langgenius/dify · error · HttpClientError

server_4xx_other

server_4xx_other

Error message

device/code: HTTP ${res.status}

What it means

Raised when HitTestingService.retrieve raises LLMBadRequestError, which perform_hit_testing translates into ProviderNotInitializeError with this message. It indicates the configured model provider cannot fulfil the embedding or reranking call — most often because no embedding/reranking model is configured for the tenant, or the configured provider's credentials are missing/invalid. This is a configuration error, not a dataset or query error.

Source

Thrown at cli/src/api/oauth-device.ts:82

export class DeviceFlowApi {
  private readonly http: HttpClient

  constructor(http: HttpClient) {
    this.http = http
  }

  async requestCode(req: CodeRequest): Promise<CodeResponse> {
    if (req.device_label === '') {
      throw new BaseError({
        code: ErrorCode.UsageMissingArg,
        message: 'device_label is required',
      })
    }
    const body = { client_id: req.client_id ?? DEFAULT_CLIENT_ID, device_label: req.device_label }
    const res = await this.http.fetch('oauth/device/code', { method: 'POST', json: body })
    if (res.status === 404) throw versionSkew()
    if (!res.ok) {
      throw new HttpClientError({
        code: ErrorCode.Server4xxOther,
        message: `device/code: HTTP ${res.status}`,
        httpStatus: res.status,
      })
    }
    return (await res.json()) as CodeResponse
  }

  async pollOnce(req: PollRequest): Promise<PollResult> {
    if (req.device_code === '') {
      throw new BaseError({
        code: ErrorCode.UsageMissingArg,
        message: 'device_code is required',
      })
    }
    const body = { client_id: req.client_id ?? DEFAULT_CLIENT_ID, device_code: req.device_code }
    const res = await this.http.fetch('oauth/device/token', { method: 'POST', json: body })
    if (res.status === 404) throw versionSkew()

View on GitHub (pinned to ef8544b173)

Solutions

1. Open Settings -> Model Provider in the console and configure a valid embedding (and, if used, reranking) provider.
2. Verify the dataset's configured retrieval_model references providers that are currently enabled for the tenant.
3. After configuring, re-run the hit-testing request; no code change is required.

Example fix

// before — dataset retrieval_model references a provider with no credentials
retrieval_model: { reranking_enable: true, reranking_model: { provider: 'cohere', model: 'rerank-multilingual-v3.0' } }
// after — configure provider in Settings -> Model Provider, or disable reranking
retrieval_model: { reranking_enable: false }

async function hasEmbeddingProvider(client): Promise<boolean> {
  const r = await client.get('/console/api/workspaces/current/model-providers');
  const providers = await r.json();
  return providers.some(p => p.models?.some(m => m.model_type === 'text-embedding') && p.is_valid);
}
if (!(await hasEmbeddingProvider(client))) {
  throw new Error('Configure an embedding provider in Settings -> Model Provider first');
}

interface ProviderConfig { model_type: string; is_valid: boolean; }
function hasValidEmbedding(providers: ProviderConfig[]): boolean {
  return providers.some(p => p.model_type === 'text-embedding' && p.is_valid);
}

try {
  await hitTesting(client, datasetId, query);
} catch (e) {
  if (e.response?.status === 400 && /Embedding Model/i.test(e.response.data?.message || '')) {
    redirectTo('/settings/model-provider');  // guide user to configure
    return;
  }
  throw e;
}

When it happens

Trigger: POST /datasets/{dataset_id}/hit-testing on a dataset whose indexing_mode requires an embedding model, but the tenant has no default embedding provider configured; or a reranking model is selected in retrieval_model but the provider credentials were removed.

Common situations: Freshly deployed instance where the model provider was never set up; tenant admin removed a provider credential that the dataset relies on; switched dataset retrieval mode to one requiring reranking without configuring a rerank model.

Related errors

• provider_not_initialize
• provider_not_initialize
• provider_not_support_speech_to_text
• model_currently_not_support
• provider_not_support_speech_to_text

AI-assisted analysis of langgenius/dify@ef8544b173 (2026-08-12). Data as JSON: /api/errors/a032426f97eab2be. Report an issue: GitHub.