feat(streaming): add chatStreaming method with token callback for real-time generation progress

- Add StreamingChatOptions interface with onToken callback
- Add optional chatStreaming method to MultiModalModel abstract class
- Implement chatStreaming in OllamaProvider using collectStreamResponse

ts/abstract.classes.multimodal.ts

@@ -17,6 +17,14 @@ export interface ChatOptions {
   messageHistory: ChatMessage[];
 }
 
+/**
+ * Options for streaming chat interactions
+ */
+export interface StreamingChatOptions extends ChatOptions {
+  /** Callback fired for each token during generation */
+  onToken?: (token: string) => void;
+}
+
 /**
  * Response format for chat interactions
  */
@@ -152,7 +160,7 @@ export abstract class MultiModalModel {
    * @returns Promise resolving to the assistant's response
    */
   public abstract chat(optionsArg: ChatOptions): Promise<ChatResponse>;
-  
+
   /**
    * Streaming interface for chat interactions
    * Allows for real-time responses from the model
@@ -161,6 +169,14 @@ export abstract class MultiModalModel {
    */
   public abstract chatStream(input: ReadableStream<Uint8Array>): Promise<ReadableStream<string>>;
 
+  /**
+   * Streaming chat with token callback
+   * Calls onToken for each token generated, returns final response
+   * @param optionsArg Options containing system message, user message, message history, and onToken callback
+   * @returns Promise resolving to the assistant's response
+   */
+  public chatStreaming?(optionsArg: StreamingChatOptions): Promise<ChatResponse>;
+
   /**
    * Text-to-speech conversion
    * @param optionsArg Options containing the message to convert to speech

ts/provider.ollama.ts

@@ -9,7 +9,8 @@ import type {
   ResearchResponse,
   ImageGenerateOptions,
   ImageEditOptions,
-  ImageResponse
+  ImageResponse,
+  StreamingChatOptions
 } from './abstract.classes.multimodal.js';
 
 /**
@@ -235,6 +236,34 @@ export class OllamaProvider extends MultiModalModel {
     };
   }
 
+  /**
+   * Streaming chat with token callback (implements MultiModalModel interface)
+   * Calls onToken for each token generated during the response
+   */
+  public async chatStreaming(optionsArg: StreamingChatOptions): Promise<ChatResponse> {
+    const onToken = optionsArg.onToken;
+
+    // Use existing collectStreamResponse with callback
+    const response = await this.collectStreamResponse(
+      {
+        systemMessage: optionsArg.systemMessage,
+        userMessage: optionsArg.userMessage,
+        messageHistory: optionsArg.messageHistory,
+      },
+      (chunk) => {
+        if (onToken) {
+          if (chunk.thinking) onToken(chunk.thinking);
+          if (chunk.content) onToken(chunk.content);
+        }
+      }
+    );
+
+    return {
+      role: 'assistant' as const,
+      message: response.message,
+    };
+  }
+
   /**
    * Streaming chat with async iteration and options support
    */