Use ZhiFlo in an OpenAI-compatible client

Use this path when your client is not listed here but accepts a custom OpenAI service. The goal is not merely to fill the fields. The goal is to get that client to return OK.

1. Field mapping

Client field	ZhiFlo value	Meaning
Provider type / API type	OpenAI or OpenAI-compatible	Selects the compatible request format
Base URL / API URL	https://api.zhiflo.com/v1	Address where requests are sent
API Key / Token	A ZhiFlo Token for this device	Password for API access
Model / Model ID	A complete model ID from your account	Exact model recognized by the API

If the client also asks for a display name, use ZhiFlo. A display name never replaces the model ID.

2. Save, enable, and select the model

1. Find the client’s Provider, Model service, or API settings.

2. Add a custom provider.

3. Select the OpenAI-compatible type.

4. Enter the Base URL from the table.

5. Paste the Token into the dedicated API Key or Token field.

   Keep the Token out of the Base URL, provider name, model name, and screenshots.

6. Add the complete model ID shown for your account.

7. Save the provider.

8. Enable the provider.

9. Make the new model current.

10. If the client explicitly supports live configuration updates, reload its model list; otherwise, exit and reopen the client.

A saved provider is not necessarily active. Confirm both the enable state and current model selection.

3. Base URL versus full endpoint

Most clients ask for a Base URL. Enter:

https://api.zhiflo.com/v1

Only when a field explicitly requests the full Chat Completions endpoint, enter:

https://api.zhiflo.com/v1/chat/completions

Putting the full endpoint in a Base URL field may make the client append /chat/completions a second time and return 404.

4. Send a minimal message

Select the ZhiFlo model and send:

Reply with OK only.

An OK reply confirms a real request. A passed API-key check, green status, or visible model name cannot replace this step.

If it fails, check in this order: provider type, Token state, model ID, Base URL, quota, then network. Match the visible symptom in Troubleshoot connection problems.

Next: Make your first API request.