Metadata-Version: 2.4
Name: machinevision-cloud
Version: 2.1.0
Summary: Python SDK for machinevision.cloud - vision inference API: natural-language tasks, structured results, visual-context accuracy tuning
License: MIT
Project-URL: Homepage, https://machinevision.cloud
Project-URL: Documentation, https://machinevision.cloud/docs/quickstart
Requires-Python: >=3.9
Description-Content-Type: text/markdown
Requires-Dist: httpx>=0.27

# machinevision-cloud

Python SDK for [machinevision.cloud](https://machinevision.cloud) — the vision
inference API: describe a visual task in natural language, POST an image, get
structured JSON back.

```python
from machinevision import Client

mvc = Client(api_key="mvc_test_...")  # sandbox key: free, deterministic mock results

task = mvc.create_task(
    name="damage_check",
    task_type="visual_inspection",
    instruction="Inspect the returned item for scratches, dents, or cracks.",
)

result = mvc.inspect(task["task_id"], image_url="https://example.com/photo.jpg")
print(result["result"]["verdict"], result["usage"])
```

- Every method returns the API's JSON as a dict; the schema of every payload is
  the generated OpenAPI document: https://api.machinevision.cloud/openapi.json
- Errors raise `MachineVisionError` with the API's stable `code`
  (`SPEND_CAP_EXCEEDED`, `TASK_NOT_FOUND`, ...).
- Pass `idempotency_key=` to `inspect()` to make retries safe — replays are not
  re-billed.
- Tune accuracy with visual contexts: `create_context`, `add_context_item`,
  then `update_task(task_id, inference_config={"mode": "few_shot", "context_id": ...})`.

Docs: https://machinevision.cloud/docs/quickstart
