Skip to main content

Chat with Cherry Studio

Use a Primex model in a Cherry Studio desktop chat assistant. Start with a model from Cherry Studio Setup, which lists live offerings that publish Chat Completions support.

Add Primex as a provider​

This guide reviews Cherry Studio 2.1.8. Download the desktop application from its official releases. The official custom-provider guide describes the current settings; labels can change between releases.

  1. Open Settings → Model Services → Add Provider.
  2. Name the custom provider Primex. Add its OpenAI endpoint and enter https://gateway.primexos.com/v1 as the base URL.
  3. Paste a dedicated Primex API key into the provider's API Key field, save, and enable the provider.
  4. Add your exact offering ID manually. Choose the model's OpenAI / Chat Completions endpoint, then select that provider and model in your chat assistant.
FieldValue
Provider namePrimex
OpenAI base URLhttps://gateway.primexos.com/v1
API keyA dedicated key from API Keys
Model IDThe exact live ID selected in Setup
Model endpointOpenAI / Chat Completions

The values copied from Setup are a settings checklist. Enter them in the provider and model forms. Keep the key out of chat messages and shared configuration or backup exports.

Cherry Studio supports several endpoint types. This guide configures only Chat Completions. Selecting Responses or adding an Anthropic endpoint changes the request contract; verify the selected model's published support before using those paths. See API formats.

Choose the model deliberately​

Manual selection gives you an exact model ID for the first check. Get model list can discover models, but a returned model name does not qualify every endpoint, tool or input type. Primex Setup remains the guide's compatibility filter.

If you edit the model's context capacity, use its current published Catalog limit. Do not treat a client default as the model's capacity. Check missing limits before long conversations. Leave image, reasoning, tool and agent capabilities disabled until their exact model/client path is verified.

Default models for translation, topic naming and other assistant features can cause additional requests. Review them separately. Local price fields or client cost estimates do not replace Primex Usage.

Verify a conversation​

Start a new ordinary chat without a knowledge base, MCP or tools.

  1. Send a short greeting and wait for the complete answer.
  2. Send a follow-up that refers to the first answer. Confirm that the conversation history is retained and the follow-up receives a complete response.
  3. In Usage, check the exact model, API key and both requests. Review extra entries caused by model checks or auxiliary features.

Treat a connection-check button as a request that may perform inference. A desktop spinner, failed check or partial stream does not establish the final account charge. Inspect Usage before retrying. This guide does not establish the desktop application's deadlines or automatic retry behavior.

Common fixes​

  • 401: check the provider's key field and whether the dedicated key is active.
  • Model missing or wrong endpoint: enable Primex, add the exact ID manually, and select OpenAI / Chat Completions for that model.
  • 404 or a duplicated path: use the /v1 base URL, not the full /v1/chat/completions URL in a base-URL field. Review the endpoint actually sent by your installed release.
  • 429, incomplete replies or timeouts: inspect Usage and the error guide before repeating the prompt.
  • Unexpected requests or charges: review auxiliary/default models, model checks, tools and knowledge-base settings independently.

Remove the connection​

Choose another assistant model, disable or remove the Primex provider, and review any default-model assignments. Remove stored credentials and any shared backup containing them. Revoke the dedicated Primex key when it is no longer needed.

Need help connecting? Report a setup issue.