tensorplate-python 0.1.3

First-party Python SDK for calling deployed TensorPlate detection and vision serving models.

sdk/python/

tensorplate-python — the first-party Python SDK for calling deployed TensorPlate detection and vision serving models over the v0.1 /infer HTTP envelope.

The import package is tensorplate:

import tensorplate

Install

pip install tensorplate-python            # core client
pip install "tensorplate-python[vision]"  # + numpy & Pillow for VisionClient.detect

The wheel + sdist are also attached to each signed GitHub Release for checksum-verified or air-gapped installs. See the SDK quickstart for that flow and the [numpy] / [vision] extras.

Ownership

• Layer: client / user space (no runtime or serving-worker code)
• Language: Python (3.10+)
• Distribution: tensorplate-python (PEP 621 pyproject.toml)
• Import package: tensorplate

Documentation

• Quickstart and API reference — ServingClient, VisionClient, Detection, tensors, and errors.
• Detection workflow — preprocessing, the yolo_v8_single_output contract, and postprocessing.
• Endpoint resolution — CLI-parity precedence and URL canonicalization.
• Examples — single-image and user-space camera samples.

Scope

The SDK is a client-side library: it calls already-deployed models and does not change runtime serving. In-runtime camera/video ingest, a DeepStream sink, and a streaming session API are out of scope and are not provided here.

License

Apache-2.0. See LICENSE.