RightsLedger Agent

• 

  Executive Overview — Live runtime dashboard showing production details and cloud agent status.

• 

  Intake Workspace — Gemini 2.5 Flash structured extraction for legal documents and contracts.

• 

  Project Settings — Active cloud runtime verification showing Gemini API and ClickHouse status.

• 

  Agent Trace Workspace — Full execution provenance, step timings, and auditability trace.

• 

  Release Simulator — In-memory what-if scenario testing for territories and streaming platforms.

Inspiration

What it does

How we built it

Challenges we ran into### Inspiration

In global entertainment distribution, verifying content rights across complex territories, windows, and platforms is a massive manual bottleneck. Traditional workflows lead to licensing oversights, missed release dates, and legal exposure. We created RightsLedger Agent to bring deterministic safety, transparent reasoning, and AI automation to enterprise media release clearance.

What it does

RightsLedger Agent acts as an autonomous decision-support platform for media operations:

• Structured Legal Extraction: Ingests legal documents, licenses, and cue sheets to extract precise rights clauses with verbatim citations using Gemini 2.5 Flash.
• Deterministic Audit Engine: Audits candidate film cuts against rights on file for specific release plans (e.g., India Theatrical vs. Global SVOD) without relying on speculative model guesses for policy severity.
• Release Simulator: Allows rights managers to run "what-if" scenarios across prospective platforms, languages, and territories in-memory without altering production clearance history.
• Human-in-the-Loop Clearance: Flags blocking findings into a managed queue. Hard safety rules dictate that blockers require explicit human authorization to waive or clear.
• Complete Observability: Features an Agent Trace workspace that logs execution paths, execution provenance, and timing for total compliance auditability.

How we built it

• Frontend: Next.js & React built with responsive dashboard interfaces.
• Backend: FastAPI service orchestrating Agent logic and API boundaries.
• AI & Language Models: Gemini 2.5 Flash (via Google Cloud Agent Framework) for high-speed structured extraction, citation linking, and summary narration.
• Analytics & Storage: ClickHouse for high-throughput execution trace logging and telemetry storage.

Challenges we ran into

Integrating LLM-driven intelligence with deterministic business logic required strict architectural boundaries. We needed to ensure that LLM outputs never directly override legal blockers or severity levels. To solve this, we isolated Gemini to extraction, citation, and narration tasks, leaving hard compliance evaluation to a deterministic rules engine.

Accomplishments that we're proud of

• Built a complete end-to-end agentic governance architecture with real-time ClickHouse and Gemini API integration at runtime.
• Achieved zero-speculation safety where all extraction steps retain verbatim provenance.

What we learned

Building enterprise-grade AI agents requires a clear separation between LLM-based reasoning and deterministic code paths, along with complete trace observability.

What's next for RightsLedger Agent

• Expanding multi-tenant team workflows with role-based sign-off chains.
• Direct integration with enterprise DAM (Digital Asset Management) systems.

Accomplishments that we're proud of

What we learned

What's next for RightsLedger Agent

Built With

• clickhouse
• fastapi
• gemini-2.5-flash
• google-cloud
• next.js
• python
• react
• render
• typescript
• vercel