Connect ChatBox
Use your Primex key in the ChatBox app. You need access to custom model-provider settings and AI credit for requests. Choose a model in ChatBox Setup.
Add a custom provider
Open Settings → Model Provider, add a Custom Provider, and choose the OpenAI API type for Chat Completions.
| Setting | Value |
|---|---|
| Provider name | Primex |
| API type | OpenAI / Chat Completions |
| API Host | https://gateway.primexos.com/v1 |
| API Path | /chat/completions |
| API Key | Your dedicated Primex key, entered privately |
| Model ID | The exact ID selected in Setup |
| Conversation Temperature | 1 |
Save and select this provider/model in a new conversation. The combined request
URL is https://gateway.primexos.com/v1/chat/completions. Use the separate host/path
fields; avoid entering /v1 twice. Model discovery uses the host's /models route.
If your version offers a model check, follow it with the first conversation below.
The settings JSON in Setup is a checklist, not a ChatBox backup/import format. Keep keys out of conversations and review exports or synchronization settings before sharing them. Your Primex requests use Primex credit; optional ChatBox hosted services have their own terms and billing.
First conversation
Select Primex explicitly and leave attachments, search and other tools off. Send “Reply with a short greeting”, then ask about the answer. Confirm complete replies and the selected model in Usage.
Use Temperature1 for models that reject other sampling values. Keep output limits within the selected model's published bounds. If a reply is interrupted, check Usage before repeating it rather than assuming no work completed.
Other provider types
OpenAI Responses, Anthropic and Gemini are distinct adapter choices. This guide uses Chat Completions; switching the dropdown requires the corresponding protocol, base URL and compatible model. Do not select Gemini for this Gateway configuration. Tools, vision, reasoning controls and embedding workflows have their own model and client requirements.
Fix a connection or remove it
- 401: check the key in this provider, not another saved provider.
- 404: inspect API Host, API Path and the exact model ID.
- 400: remove unsupported options and check Temperature/limits.
- 402 / 429 / stream failure: inspect Errors & limits and Usage.
Choose another provider in dependent conversations, remove the Primex custom provider and revoke the dedicated key when no longer needed.
See the official project and custom OpenAI provider source. Report a setup issue.