community[minor]: upgraded @mlc/web-llm dependency and updated it's ChatModel (#5637)

* chore(community/webllm): upgraded @mlc/web-llm dependency and updated it's ChatModel

* Format

---------

Co-authored-by: jacoblee93 <jacoblee93@gmail.com>

Author: tnfssc

libs/langchain-community/package.json

@@ -86,7 +86,7 @@
     "@langchain/scripts": "~0.0.14",
     "@layerup/layerup-security": "^1.5.12",
     "@mendable/firecrawl-js": "^0.0.13",
-    "@mlc-ai/web-llm": "^0.2.35",
+    "@mlc-ai/web-llm": "^0.2.40",
     "@mozilla/readability": "^0.4.4",
     "@neondatabase/serverless": "^0.9.1",
     "@notionhq/client": "^2.2.10",
@@ -245,7 +245,7 @@
     "@huggingface/inference": "^2.6.4",
     "@layerup/layerup-security": "^1.5.12",
     "@mendable/firecrawl-js": "^0.0.13",
-    "@mlc-ai/web-llm": "^0.2.35",
+    "@mlc-ai/web-llm": "^0.2.40",
     "@mozilla/readability": "*",
     "@neondatabase/serverless": "*",
     "@notionhq/client": "^2.2.10",

libs/langchain-community/src/chat_models/webllm.ts

@@ -43,7 +43,7 @@ export interface WebLLMCallOptions extends BaseLanguageModelCallOptions {}
 export class ChatWebLLM extends SimpleChatModel<WebLLMCallOptions> {
   static inputs: WebLLMInputs;
 
-  protected engine: webllm.EngineInterface;
+  protected engine: webllm.MLCEngine;
 
   appConfig?: webllm.AppConfig;
 
@@ -63,40 +63,33 @@ export class ChatWebLLM extends SimpleChatModel<WebLLMCallOptions> {
     this.chatOptions = inputs.chatOptions;
     this.model = inputs.model;
     this.temperature = inputs.temperature;
+    this.engine = new webllm.MLCEngine();
   }
 
   _llmType() {
     return "web-llm";
   }
 
   async initialize(progressCallback?: webllm.InitProgressCallback) {
-    this.engine = new webllm.Engine();
     if (progressCallback !== undefined) {
       this.engine.setInitProgressCallback(progressCallback);
     }
     await this.reload(this.model, this.chatOptions, this.appConfig);
-    this.engine.setInitProgressCallback(() => {});
   }
 
   async reload(
     modelId: string,
     newAppConfig?: webllm.AppConfig,
     newChatOpts?: webllm.ChatOptions
   ) {
-    if (this.engine !== undefined) {
-      await this.engine.reload(modelId, newAppConfig, newChatOpts);
-    } else {
-      throw new Error("Initialize model before reloading.");
-    }
+    await this.engine.reload(modelId, newChatOpts, newAppConfig);
   }
 
   async *_streamResponseChunks(
     messages: BaseMessage[],
     options: this["ParsedCallOptions"],
     runManager?: CallbackManagerForLLMRun
   ): AsyncGenerator<ChatGenerationChunk> {
-    await this.initialize();
-
     const messagesInput: ChatCompletionMessageParam[] = messages.map(
       (message) => {
         if (typeof message.content !== "string") {
@@ -124,15 +117,12 @@ export class ChatWebLLM extends SimpleChatModel<WebLLMCallOptions> {
       }
     );
 
-    const stream = this.engine.chatCompletionAsyncChunkGenerator(
-      {
-        stream: true,
-        messages: messagesInput,
-        stop: options.stop,
-        logprobs: true,
-      },
-      {}
-    );
+    const stream = await this.engine.chat.completions.create({
+      stream: true,
+      messages: messagesInput,
+      stop: options.stop,
+      logprobs: true,
+    });
     for await (const chunk of stream) {
       // Last chunk has undefined content
       const text = chunk.choices[0].delta.content ?? "";
@@ -146,7 +136,7 @@ export class ChatWebLLM extends SimpleChatModel<WebLLMCallOptions> {
           },
         }),
       });
-      await runManager?.handleLLMNewToken(text ?? "");
+      await runManager?.handleLLMNewToken(text);
     }
   }
 

yarn.lock

@@ -9100,7 +9100,7 @@ __metadata:
     "@langchain/scripts": ~0.0.14
     "@layerup/layerup-security": ^1.5.12
     "@mendable/firecrawl-js": ^0.0.13
-    "@mlc-ai/web-llm": ^0.2.35
+    "@mlc-ai/web-llm": ^0.2.40
     "@mozilla/readability": ^0.4.4
     "@neondatabase/serverless": ^0.9.1
     "@notionhq/client": ^2.2.10
@@ -9267,7 +9267,7 @@ __metadata:
     "@huggingface/inference": ^2.6.4
     "@layerup/layerup-security": ^1.5.12
     "@mendable/firecrawl-js": ^0.0.13
-    "@mlc-ai/web-llm": ^0.2.35
+    "@mlc-ai/web-llm": ^0.2.40
     "@mozilla/readability": "*"
     "@neondatabase/serverless": "*"
     "@notionhq/client": ^2.2.10
@@ -10478,10 +10478,12 @@ __metadata:
   languageName: node
   linkType: hard
 
-"@mlc-ai/web-llm@npm:^0.2.35":
-  version: 0.2.35
-  resolution: "@mlc-ai/web-llm@npm:0.2.35"
-  checksum: 03c1d1847340f88474e1eeed7a91cc09e29299a1216e378385ffe5479c203d39a8656d98c9187864322453a91f046b874d7073662ab04033527079d9bb29bee3
+"@mlc-ai/web-llm@npm:^0.2.40":
+  version: 0.2.40
+  resolution: "@mlc-ai/web-llm@npm:0.2.40"
+  dependencies:
+    loglevel: ^1.9.1
+  checksum: 44d46178f7b7f899893ee8096fd4188b8c343589a10428c52f87b1b7e708f7a94b2b6315c8a6f8075f14d6d92aebfd8afc7f6d049a2ef60f8b8dc950b98a82e2
   languageName: node
   linkType: hard
 
@@ -28464,6 +28466,13 @@ __metadata:
   languageName: node
   linkType: hard
 
+"loglevel@npm:^1.9.1":
+  version: 1.9.1
+  resolution: "loglevel@npm:1.9.1"
+  checksum: e1c8586108c4d566122e91f8a79c8df728920e3a714875affa5120566761a24077ec8ec9e5fc388b022e39fc411ec6e090cde1b5775871241b045139771eeb06
+  languageName: node
+  linkType: hard
+
 "long@npm:*, long@npm:^5.2.1, long@npm:~5.2.3":
   version: 5.2.3
   resolution: "long@npm:5.2.3"