CherryHQ/cherry-studio · error · Error

ModelScope provider requires a non-empty `baseURL`.

Error message

ModelScope provider requires a non-empty `baseURL`.

What it means

Thrown by `createModelscopeProvider` when `settings.baseURL` is falsy. Unlike DashScope/DMXAPI, the image transport's `imageBaseURL` has a built-in default (`DEFAULT_MODELSCOPE_BASE_URL`), so this guard specifically protects the chat/embedding OpenAI-compatible endpoint which has no default. Chat and embedding calls need an explicit host.

Source

Thrown at src/main/ai/provider/custom/modelscope/modelscopeProvider.ts:51

 * registry (`resolveImageTransport`) so the job handler can rebuild the same
 * transport after a restart from the re-resolved provider settings.
 */
export function buildModelscopeTransport(settings: ModelscopeProviderSettings): ImageGenerationTransport {
  return createModelscopeTransport({
    apiKey: settings.apiKey ?? '',
    baseURL: settings.imageBaseURL || DEFAULT_MODELSCOPE_BASE_URL
  })
}

/**
 * Unified ModelScope (魔搭) provider: OpenAI-compatible chat/embedding off
 * `settings.baseURL`, plus an async submit/poll image transport off
 * `settings.imageBaseURL` (defaults to `https://api-inference.modelscope.cn`).
 */
export function createModelscopeProvider(settings: ModelscopeProviderSettings = {}): ModelscopeProvider {
  const { baseURL, fetch: customFetch } = settings
  if (!baseURL) {
    throw new Error('ModelScope provider requires a non-empty `baseURL`.')
  }

  const resolveApiKey = () =>
    loadApiKey({ apiKey: settings.apiKey, environmentVariableName: 'MODELSCOPE_API_KEY', description: 'ModelScope' })

  const authHeaders = () => ({
    Authorization: `Bearer ${resolveApiKey()}`,
    ...settings.headers
  })

  const url = ({ path }: { path: string; modelId: string }) => `${withoutTrailingSlash(baseURL)}${path}`

  const createChatModel = (modelId: string) =>
    new OpenAICompatibleChatLanguageModel(modelId, {
      provider: `${MODELSCOPE_PROVIDER_NAME}.chat`,
      url,
      headers: authHeaders,
      fetch: customFetch

View on GitHub (pinned to 726446b54c)

Solutions

1. Set `baseURL` on the ModelScope provider settings to the OpenAI-compatible host, e.g. `https://api-inference.modelscope.cn/v1/`.
2. Verify the provider config builder forwards the user-configured host.
3. If only using image generation, prefer `buildModelscopeTransport` directly (it has a default base).

Example fix

// before
createModelscopeProvider({ apiKey: 'ms-...' })
// after
createModelscopeProvider({ apiKey: 'ms-...', baseURL: 'https://api-inference.modelscope.cn/v1/' })

function assertModelscopeSettings(s: { baseURL?: string }) {
  if (!s.baseURL) throw new Error('ModelScope baseURL required before creating provider')
}
assertModelscopeSettings(settings)
const provider = createModelscopeProvider(settings)

const hasModelscopeBaseURL = (s: { baseURL?: string }): s is { baseURL: string } =>
  typeof s.baseURL === 'string' && s.baseURL.trim().length > 0

When it happens

Trigger: Instantiating the ModelScope provider with no `baseURL` — `createModelscopeProvider({})`. The image transport alone would still work (defaults apply), but chat/embedding would have no target.

Common situations: ModelScope provider added in settings with only an image endpoint configured; the API host field left blank; a migration dropped `baseURL`.

Related errors

• DashScope provider requires a non-empty `baseURL`. An empty
• DMXAPI provider requires a non-empty `baseURL` to build the
• DMXAPI provider requires a non-empty `baseURL`. An empty val
• Private key must be a non-empty string
• Cherry Assistant package configuration is invalid: ${invalid

AI-assisted analysis of CherryHQ/cherry-studio@726446b54c (2026-08-12). Data as JSON: /api/errors/4d4ba8691d1449e7. Report an issue: GitHub.