Encoderfile Models

Pre-built encoderfiles for popular Hugging Face embedding models — self-contained executables that run as embedding servers with no Python or ML dependencies required.

Available Models

Model	Details
sentence-transformers/all-MiniLM-L6-v2	384-dim, English sentence embeddings
deepset/deberta-v3-base-injection	DeBERTa-variant prompt injection classification model
JasperLS/deberta-v3-base-injection	DeBERTa-variant prompt injection classification model
JasperLS/gelectra-base-injection	gELECTRA-variant prompt injection classification model
protectai/deberta-v3-base-prompt-injection-v2	DeBERTa-variant prompt injection classification model
protectai/deberta-v3-small-prompt-injection-v2	DeBERTa-variant prompt injection classification model
protectai/deberta-v3-base-prompt-injection	DeBERTa-variant prompt injection classification model
protectai/distilroberta-base-rejection-v1	DistilRoBERTa-variant prompt injection classification model

More models coming soon.

Usage

Each model directory contains platform-specific binaries. Download the one for your platform, make it executable, and run:

# Serve embeddings over HTTP
./all-MiniLM-L6-v2.aarch64-apple-darwin.encoderfile serve

# Or infer directly from the CLI
./all-MiniLM-L6-v2.aarch64-apple-darwin.encoderfile infer "this is a test"

If you don't see the model you want or are using an exotic architecture, check out our guide on Building encoderfiles.

About

These encoderfiles are built and published by mozilla-ai using the Encoderfile tool. To build your own, see the Encoderfile documentation.