Cherry Studio with Nebius Token Factory
Cherry Studio already supports custom OpenAI-compatible providers. This recipe configures its existing OpenAI Chat Completions endpoint for Token Factory. It does not add a provider adapter or enable stateful Responses behavior.
Cherry Studio stores custom providers in its application database, so config.json is a reviewable reference manifest—not a file to import into the application. Follow the UI steps below.
Prerequisites
- Cherry Studio installed
- a Token Factory API key
Add the provider
- Open Settings → Model Services.
- Select Add Provider (
+), then Add Custom Provider. - Enter:
| Field | Value |
|---|---|
| Provider name | Nebius Token Factory |
| API key | Your Token Factory key |
| OpenAI / Chat Completions Base URL | https://api.tokenfactory.nebius.com/v1 |
- Keep OpenAI / Chat Completions selected as the default text endpoint.
- Leave OpenAI Responses under More Endpoints empty. Token Factory's current Responses surface is stateless Stage 1 and is not the default wire contract for this recipe.
- Save the provider.
Cherry Studio previews the final Chat Completions request as:
https://api.tokenfactory.nebius.com/v1/chat/completions
Do not paste the full /chat/completions path into the Base URL field; Cherry Studio appends it.
Discover and add models
- Select the new Nebius Token Factory provider.
- Choose Get model list / Pull models.
- Cherry Studio calls
GET https://api.tokenfactory.nebius.com/v1/modelswith the configured bearer key. - Select a current coding model and apply the changes. This recipe was checked with
moonshotai/Kimi-K2.7-Code. - If you add the model manually instead, use the exact raw model ID and set its chat protocol/endpoint type to OpenAI Chat Completions. Do not prepend
openai/ornebius/. - Enable the provider, choose the model in a chat, and send a small prompt.
Model availability changes independently of Cherry Studio. Prefer Get model list over copying a static catalog, and re-pull before relying on a model in a durable workflow.
Scope
This setup covers Token Factory's OpenAI-compatible Chat Completions path, including normal Cherry Studio chat and model features supported by the selected model.
It deliberately does not configure Cherry Studio's separate OpenAI Responses endpoint. Do not infer support for previous_response_id, server-side conversation chaining, or replay of prior response output/reasoning items from Cherry Studio's ability to speak that protocol to other providers.
Validate the reference offline
The reference manifest contains no API key. Validate it with Node's built-in tools:
node validate-config.mjs
node --test validate-config.test.mjs
The checks make no network calls. They verify the provider name, canonical Base URL, Chat Completions default and model endpoint, derived /chat/completions and /models routes, raw model ID, absence of a Responses endpoint, and absence of stored secrets.
Troubleshooting
401 Unauthorized
Open the provider's authentication section and replace the API key with a current Token Factory key. A key is stored by Cherry Studio locally; never add it to config.json, source control, screenshots, or support logs.
404 or wrong host
The Base URL must be exactly https://api.tokenfactory.nebius.com/v1. Retired api.studio.nebius.* hosts are not valid. Do not omit /v1, and do not append /chat/completions or /models yourself.
Get model list fails
Confirm the provider has an enabled API key and the Chat Completions Base URL is set. The generic Cherry Studio model fetcher derives /models from that Base URL and sends bearer authentication. You can compare the upstream response directly:
curl -sS https://api.tokenfactory.nebius.com/v1/models \
-H "Authorization: Bearer $NEBIUS_API_KEY"
Model not found
Pull the model list again. If the model was removed, select another current model suitable for coding and tool use. Keep the provider-owned model ID unchanged.
Requests unexpectedly use Responses
Edit the provider and confirm OpenAI / Chat Completions is the default endpoint. Remove any OpenAI Responses Base URL, then edit the model and ensure its endpoint type contains only OpenAI Chat Completions.
Evidence checked
The recipe follows Cherry Studio's current public source behavior:
- custom providers expose independent OpenAI Chat Completions and OpenAI Responses URL fields;
- Chat Completions is the default when it is the configured text endpoint;
- request previews append
/chat/completionsto the configured API root; and - the generic model fetcher appends
/modelsand sends bearer authentication.
No Cherry Studio provider preset or adapter is required.