Skip to main content

OpenAI Provider

💡TL;DR

The OpenAI preset uses Notemd's shared OpenAI-compatible Chat Completions runtime. Configure an API key and a model accepting that protocol. The preset model is a starting value, not a guarantee of current account access or a recommendation for every task.

Setup​

Select OpenAI in Notemd settings, enter the API key, and keep https://api.openai.com/v1 for the official endpoint. Choose a model available to the account and compatible with chat completions. Models exposed only through another API are not made compatible by appearing in a model list.

For a gateway, prefer its matching preset so its endpoint and request headers are used. For Azure deployments, use Azure OpenAI and configure the deployment and API version separately.

Endpoint And Authentication​

FieldPreset value
Transportopenai-compatible
API keyRequired
Base URLhttps://api.openai.com/v1
Default modelgpt-4o
Temperature0.5
Connection testmodels-then-chat

The runtime normalizes known base-URL forms before constructing the model-list or chat route. Start with the preset base URL and inspect the resolved route when debugging, rather than appending extra endpoint segments by guesswork.

Model Discovery​

Fetch models queries the configured endpoint. Test connection can succeed on the model-list request without validating chat generation. Run a disposable note through the exact provider/model after testing connectivity.

Reasoning, thinking, sampling and output-token controls depend on the preset and model-specific runtime rules. Validate the combination used by your task; listing a model does not guarantee acceptance of every parameter.

Troubleshooting​

  • 401 or 403: check API credentials, project permissions, model entitlement and account status.
  • 404: check endpoint, model ID and whether the selected API supports that model.
  • Model listing works but the task fails: inspect the actual chat error and requested model/settings.
  • Long output fails or truncates: review the requested output limit, source size and provider response; a short connection test cannot reproduce every long-request failure.

When To Use​

Use this preset for direct OpenAI models accepting the implemented Chat Completions protocol. Evaluate a representative note before choosing it for a workload. Provider pricing and model availability can change independently of a Notemd release.

Next Steps​