Method sendPrompt

app/src/utils/ModelManager.ts:650–698

* Send a prompt through the appropriate backend for the given model. * - Local models (browser_local, llama_cpp_local): routed directly, no inference params * - Remote models: fetches per-model params and delegates to sendApi.fetchResponse

(
    modelName: string,
    preprocessResult: PreProcessorResult,
    token?: string,
    enableStreaming: boolean = false,
    onStreamChunk?: (chunk: string) => void,
    onReasoningChunk?: (chunk: string) => void
  )

   * - Remote models: fetches per-model params and delegates to sendApi.fetchResponse
   */
  public async sendPrompt(
    modelName: string,
    preprocessResult: PreProcessorResult,
    token?: string,
    enableStreaming: boolean = false,
    onStreamChunk?: (chunk: string) => void,
    onReasoningChunk?: (chunk: string) => void
  ): Promise<string> {
    // Resolve model → server
    let modelsResponse = this.listModels();
    let model = modelsResponse.models.find(m => m.name === modelName);
    if (!model) {
      modelsResponse = await this.fetchModels();
      model = modelsResponse.models.find(m => m.name === modelName);
    }
    if (!model) throw new Error(`Model '${modelName}' not found in available models`);

    const serverAddress = model.server;

    // Build messages from preprocessResult
    const hasImages = preprocessResult.images && preprocessResult.images.length > 0;
    let content: any = preprocessResult.modifiedPrompt;
    if (hasImages) {
      content = [
        { type: 'text', text: preprocessResult.modifiedPrompt },
        ...preprocessResult.images!.map(img => ({
          type: 'image',
          image: `data:image/png;base64,${img}`,
        })),
      ];
    }
    const messages = [{ role: 'user', content }];

    // Local models: route directly, they manage their own settings
    if (this.isLocalModel(serverAddress)) {
      if (!this.isLocalModelReady(serverAddress)) {
        throw new Error('Local model not loaded. Please load it from the Add Model panel.');
      }
      return this.generateWithLocalModel(serverAddress, messages, onStreamChunk, onReasoningChunk);
    }

    // Remote models: attach per-model inference params
    if (serverAddress.includes('api.observer-ai.com') && token) {
      this.optimisticUpdateQuota();
    }
    const { fetchResponse } = await import('./sendApi');
    const params = { ...DEFAULT_INFERENCE_PARAMS, ...this.getModelParams(modelName) };
    return fetchResponse(serverAddress, messages, modelName, token, enableStreaming, onStreamChunk, params, onReasoningChunk);
  }

  // ===========================================================================
  // Utilities