linshenkx/prompt-optimizer · warning · RequestConfigError

Model is not enabled

Error message

Model is not enabled

What it means

Thrown by validateModelConfig when the model config is not enabled, unless the caller passes { allowDisabled: true }. Disabled models are deliberately blocked from normal requests so users don't accidentally hit deactivated providers; only connection testing bypasses the check.

Source

Thrown at packages/core/src/services/llm/service.ts:75

   * 验证模型配置
   */
  private validateModelConfig(
    modelConfig: TextModelConfig,
    options: { allowDisabled?: boolean } = {}
  ): void {
    if (!modelConfig) {
      throw new RequestConfigError('Model config cannot be empty');
    }
    if (!modelConfig.providerMeta || !modelConfig.providerMeta.id) {
      throw new RequestConfigError('Model provider metadata cannot be empty');
    }
    if (!modelConfig.modelMeta || !modelConfig.modelMeta.id) {
      throw new RequestConfigError('Model metadata cannot be empty');
    }
    // Default behavior: disabled models cannot be used for normal requests.
    // Connection testing is allowed to bypass this check (align with image model test behavior).
    if (!options.allowDisabled && !modelConfig.enabled) {
      throw new RequestConfigError('Model is not enabled');
    }
  }

  /**
   * 发送消息（结构化格式）
   */
  async sendMessageStructured(messages: Message[], provider: string): Promise<LLMResponse> {
    try {
      if (!provider) {
        throw new RequestConfigError('Model provider cannot be empty');
      }

      const modelConfig = await this.modelManager.getModel(provider);
      if (!modelConfig) {
        throw new RequestConfigError(`Model ${provider} not found`);
      }

      this.validateModelConfig(modelConfig);

View on GitHub (pinned to 3e677b1d9f)

Solutions

1. Enable the model in your model manager / settings before sending messages (config.enabled = true)
2. If the disabled state is unexpected, inspect the stored config to see why enabled is false or missing
3. Ensure new model configs are created with enabled explicitly set to true when they should be usable

Example fix

// before
await modelManager.updateModel('openai', { enabled: false });
await llmService.sendMessageStream(msgs, 'openai', onChunk); // throws

// after
await modelManager.updateModel('openai', { enabled: true });
await llmService.sendMessageStream(msgs, 'openai', onChunk);

const cfg = await modelManager.getModel(provider);
if (!cfg?.enabled) {
  await modelManager.updateModel(provider, { enabled: true }); // or block the send
}

const isEnabledModel = (c?: ModelConfig | null): c is ModelConfig => !!c && c.enabled === true;

catch (e) { if (e instanceof RequestConfigError && e.message === 'Model is not enabled') { promptUserToEnable(provider); } }

When it happens

Trigger: Calling sendMessageStructured, sendMessageStream, or sendMessageStreamWithTools on a model whose config.enabled is false (or undefined, which is falsy). testConnection does not throw this because it calls validateModelConfig with allowDisabled: true.

Common situations: User disabled a provider in the settings UI but a chat session still references it; a persisted config where the enabled flag defaults to false; toggling a model off while background jobs still use its id.

Related errors

• Model provider metadata cannot be empty
• Model metadata cannot be empty
• Evaluation mode configuration must not be empty.
• Workspace prompt must not be empty.
• Image understanding service is not available for multimodal

AI-assisted analysis of linshenkx/prompt-optimizer@3e677b1d9f (2026-08-27). Data as JSON: /api/errors/9956f34db8141946. Report an issue: GitHub.