soruly/trace.moe

What it solves

It allows users to identify the source of an anime screenshot by tracing it back to the specific anime, episode, and exact timestamp where the scene appears.

How it works

The system uses a dedicated library to analyze image features and generate image vectors. These vectors are used for image search against a database of anime scenes. The project is composed of several integrated parts, including an API server for search and database updates, a web frontend, a Telegram bot, and an MCP server for AI agents.

Who it’s for

Anime fans and developers who want to identify anime scenes from images or integrate this search functionality into their own apps and bots.

Highlights

• Comprehensive Ecosystem: Includes a web interface, API, Telegram bot, and browser extensions.
• AI Agent Integration: Provides an MCP server specifically for AI agents to utilize the search capability.
• Self-Hostable: Can be deployed via Docker Compose with support for custom video indexing.
• Pre-hashed Data: Offers a database dump for those who want to set up the system using existing pre-processed image vectors.