Add model parameter to InvokeLLM for per-call model selection

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: faygade

src/modules/integrations.types.ts

@@ -46,6 +46,11 @@ export type IntegrationPackage = {
 export interface InvokeLLMParams {
   /** The prompt text to send to the model */
   prompt: string;
+  /** Optionally specify a model to override the app-level model setting for this specific call.
+   *
+   * Options: `"gpt_5_mini"`, `"gemini_3_flash"`, `"gpt_5"`, `"gemini_3_pro"`, `"claude_sonnet_4_6"`, `"claude_opus_4_6"`
+   */
+  model?: 'gpt_5_mini' | 'gemini_3_flash' | 'gpt_5' | 'gemini_3_pro' | 'claude_sonnet_4_6' | 'claude_opus_4_6';
   /** If set to `true`, the LLM will use Google Search, Maps, and News to gather real-time context before answering.
    * @default false
    */