timber-compiler

What problem does it solve?

This Skill converts ML models from various frameworks (XGBoost, LightGBM, scikit-learn, CatBoost, ONNX) into self-contained C99 inference binaries, enabling fast and efficient model deployment.

Core Features & Use Cases

• Model Compilation: Converts ML models into C99 binaries for zero-overhead inference.
• HTTP Server: Serves compiled models through a simple Ollama-compatible API.
• Fast Inference: Achieves single-sample latencies of ~2 µs, outperforming Python inference by over 336×.
• Zero Runtime Dependencies: Binaries are self-contained with no external dependencies.
• Use Case: Ideal for fraud detection systems, edge devices, and IoT applications where fast, reliable inference is critical.

Quick Start

Install timber-compiler with pip install timber-compiler. Use timber serve https://your-model-url to start serving the model. Make an inference request to http://localhost:11434/api/predict with the desired input data.