CrimeBoard

Upload evidence, and AI detective agents automatically build a floating laser-linked investigation board with suspects, guilt probabilities, and a courtroom-ready case file.

Comment

Agents of Deduction: An Interactive Murder Board Powered by Multi-Agent AI

Inspiration

Every great mystery has a murder board. From Agatha Christie to Sherlock and CSI, investigators don’t solve cases by reading documents in isolation. They solve them by seeing connections between evidence, people, and timelines.

Agents of Deduction was inspired by the idea of turning that classic murder board into a living, AI-powered system that builds itself and helps you reason about a case from multiple perspectives at once.

What It Does

Agents of Deduction lets you create a case, upload evidence like police reports, images, and witness statements, and automatically generates an interactive investigation board.

Behind the scenes, multiple AI agents analyze the case to:

  • Extract facts and timelines from unstructured data.
  • Identify suspects, motives, and relationships.
  • Rank suspects with confidence percentages.
  • Detect inconsistencies and missing links in the narrative.

The result is a visual evidence board plus a structured investigation report explaining what happened and what to look at next.

How We Built It (Powered by DigitalOcean)

We leveraged DigitalOcean’s integrated cloud services to build a scalable, multi-component application:

Component Technology Role
Frontend Next.js Interactive floating evidence board UI
Backend Serverless APIs (DigitalOcean App Platform) Orchestration and data handling
Database DigitalOcean Managed PostgreSQL Structured data storage for cases, nodes, and relationships
File Storage DigitalOcean Spaces Secure evidence storage using signed uploads
AI Engine DigitalOcean Gradient AI Agents Multi-agent analysis and structured output generation

We use Gradient AI as real backend infrastructure, orchestrating multiple investigative roles through structured JSON outputs.

Challenges We Ran Into

  • Consistency: Coordinating multiple AI agents without losing narrative consistency.
  • Schema Enforcement: Enforcing strict, reliable JSON outputs from the AI for production use.
  • Infrastructure: Handling secure file uploads and managing SSL differences between local and production environments.
  • Visualization: Designing a board layout that remains readable and functional as the volume of evidence grows.

Accomplishments We’re Proud Of

  • A working multi-agent AI system deployed end-to-end on the DigitalOcean platform.
  • Automatic generation of evidence boards from completely unstructured input data.
  • Secure, scalable uploads using Spaces and signed URLs.
  • Successfully treating AI as an integral part of the system’s logic, not just a UI feature.

What We Learned

  • Agent Design: Multi-agent AI works best with clear, narrow roles and strong output schemas.
  • User Experience: Visual context (the board) dramatically improves the user's ability to reason and find connections.
  • Stack Integration: The combination of serverless APIs, managed databases, and dedicated storage significantly speeds up complex system builds.

What’s Next for Agents of Deduction

  • Smarter default board layouts and chronological timelines.
  • Suspect face composites and visual profiling tools.
  • Cross-case pattern detection and knowledge base integration.
  • Collaboration tools for teams and researchers working together.

Built With

Share this project:

Updates