Metadata-Version: 2.4
Name: kynasmith
Version: 0.1.0
Summary: Official Python SDK for Kynasmith.
Project-URL: Documentation, https://docs.kynasmith.dev/sdks/python
Project-URL: Homepage, https://docs.kynasmith.dev
License: Proprietary
Requires-Python: >=3.12
Requires-Dist: httpx>=0.28.1
Requires-Dist: kynasmith-contracts<0.2.0,>=0.1.0
Requires-Dist: mediapipe==0.10.32
Requires-Dist: opencv-python-headless>=4.13.0.92
Requires-Dist: websockets>=16.0
Description-Content-Type: text/markdown

## Kynasmith Python SDK

The official Python SDK for Kynasmith.

### Install

```bash
pip install kynasmith
```

### What it includes

- machine-credential bootstrap and access-token helpers
- detection session and realtime helpers
- bundled MediaPipe Pose Landmarker support
- local MP4 replay and local camera helpers

### Quick start

```python
from kynasmith import KynasmithClient
from kynasmith.models import DetectionSessionCreate

client = KynasmithClient(
    base_url="https://api.kynasmith.dev",
    project_id="proj_123",
    api_key_id="key_123",
    api_key_secret="secret_123",
)

session = client.detection.create_session(
    DetectionSessionCreate(
        project_id="proj_123",
        movespec_version_id="fsv_123",
        mode="replay",
    )
)
```

### Runtime requirements

- Python 3.12 or newer
- the default install includes `mediapipe` and `opencv-python-headless`
- the first detector run may download the default `.task` model into the local cache unless you provide `model_path`

### Notes

- machine-credential clients automatically remint short-lived access tokens when the cached token is near expiry
- raw video stays local; the SDK sends pose frames, not uploaded MP4 bytes
- see the public docs for auth, troubleshooting, and the end-to-end happy path: https://docs.kynasmith.dev/sdks/python
