Skip to content
Knotest Learn
Join waitlist

Connect an inference API

Verify a sample request and map the response to predictions.

An API profile connects a Dataset to an external prediction service. It is separate from the provider that powers the Knotest assistant.

  1. Configure the request

    In Vault → API profiles, add a profile. Set the connection and authentication. Import from cURL can fill a draft for review.

    For images, choose a JSON field with base64 or a data URI, multipart upload, raw bytes, or no body for a GET endpoint. For tables, map the feature object into a JSON field. Advanced settings include headers, query parameters, batching, limits, and asynchronous jobs.

  2. Send a real sample

    Choose one Dataset sample and inspect the response. Map the response fields to task-specific predictions. Check class names, coordinates, and other task information against the Dataset.

  3. Verify the saved profile

    Use Verify with the current settings and credential. A changed or stale profile needs verification before generating predictions. Resolve the returned problem rather than assuming that a successful HTTP response means the prediction mapping is correct.

Use the connection deliberately

Next steps

Where your data goes

Understand the host, provider, and sharing choices behind the work.

Prepare and run tests

Select the inputs, inspect the package, and approve execution.