SigLIP 2 text embedding on CPU with Rust + ONNX

We’re building a robotics data platform with a lot of images, video, and text metadata. For search, we use SigLIP 2. GPUs handle batched asynchronous image/video embedding and indexing, while this small Rust + ONNX Runtime service handles live text queries on CPU. Both land in the same embedding spa

中文全文翻译正在补齐，可先查看下方英文原文。

We’re building a robotics data platform with a lot of images, video, and text metadata. For search, we use SigLIP 2. GPUs handle batched asynchronous image/video embedding and indexing, while this small Rust + ONNX Runtime service handles live text queries on CPU. Both land in the same embedding spa