always verify the build actually runs.

What's next for cinemaos

Next: finish Schedule Optimizer, Budget Sentinel, Continuity Watch, and Call Sheet Drafter agents on the existing pattern. Move agent execution to a background queue for long screenplays. Add real auth and role-based approvals per ProjectMember. Support PDF/FDX screenplay uploads via Vercel Blob, not just pasted text. Wire up ElevenLabs spoken briefings and Slack risk alerts. Add real-time agent status via SSE instead of polling. Eventually: multi-project support and a true calendar view for missions.

Built With

• framer-motion
• google-gemini-api
• next.js-14
• prisma-+-postgresql
• react-flow
• tailwind-css
• typescript
• zod
• zustand