<div align="center">
<h1>E-gemma</h1>
</div>
<div align="center">
Rust ONNX inference library for Google's [EmbeddingGemma](`google/embeddinggemma-300m`) text embeddings. Produces 768-dim
L2-normalized sentence embeddings via [`ort`] and [`tokenizers`].
[<img alt="github" src="https://img.shields.io/badge/github-findit--ai/egemma-8da0cb?style=for-the-badge&logo=Github" height="22">][Github-url]
<img alt="LoC" src="https://img.shields.io/endpoint?url=https%3A%2F%2Fgist.githubusercontent.com%2Fal8n%2F327b2a8aef9003246e45c6e47fe63937%2Fraw%2Fegemma" height="22">
[<img alt="Build" src="https://img.shields.io/github/actions/workflow/status/findit-ai/egemma/ci.yml?logo=Github-Actions&style=for-the-badge" height="22">][CI-url]
[<img alt="codecov" src="https://img.shields.io/codecov/c/gh/findit-ai/egemma?style=for-the-badge&token=6R3QFWRWHL&logo=codecov" height="22">][codecov-url]
[<img alt="docs.rs" src="https://img.shields.io/badge/docs.rs-egemma-66c2a5?style=for-the-badge&labelColor=555555&logo=data:image/svg+xml;base64,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" height="20">][doc-url]
[<img alt="crates.io" src="https://img.shields.io/crates/v/egemma?style=for-the-badge&logo=data:image/svg+xml;base64,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" height="22">][crates-url]
[<img alt="crates.io" src="https://img.shields.io/crates/d/egemma?color=critical&logo=data:image/svg+xml;base64,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&style=for-the-badge" height="22">][crates-url]
<img alt="license" src="https://img.shields.io/badge/License-Apache%202.0/MIT-blue.svg?style=for-the-badge&fontColor=white&logoColor=f5c076&logo=data:image/svg+xml;base64,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" height="22">
</div>
## Install
```toml
[dependencies]
egemma = "0.1"
```
Download the canonical fp32 export from
[`onnx-community/embeddinggemma-300m-ONNX`][EmbeddingGemma]
(`model.onnx` plus its `model.onnx_data` sidecar, and `tokenizer.json`).
The model card flags fp16 as an unsupported activation dtype for this
graph; pass `model_fp16.onnx` only if you've validated it for your
workload.
## Cargo features
| Feature | Default | Effect |
|--------------|:-------:|-----------------------------------------------------------------------------------------------------------|
| `inference` | ✅ | Pulls `ort` + `tokenizers`; activates [`TextEncoder`]. Native targets only. |
| `serde` | | `Serialize` / `Deserialize` on `Options`, `BatchOptions`, `ThreadOptions`. |
| `cuda` | | NVIDIA GPUs (Linux/Windows). Requires CUDA toolkit + cuDNN at build and run time. |
| `tensorrt` | | NVIDIA, optimized inference. Falls back to CUDA, then CPU. Requires CUDA + TensorRT. |
| `directml` | | Windows GPUs (any vendor) via DirectX 12. |
| `rocm` | | AMD GPUs (Linux). Requires ROCm SDK. |
| `coreml` | | macOS / iOS via Core ML (Neural Engine + GPU + Metal Performance Shaders). |
The execution-provider features are off by default — none are needed
for CPU inference, and each requires the corresponding vendor SDK at
build time.
[`TextEncoder`]: https://docs.rs/egemma/latest/egemma/struct.TextEncoder.html
## Target / feature contract
The `inference` feature is **native-only**. It pulls `ort` (ONNX
Runtime FFI) and `tokenizers` (which transitively depends on C-only
libraries like `onig_sys`); neither builds on `wasm32-*` today.
Building wasm with default features fails deep in `getrandom` /
`onig_sys` before this crate's code is reached.
**Wasm consumers must opt out:**
```bash
cargo check --target wasm32-unknown-unknown --no-default-features
```
Without `inference`, the public surface is the `Embedding` type,
`Options` / `BatchOptions` / `ThreadOptions`, and the `Error` enum
— useful when inference itself happens elsewhere (a server, a
different runtime) and only the value types and similarity primitive
need to be present.
## API surface
The crate exposes:
- `TextEncoder` — owns one `ort::Session` and one
`tokenizers::Tokenizer`. `embed`, `embed_batch`, `warmup`.
`Send + !Sync` (mirrors `ort::Session`); for parallelism, instantiate
one encoder per thread, or share one behind a `Mutex`.
- `Embedding(Arc<[f32]>)` — 768-dim L2-normalized sentence embedding.
`try_cosine` returns `Result<f32, Error>` (no panic on dim mismatch).
- `Options` / `BatchOptions` / `ThreadOptions` — session, batch, and
threading configuration. `with_*` / `set_*` builders are `const fn`
where the underlying types permit.
- `Error` (`#[non_exhaustive]`, `thiserror`-derived).
`Embedding` deliberately does **not** implement `Serialize` /
`Deserialize` — see its docstring for the validated round-trip pattern
through the inner slice.
## SIMD
`Embedding::try_cosine` dispatches the 768-element f32 dot product
through a runtime-detected backend:
- **NEON** on aarch64 (baseline ISA feature, always available).
- **AVX2 + FMA** on x86_64 when both are detected.
- **Scalar** four-accumulator fallback elsewhere.
The unsafe per-arch kernels take `&[f32; 768]` rather than `&[f32]` —
the type-level length invariant is what makes the raw-pointer reads
sound, and a wrong-length slice can never reach the unsafe boundary.
The dispatcher short-circuits to scalar under `cfg!(miri)` so Miri
matrices exercise the same call sites without entering platform
intrinsics it can't model.
## License
Dual-licensed under MIT or Apache-2.0, at your option.
See [LICENSE-MIT](LICENSE-MIT) and [LICENSE-APACHE](LICENSE-APACHE).
[EmbeddingGemma]: https://huggingface.co/onnx-community/embeddinggemma-300m-ONNX
[`ort`]: https://crates.io/crates/ort
[`tokenizers`]: https://crates.io/crates/tokenizers
[Github-url]: https://github.com/Findit-AI/egemma
[CI-url]: https://github.com/Findit-AI/egemma/actions/workflows/ci.yml
[doc-url]: https://docs.rs/egemma
[crates-url]: https://crates.io/crates/egemma
[codecov-url]: https://app.codecov.io/gh/Findit-AI/egemma