AI Models / Armiger
Armiger
Classify torrent metadata on your own hardware.
Armiger uses release names and file lists to predict media type, language and other attributes. The standalone ONNX model runs on a CPU. Integrate it directly, or use Armiger Ultra with Bitmagnet or Rats Search.
Armiger or Armiger Ultra?
Armiger
Use the standalone INT8 ONNX classifier to label supplied torrent metadata in your own Python workflow. The model card documents the input and output.
Armiger Ultra
Use the extended pipeline to combine classification with optional title matching, language rules and swarm availability estimates. See the Armiger Ultra model card for model variants and setup.
Results are predictions from metadata. Armiger does not download or inspect media payloads. The evaluation and limitations describe what the published tests measure.
Run it with your torrent index
Use a Docker Compose guide for Bitmagnet or Rats Search. Each setup includes Armiger Ultra. To run the classifier separately, follow the standalone quickstart.
The project and model weights are published under Apache-2.0. Check the resource notes for disk and memory requirements.