feat: requested changes

Author: NorbertKlockiewicz

docs/docs/03-hooks/01-natural-language-processing/useLLM.md

@@ -531,7 +531,7 @@ The `imagePath` should be a local file path on the device.
 
 ### Functional generation with images
 
-You can also use `generate` directly, passing `imagePaths` as the third argument:
+You can also use `generate` directly by setting `mediaPath` on user messages:
 
 ```tsx
 const llm = useLLM({ model: LFM2_VL_1_6B_QUANTIZED });
@@ -540,14 +540,12 @@ const handleGenerate = async () => {
   const chat: Message[] = [
     {
       role: 'user',
-      content: [
-        { type: 'image' },
-        { type: 'text', text: 'Describe this image.' },
-      ],
+      content: 'Describe this image.',
+      mediaPath: '/path/to/image.jpg',
     },
   ];
 
-  const response = await llm.generate(chat, undefined, ['/path/to/image.jpg']);
+  const response = await llm.generate(chat);
   console.log(response);
 };
 ```

docs/docs/04-typescript-api/01-natural-language-processing/LLMModule.md

@@ -147,20 +147,18 @@ const response = await llm.sendMessage('What is in this image?', {
 });
 ```
 
-Or use [`generate`](../../06-api-reference/classes/LLMModule.md#generate) with `imagePaths` directly:
+Or use [`generate`](../../06-api-reference/classes/LLMModule.md#generate) with `mediaPath` on the message:
 
 ```typescript
 const chat: Message[] = [
   {
     role: 'user',
-    content: [
-      { type: 'image' },
-      { type: 'text', text: 'Describe this image.' },
-    ],
+    content: 'Describe this image.',
+    mediaPath: '/path/to/image.jpg',
   },
 ];
 
-const response = await llm.generate(chat, undefined, ['/path/to/image.jpg']);
+const response = await llm.generate(chat);
 ```
 
 ## Deleting the model from memory

packages/react-native-executorch/common/rnexecutorch/threads/GlobalThreadPool.h

@@ -35,6 +35,8 @@ class GlobalThreadPool {
       }
 
       numThreads = std::max(numThreads.value(), 2u);
+      log(rnexecutorch::LOG_LEVEL::Info, "Initializing global thread pool with",
+          numThreads, "threads");
       instance = std::make_unique<HighPerformanceThreadPool>(numThreads.value(),
                                                              config);
       // Disable OpenCV's internal threading to prevent it from overriding our

packages/react-native-executorch/src/constants/modelUrls.ts

@@ -385,7 +385,7 @@ export const LFM2_VL_1_6B_QUANTIZED = {
   modelSource: LFM2_VL_1_6B_QUANTIZED_MODEL,
   tokenizerSource: LFM2_VL_TOKENIZER,
   tokenizerConfigSource: LFM2_VL_TOKENIZER_CONFIG,
-};
+} as const;
 
 // Classification
 const EFFICIENTNET_V2_S_MODEL =

packages/react-native-executorch/src/controllers/LLMController.ts

@@ -288,8 +288,7 @@ export class LLMController {
 
   public async generate(
     messages: Message[],
-    tools?: LLMTool[],
-    imagePaths?: string[]
+    tools?: LLMTool[]
   ): Promise<string> {
     if (!this._isReady) {
       throw new RnExecutorchError(
@@ -309,6 +308,10 @@ export class LLMController {
       );
     }
 
+    const imagePaths = messages
+      .filter((m) => m.mediaPath)
+      .map((m) => m.mediaPath!);
+
     const renderedChat: string = this.applyChatTemplate(
       messages,
       this.tokenizerConfig,
@@ -317,7 +320,10 @@ export class LLMController {
       { tools_in_user_message: false, add_generation_prompt: true }
     );
 
-    return await this.forward(renderedChat, imagePaths);
+    return await this.forward(
+      renderedChat,
+      imagePaths.length > 0 ? imagePaths : undefined
+    );
   }
 
   public async sendMessage(
@@ -367,14 +373,9 @@ export class LLMController {
         countTokensCallback
       );
 
-    const imagePaths = messageHistoryWithPrompt
-      .filter((m) => m.mediaPath)
-      .map((m) => m.mediaPath!);
-
     const response = await this.generate(
       messageHistoryWithPrompt,
-      this.toolsConfig?.tools,
-      imagePaths.length > 0 ? imagePaths : undefined
+      this.toolsConfig?.tools
     );
 
     if (!this.toolsConfig || this.toolsConfig.displayToolCalls) {

packages/react-native-executorch/src/hooks/natural_language_processing/useLLM.ts

@@ -97,9 +97,9 @@ export function useLLM({
   );
 
   const generate = useCallback(
-    (messages: Message[], tools?: LLMTool[], imagePaths?: string[]) => {
+    (messages: Message[], tools?: LLMTool[]) => {
       setResponse('');
-      return controllerInstance.generate(messages, tools, imagePaths);
+      return controllerInstance.generate(messages, tools);
     },
     [controllerInstance]
   );

packages/react-native-executorch/src/modules/natural_language_processing/LLMModule.ts

@@ -108,17 +108,14 @@ export class LLMModule {
 
   /**
    * Runs model to complete chat passed in `messages` argument. It doesn't manage conversation context.
+   * For multimodal models, set `mediaPath` on user messages to include images.
    *
-   * @param messages - Array of messages representing the chat history.
+   * @param messages - Array of messages representing the chat history. User messages may include a `mediaPath` field with a local image path.
    * @param tools - Optional array of tools that can be used during generation.
    * @returns The generated response as a string.
    */
-  async generate(
-    messages: Message[],
-    tools?: LLMTool[],
-    imagePaths?: string[]
-  ): Promise<string> {
-    return await this.controller.generate(messages, tools, imagePaths);
+  async generate(messages: Message[], tools?: LLMTool[]): Promise<string> {
+    return await this.controller.generate(messages, tools);
   }
 
   /**

packages/react-native-executorch/src/types/llm.ts

@@ -103,16 +103,13 @@ export interface LLMTypeBase {
   getGeneratedTokenCount: () => number;
   /**
    * Runs model to complete chat passed in `messages` argument. It doesn't manage conversation context.
+   * For multimodal models, set `mediaPath` on user messages to include images.
    *
-   * @param messages - Array of messages representing the chat history.
+   * @param messages - Array of messages representing the chat history. User messages may include a `mediaPath` field with a local image path.
    * @param tools - Optional array of tools that can be used during generation.
    * @returns The generated tokens as `string`.
    */
-  generate: (
-    messages: Message[],
-    tools?: LLMTool[],
-    imagePaths?: string[]
-  ) => Promise<string>;
+  generate: (messages: Message[], tools?: LLMTool[]) => Promise<string>;
   /**
    * Returns the number of total tokens from the previous generation. This is a sum of prompt tokens and generated tokens.
    *