Compact transformer encoders.
Shipped as a single binary.
Package transformer-based encoders into a single, self-contained executable. No Python, no dependencies, no runtime overhead.
Use encoderfile when you need reliable, fast, dependency-free inference for encoder-based tasks, especially in constrained or compliance-sensitive environments.
Embeddings and classifications out-of-the-box
Supports embedding/feature extraction, sequence classification, and token classification with ONNX-exported models.
Small footprint
A binary often measured in tens-to-hundreds of megabytes instead of gigabytes of packages or runtime environments.
Deterministic, offline, and safe
Ideal for environments where stability, reproducibility, and compliance matter.
Integration-ready
encoderfile integrates smoothly into existing systems as a CLI, a gRPC/HTTP microservice, or even an MCP server without refactoring your stack.
Get Started with encoderfile
Prepare your model, create a configuration file, and build your encoderfile.