Cinefiles

Inspiration

Independent filmmakers pour their passion into visual storytelling, only to hit a massive financial and legal wall in post-production: music licensing. Securing commercial tracks requires navigating complex sync and master rights, obtaining clearance documentation for distribution, and avoiding devastating E&O insurance rejections. Traditional legal clearance workflows are slow, opaque, and cost-prohibitive for indie budgets. We built Cinefiles to solve this bottleneck, bringing automated, instant audio perception and legal risk assessment directly to the edit bay.

What it does

Cinefiles is an enterprise media perception agent that automates music copyright clearance. Beyond simply analyzing an audio URL in the web interface, the upgraded platform features a comprehensive script-parsing engine that cross-references script dialogue and scene descriptions with timeline cues. The agent isolates acoustic fingerprints, queries real-time licensing metadata, and instantly returns a structured financial risk breakdown—including detailed Synchronization and Master Use fees—alongside required compliance guidance for distribution and E&O insurance.

How I built it

Cinefiles leverages a powerful, production-grade cloud stack combining Google Cloud and IBM infrastructure:

• Backend & UI Microservice: Written in Python using FastAPI (with integrated frontend UI routing to streamline navigation) and generated via IBM Bob, hosted serverlessly on Google Cloud Run (us-east1) with secure environment variable management via Google Secret Manager.
• Orchestration & Agent Logic: Built natively using Google Cloud Agent Builder and Conversational Agents, utilizing structured Playbooks, Few-Shot training examples, and precise tool-binding syntax (${TOOL: IBM_Bob_Audio_Clearance_Tool}) integrated with Gemini Enterprise to guarantee deterministic tool execution.
• Security & IAM: Configured with robust service account permissions bridging Dialogflow CX and Cloud Run via the Cloud Run Invoker role.

Challenges I ran into

During development, we tackled complex multi-system integration hurdles—ranging from resolving initial Cloud Run 503 IAM authentication boundaries between Dialogflow CX and our backend service, to aligning regional deployment to us-east1 and refining Playbook instructions and tool mappings to eliminate agent refusal behaviors. By structuring clean OpenAPI schemas, implementing rigorous few-shot examples, and aligning the FastAPI UI routing with our backend logic, we achieved a seamless, highly reliable end-to-end execution loop.

Accomplishments that I'm proud of

• Successfully delivered a fully functional, production-ready AI legal agent that bridges generative audio workflows and script parsing with real-world financial risk calculations.
  • Orchestrated a seamless multi-cloud and multi-tool architecture (Gemini Enterprise, Cloud Run, IBM Bob backend generation, and AudD fingerprinting) operating in real-time under tight tool latencies.

What I learned

• We gained deep technical insight into advanced prompt engineering for agentic playbooks, managing precise tool-parameter bindings in Agent Builder, configuring FastAPI-driven frontend-backend routing, and securing serverless microservices with proper IAM role propagation.

What's next for Cinefiles

• Expanding the backend mock engine to dynamically scale licensing fees based on media duration, territory distribution scope, and track popularity metrics.
• Deepening native video timeline parsing to automatically timestamp multiple background cues within a full feature-length rough cut alongside automated script cue matching.

Built With

• gemini
• geminiai
• google
• google-cloud
• ibm
• python