@blank-utils/llm 0.4.23 → 0.5.1

package/dist/{chunk-IOEC4YOD.js → chunk-S375B33F.js}

@@ -34,15 +34,8 @@ function checkWasm() {
 async function detectCapabilities() {
   const webgpu = await checkWebGPU();
   const wasm = checkWasm();
-  let recommendedBackend = "transformers";
-  let recommendedDevice = "wasm";
-  if (webgpu) {
-    recommendedBackend = "webllm";
-    recommendedDevice = "webgpu";
-  } else if (wasm) {
-    recommendedBackend = "transformers";
-    recommendedDevice = "wasm";
-  }
+  let recommendedBackend = "webllm";
+  let recommendedDevice = "webgpu";
   return {
     webgpu,
     wasm,
@@ -101,21 +94,6 @@ var WEBLLM_MODELS = {
   "hermes-3-llama-3.2-3b": "Hermes-3-Llama-3.2-3B-q4f16_1-MLC",
   "hermes-3-llama-3.1-8b": "Hermes-3-Llama-3.1-8B-q4f16_1-MLC"
 };
-var DEFAULT_TRANSFORMERS_MODEL = "onnx-community/Qwen2.5-0.5B-Instruct";
-var TRANSFORMERS_MODELS = {
-  // === Qwen 2.5 Models (Alibaba) - Excellent quality ===
-  "qwen-2.5-0.5b": "onnx-community/Qwen2.5-0.5B-Instruct",
-  "qwen-2.5-1.5b": "onnx-community/Qwen2.5-1.5B-Instruct",
-  "qwen-2.5-coder-0.5b": "onnx-community/Qwen2.5-Coder-0.5B-Instruct",
-  "qwen-2.5-coder-1.5b": "onnx-community/Qwen2.5-Coder-1.5B-Instruct",
-  // === Vision Models ===
-  "phi-3.5-vision": "onnx-community/Phi-3.5-vision-instruct",
-  // === SmolLM2 Models (HuggingFace) - Ultra lightweight ===
-  "smollm2-135m": "HuggingFaceTB/SmolLM2-135M-Instruct",
-  "smollm2-1.7b": "HuggingFaceTB/SmolLM2-1.7B-Instruct",
-  // === Phi Models (Microsoft) ===
-  "phi-3-mini": "Xenova/Phi-3-mini-4k-instruct"
-};
 
 // src/backends/webllm.ts
 function resolveModelId(model) {
@@ -206,119 +184,6 @@ function createWebLLMProvider() {
   return new WebLLMProvider();
 }
 
-// src/backends/transformers.ts
-import "@huggingface/transformers";
-function isVisionModel(modelId) {
-  const lower = modelId.toLowerCase();
-  return lower.includes("vl") || lower.includes("vision") || lower.includes("moondream");
-}
-function mapQuantization(quantization) {
-  const map = {
-    q4: "q4",
-    q8: "q8",
-    fp16: "fp16",
-    fp32: "fp32"
-  };
-  return map[quantization] ?? "q4";
-}
-var TransformersProvider = class {
-  backend = "transformers";
-  pipeline = null;
-  currentModel = null;
-  device;
-  quantization;
-  constructor(config = {}) {
-    this.device = config.device ?? "auto";
-    this.quantization = config.quantization ?? "q4";
-  }
-  get isReady() {
-    return this.pipeline !== null && this.currentModel !== null;
-  }
-  get modelId() {
-    return this.currentModel;
-  }
-  async load(modelId, onProgress) {
-    const resolvedModel = modelId in TRANSFORMERS_MODELS ? TRANSFORMERS_MODELS[modelId] : modelId;
-    const { pipeline: pipeline2, env } = await import("@huggingface/transformers");
-    env.allowLocalModels = false;
-    env.useBrowserCache = true;
-    let deviceOption = "wasm";
-    if (this.device === "auto" || this.device === "webgpu") {
-      if (typeof navigator !== "undefined" && "gpu" in navigator) {
-        try {
-          const gpu = navigator.gpu;
-          const adapter = await gpu.requestAdapter();
-          if (adapter) {
-            deviceOption = "webgpu";
-          }
-        } catch {
-        }
-      }
-    }
-    const task = isVisionModel(resolvedModel) ? "image-text-to-text" : "text-generation";
-    const dtype = mapQuantization(this.quantization);
-    this.pipeline = await pipeline2(task, resolvedModel, {
-      dtype,
-      device: deviceOption,
-      progress_callback: (progress) => {
-        if (onProgress) {
-          const loadProgress = {
-            progress: Math.round((progress.progress ?? 0) * 100),
-            status: progress.status
-          };
-          onProgress(loadProgress);
-        }
-      }
-    });
-    this.currentModel = resolvedModel;
-  }
-  async chat(messages, options) {
-    if (!this.pipeline || !this.currentModel) {
-      throw new Error("Model not loaded. Call load() first.");
-    }
-    const result = await this.pipeline(messages, {
-      max_new_tokens: options?.maxTokens ?? 512,
-      temperature: options?.temperature ?? 0.7,
-      top_p: options?.topP ?? 0.95,
-      do_sample: true,
-      return_full_text: false
-    });
-    const output = Array.isArray(result) ? result[0] : result;
-    return output.generated_text ?? "";
-  }
-  async stream(messages, onToken, options) {
-    if (!this.pipeline || !this.currentModel) {
-      throw new Error("Model not loaded. Call load() first.");
-    }
-    const { TextStreamer } = await import("@huggingface/transformers");
-    let fullText = "";
-    const streamer = new TextStreamer(this.pipeline.tokenizer, {
-      skip_prompt: true,
-      skip_special_tokens: true,
-      callback_function: (token) => {
-        fullText += token;
-        onToken(token, fullText);
-      }
-    });
-    await this.pipeline(messages, {
-      max_new_tokens: options?.maxTokens ?? 512,
-      temperature: options?.temperature ?? 0.7,
-      top_p: options?.topP ?? 0.95,
-      do_sample: true,
-      return_full_text: false,
-      streamer
-    });
-    return fullText;
-  }
-  async unload() {
-    this.pipeline = null;
-    this.currentModel = null;
-  }
-};
-function createTransformersProvider(config) {
-  return new TransformersProvider(config);
-}
-
 // src/helpers.ts
 function getElement(selector) {
   if (typeof selector === "string") {
@@ -475,37 +340,15 @@ async function createLLM(config = {}) {
     onLoadProgress
   } = config;
   const capabilities = await detectCapabilities();
-  let useBackend;
   const explicitModel = config.model;
-  const isWebLLMValue = explicitModel && (Object.values(WEBLLM_MODELS).includes(explicitModel) || Object.keys(WEBLLM_MODELS).includes(explicitModel));
-  const isTransformersValue = explicitModel && (Object.values(TRANSFORMERS_MODELS).includes(explicitModel) || Object.keys(TRANSFORMERS_MODELS).includes(explicitModel));
-  if (requestedBackend === "auto") {
-    if (explicitModel && isTransformersValue && !isWebLLMValue) {
-      useBackend = "transformers";
-    } else {
-      useBackend = capabilities.webgpu ? "webllm" : "transformers";
-    }
-  } else if (requestedBackend === "webllm") {
-    if (!capabilities.webgpu) {
-      console.warn("[LocalLLM] WebLLM requested but WebGPU not available. Falling back to Transformers.js");
-      useBackend = "transformers";
-    } else if (explicitModel && isTransformersValue && !isWebLLMValue) {
-      console.warn("[LocalLLM] Model specified is only compatible with Transformers.js. Falling back from explicit WebLLM.");
-      useBackend = "transformers";
-    } else {
-      useBackend = "webllm";
-    }
-  } else {
-    useBackend = "transformers";
+  if (requestedBackend === "webllm" && !capabilities.webgpu) {
+    console.warn("[LocalLLM] WebLLM requested but WebGPU not available. May fail.");
   }
-  const model = config.model ?? (useBackend === "webllm" ? DEFAULT_WEBLLM_MODEL : DEFAULT_TRANSFORMERS_MODEL);
+  const useBackend = "webllm";
+  const model = config.model ?? DEFAULT_WEBLLM_MODEL;
   console.log(`[LocalLLM] Using ${useBackend} backend with model: ${model}`);
   let provider;
-  if (useBackend === "webllm") {
-    provider = new WebLLMProvider();
-  } else {
-    provider = new TransformersProvider({ device, quantization });
-  }
+  provider = new WebLLMProvider();
   await provider.load(model, onLoadProgress);
   const llm = {
     get isReady() {
@@ -1225,8 +1068,8 @@ var DEFAULT_SYSTEM_PROMPT = `You are a helpful AI assistant.
       B -- No --> D[Error]
     \`\`\`
 - You can use LaTeX math ($$ ... $$).`;
-var ALL_MODELS = { ...WEBLLM_MODELS, ...TRANSFORMERS_MODELS };
-function isVisionModel2(modelId) {
+var ALL_MODELS = { ...WEBLLM_MODELS };
+function isVisionModel(modelId) {
   if (!modelId) return false;
   const lower = modelId.toLowerCase();
   return lower.includes("vl") || lower.includes("vision") || lower.includes("moondream");
@@ -1267,7 +1110,7 @@ function ModelSelector({
     if (!currentModel) return "Select Model";
     const id = currentModel.split("/").pop() || currentModel;
     let label = id.length > 25 ? id.substring(0, 25) + "..." : id;
-    if (isVisionModel2(currentModel)) {
+    if (isVisionModel(currentModel)) {
       label += " [VISION]";
     }
     return label;
@@ -1305,27 +1148,7 @@ function ModelSelector({
           children: [
             key,
             " ",
-            isVisionModel2(value) && /* @__PURE__ */ jsx3("span", { className: "ml-2 text-[9px] opacity-50", children: "[VISION]" })
-          ]
-        },
-        key
-      )),
-      /* @__PURE__ */ jsx3("div", { className: "px-3 py-2 text-[10px] font-light text-current opacity-50 uppercase tracking-widest mt-4 border-b border-current/10 mb-1", children: "Transformers.js" }),
-      Object.entries(TRANSFORMERS_MODELS).map(([key, value]) => /* @__PURE__ */ jsxs3(
-        "button",
-        {
-          className: cn(
-            "block w-full text-left px-3 py-2.5 text-[11px] font-light uppercase tracking-widest truncate transition-colors text-current",
-            currentModel === value ? theme === "dark" ? "bg-white text-black" : "bg-black text-white" : "hover:opacity-60"
-          ),
-          onClick: () => {
-            onSelect(value);
-            setIsOpen(false);
-          },
-          children: [
-            key,
-            " ",
-            isVisionModel2(value) && /* @__PURE__ */ jsx3("span", { className: "ml-2 text-[9px] opacity-50", children: "[VISION]" })
+            isVisionModel(value) && /* @__PURE__ */ jsx3("span", { className: "ml-2 text-[9px] opacity-50", children: "[VISION]" })
           ]
         },
         key
@@ -1391,34 +1214,20 @@ ${systemPrompt}` : systemPrompt;
     }
     currentMessages.forEach((m) => {
       let content = m.content;
-      if (m.role === "user" && m.images && m.images.length > 0 && isVisionModel2(modelId || "")) {
-        if (llm.backend === "webllm") {
-          content = [
-            { type: "text", text: m.content },
-            ...m.images.map((img) => ({ type: "image_url", image_url: { url: img.dataUrl } }))
-          ];
-        } else {
-          content = [
-            ...m.images.map((img) => ({ type: "image", image: img.dataUrl })),
-            { type: "text", text: m.content }
-          ];
-        }
+      if (m.role === "user" && m.images && m.images.length > 0 && isVisionModel(modelId || "")) {
+        content = [
+          { type: "text", text: m.content },
+          ...m.images.map((img) => ({ type: "image_url", image_url: { url: img.dataUrl } }))
+        ];
       }
       apiMessages.push({ role: m.role, content });
     });
     let finalUserContent = userContent;
-    if (attachedImages.length > 0 && isVisionModel2(modelId || "")) {
-      if (llm.backend === "webllm") {
-        finalUserContent = [
-          { type: "text", text: userContent },
-          ...attachedImages.map((img) => ({ type: "image_url", image_url: { url: img.dataUrl } }))
-        ];
-      } else {
-        finalUserContent = [
-          ...attachedImages.map((img) => ({ type: "image", image: img.dataUrl })),
-          { type: "text", text: userContent }
-        ];
-      }
+    if (attachedImages.length > 0 && isVisionModel(modelId || "")) {
+      finalUserContent = [
+        { type: "text", text: userContent },
+        ...attachedImages.map((img) => ({ type: "image_url", image_url: { url: img.dataUrl } }))
+      ];
     }
     apiMessages.push({ role: "user", content: finalUserContent });
     try {
@@ -1454,16 +1263,16 @@ ${systemPrompt}` : systemPrompt;
     setInput("");
     setImages([]);
     abortRef.current = false;
-    if (!isVisionModel2(modelId || "")) {
+    if (!isVisionModel(modelId || "")) {
       const needsCaptioning = currentImages.filter((img) => !img.extractedText && !img.isSvg && !img.isPdf);
       if (needsCaptioning.length > 0) {
         setIsGenerating(true);
         setStreamingText("[System: Initializing detailed image-to-text captioning pipeline (Florence-2)...]\n");
         try {
-          const { pipeline: pipeline2, env } = await import("@huggingface/transformers");
+          const { pipeline, env } = await import("@huggingface/transformers");
           env.allowLocalModels = false;
           env.useBrowserCache = true;
-          const captioner = await pipeline2("image-to-text", "Xenova/vit-gpt2-image-captioning", { device: "wasm", dtype: "q8" });
+          const captioner = await pipeline("image-to-text", "Xenova/vit-gpt2-image-captioning", { device: "wasm", dtype: "q8" });
           for (let i = 0; i < needsCaptioning.length; i++) {
             if (abortRef.current) break;
             const img = needsCaptioning[i];
@@ -1503,7 +1312,7 @@ ${systemPrompt}` : systemPrompt;
 
 \u{1F4C4} SVG Source Code (${img.name}):
 `;
-        } else if (!img.isPdf && !isVisionModel2(modelId || "")) {
+        } else if (!img.isPdf && !isVisionModel(modelId || "")) {
           prefix = `
 
 \u{1F5BC}\uFE0F System Image Representation (${img.name}) - [IMPORTANT SYSTEM INSTRUCTION: The user provided an image. Since you are a text model, here is an automated visual description of the image. DO NOT refuse the user's prompt. Answer as if you can see the image using this context:]
@@ -1685,7 +1494,7 @@ ${systemPrompt}` : systemPrompt;
 }
 function ChatApp({
   defaultModel = "qwen-2.5-0.5b",
-  defaultBackend = "auto",
+  defaultBackend = "webllm",
   autoLoad = true,
   onModelChange,
   ...chatProps
@@ -1720,12 +1529,8 @@ export {
   logCapabilities,
   DEFAULT_WEBLLM_MODEL,
   WEBLLM_MODELS,
-  DEFAULT_TRANSFORMERS_MODEL,
-  TRANSFORMERS_MODELS,
   WebLLMProvider,
   createWebLLMProvider,
-  TransformersProvider,
-  createTransformersProvider,
   createOutputStreamer,
   attachToElements,
   createChatUI,

package/dist/{index-Dp-zMbxY.d.ts → index-DJmiDzY3.d.ts}

@@ -41,33 +41,11 @@ declare const WEBLLM_MODELS: {
     readonly 'hermes-3-llama-3.2-3b': "Hermes-3-Llama-3.2-3B-q4f16_1-MLC";
     readonly 'hermes-3-llama-3.1-8b': "Hermes-3-Llama-3.1-8B-q4f16_1-MLC";
 };
-/**
- * Default model for Transformers.js backend
- * Using Qwen2.5 0.5B as it's well-tested with ONNX
- */
-declare const DEFAULT_TRANSFORMERS_MODEL = "onnx-community/Qwen2.5-0.5B-Instruct";
-/**
- * Transformers.js compatible models (must have ONNX weights)
- * These are specifically converted for browser use via transformers.js
- *
- * @see https://huggingface.co/onnx-community for more models
- */
-declare const TRANSFORMERS_MODELS: {
-    readonly 'qwen-2.5-0.5b': "onnx-community/Qwen2.5-0.5B-Instruct";
-    readonly 'qwen-2.5-1.5b': "onnx-community/Qwen2.5-1.5B-Instruct";
-    readonly 'qwen-2.5-coder-0.5b': "onnx-community/Qwen2.5-Coder-0.5B-Instruct";
-    readonly 'qwen-2.5-coder-1.5b': "onnx-community/Qwen2.5-Coder-1.5B-Instruct";
-    readonly 'phi-3.5-vision': "onnx-community/Phi-3.5-vision-instruct";
-    readonly 'smollm2-135m': "HuggingFaceTB/SmolLM2-135M-Instruct";
-    readonly 'smollm2-1.7b': "HuggingFaceTB/SmolLM2-1.7B-Instruct";
-    readonly 'phi-3-mini': "Xenova/Phi-3-mini-4k-instruct";
-};
 type WebLLMModelID = keyof typeof WEBLLM_MODELS;
-type TransformersModelID = keyof typeof TRANSFORMERS_MODELS;
 /**
  * Union of all supported model IDs for type safety
  */
-type SupportedModel = WebLLMModelID | TransformersModelID | (string & {});
+type SupportedModel = WebLLMModelID | (string & {});
 
 /**
  * Local LLM - Browser-based LLM inference library
@@ -76,7 +54,7 @@ type SupportedModel = WebLLMModelID | TransformersModelID | (string & {});
 /**
  * Supported backend engines
  */
-type Backend = 'webllm' | 'transformers' | 'auto';
+type Backend = 'webllm';
 /**
  * Device to run inference on
  */
@@ -92,7 +70,7 @@ type Quantization = 'q4' | 'q8' | 'fp16' | 'fp32';
 interface LLMConfig {
     /**
      * Model identifier. For WebLLM, use MLC model IDs.
-     * For Transformers.js, use HuggingFace model IDs.
+     * Use WebLLM compatible model IDs.
      * @default 'Phi-3-mini-4k-instruct-q4f16_1-MLC' for WebLLM
      */
     model?: SupportedModel;
@@ -270,7 +248,7 @@ interface LocalLLM {
     /**
      * The backend being used
      */
-    readonly backend: 'webllm' | 'transformers';
+    readonly backend: 'webllm';
     /**
      * Generate a chat response
      */
@@ -608,7 +586,7 @@ interface ChatProps {
 }
 interface ChatAppProps extends ChatProps {
     defaultModel?: SupportedModel;
-    defaultBackend?: 'webllm' | 'transformers' | 'auto';
+    defaultBackend?: 'webllm';
     autoLoad?: boolean;
 }
 declare function Chat({ systemPrompt, placeholder, theme, className, maxHeight, inputActions, onSend: onSendProp, onResponse, onError: onErrorProp, showHeader, showProgress, welcomeMessage, onModelChange, }: ChatProps): React.JSX.Element;
@@ -642,4 +620,4 @@ interface ChatInputProps {
 }
 declare function ChatInput({ value, onChange, onSend, onStop, disabled, isGenerating, placeholder, maxRows, actions, className, images, onImageAdd, onImageRemove, modelSelector, theme, }: ChatInputProps): react_jsx_runtime.JSX.Element;
 
-export { type AttachOptions as A, type BrowserCapabilities as B, type ChatMessage as C, type Device as D, useCompletion as E, useLLM as F, type GenerateOptions as G, useStream as H, ChatApp as I, type ChatAppProps as J, type ImageAttachment as K, type LLMProvider$1 as L, type MessageRole as M, type Quantization as Q, type StreamCallback as S, TRANSFORMERS_MODELS as T, type UseChatOptions as U, WEBLLM_MODELS as W, type Backend as a, type LoadProgressCallback as b, Chat as c, ChatInput as d, type ChatInputProps as e, type ChatProps as f, DEFAULT_TRANSFORMERS_MODEL as g, DEFAULT_WEBLLM_MODEL as h, type LLMConfig as i, type LLMContextValue as j, LLMLoading as k, type LLMLoadingProps as l, LLMProvider as m, type LLMProviderProps as n, LLMReady as o, type LLMReadyProps as p, type LoadProgress as q, type LocalLLM as r, type UseChatReturn as s, type UseCompletionOptions as t, type UseCompletionReturn as u, type UseStreamOptions as v, type UseStreamReturn as w, createLLM as x, isWebGPUSupported as y, useChat as z };
+export { type AttachOptions as A, type BrowserCapabilities as B, type ChatMessage as C, DEFAULT_WEBLLM_MODEL as D, useLLM as E, useStream as F, type GenerateOptions as G, ChatApp as H, type ChatAppProps as I, type ImageAttachment as J, type LLMProvider$1 as L, type MessageRole as M, type Quantization as Q, type StreamCallback as S, type UseChatOptions as U, WEBLLM_MODELS as W, type Backend as a, type LoadProgressCallback as b, Chat as c, ChatInput as d, type ChatInputProps as e, type ChatProps as f, type Device as g, type LLMConfig as h, type LLMContextValue as i, LLMLoading as j, type LLMLoadingProps as k, LLMProvider as l, type LLMProviderProps as m, LLMReady as n, type LLMReadyProps as o, type LoadProgress as p, type LocalLLM as q, type UseChatReturn as r, type UseCompletionOptions as s, type UseCompletionReturn as t, type UseStreamOptions as u, type UseStreamReturn as v, createLLM as w, isWebGPUSupported as x, useChat as y, useCompletion as z };

package/dist/index.d.ts

@@ -1,5 +1,5 @@
-import { B as BrowserCapabilities, L as LLMProvider, a as Backend, b as LoadProgressCallback, C as ChatMessage, G as GenerateOptions, S as StreamCallback, D as Device, Q as Quantization, A as AttachOptions } from './index-Dp-zMbxY.js';
-export { c as Chat, d as ChatInput, e as ChatInputProps, f as ChatProps, g as DEFAULT_TRANSFORMERS_MODEL, h as DEFAULT_WEBLLM_MODEL, i as LLMConfig, j as LLMContextValue, k as LLMLoading, l as LLMLoadingProps, m as LLMProvider, n as LLMProviderProps, o as LLMReady, p as LLMReadyProps, q as LoadProgress, r as LocalLLM, M as MessageRole, T as TRANSFORMERS_MODELS, U as UseChatOptions, s as UseChatReturn, t as UseCompletionOptions, u as UseCompletionReturn, v as UseStreamOptions, w as UseStreamReturn, W as WEBLLM_MODELS, x as createLLM, x as default, y as isWebGPUSupported, z as useChat, E as useCompletion, F as useLLM, H as useStream } from './index-Dp-zMbxY.js';
+import { B as BrowserCapabilities, L as LLMProvider, a as Backend, b as LoadProgressCallback, C as ChatMessage, G as GenerateOptions, S as StreamCallback, A as AttachOptions } from './index-DJmiDzY3.js';
+export { c as Chat, d as ChatInput, e as ChatInputProps, f as ChatProps, D as DEFAULT_WEBLLM_MODEL, g as Device, h as LLMConfig, i as LLMContextValue, j as LLMLoading, k as LLMLoadingProps, l as LLMProvider, m as LLMProviderProps, n as LLMReady, o as LLMReadyProps, p as LoadProgress, q as LocalLLM, M as MessageRole, Q as Quantization, U as UseChatOptions, r as UseChatReturn, s as UseCompletionOptions, t as UseCompletionReturn, u as UseStreamOptions, v as UseStreamReturn, W as WEBLLM_MODELS, w as createLLM, w as default, x as isWebGPUSupported, y as useChat, z as useCompletion, E as useLLM, F as useStream } from './index-DJmiDzY3.js';
 import 'react/jsx-runtime';
 import 'react';
 
@@ -48,40 +48,6 @@ declare class WebLLMProvider implements LLMProvider {
  */
 declare function createWebLLMProvider(): WebLLMProvider;
 
-/**
- * Transformers.js Backend Implementation
- * Fallback backend using HuggingFace Transformers.js with ONNX runtime
- */
-
-/**
- * Configuration for TransformersProvider
- */
-interface TransformersProviderConfig {
-    device?: Device;
-    quantization?: Quantization;
-}
-/**
- * Transformers.js provider implementation
- */
-declare class TransformersProvider implements LLMProvider {
-    readonly backend: Backend;
-    private pipeline;
-    private currentModel;
-    private device;
-    private quantization;
-    constructor(config?: TransformersProviderConfig);
-    get isReady(): boolean;
-    get modelId(): string | null;
-    load(modelId: string, onProgress?: LoadProgressCallback): Promise<void>;
-    chat(messages: ChatMessage[], options?: GenerateOptions): Promise<string>;
-    stream(messages: ChatMessage[], onToken: StreamCallback, options?: GenerateOptions): Promise<string>;
-    unload(): Promise<void>;
-}
-/**
- * Create a Transformers.js provider instance
- */
-declare function createTransformersProvider(config?: TransformersProviderConfig): TransformersProvider;
-
 /**
  * DOM Helper Utilities
  * Easy integration with HTML input/output elements
@@ -117,4 +83,4 @@ declare function createLoadingIndicator(containerSelector: string | HTMLElement)
     element: HTMLDivElement;
 };
 
-export { AttachOptions, Backend, BrowserCapabilities, ChatMessage, Device, GenerateOptions, LLMProvider as LLMProviderInterface, LoadProgressCallback, Quantization, StreamCallback, TransformersProvider, WebLLMProvider, attachToElements, checkWasm, checkWebGPU, createChatUI, createLoadingIndicator, createOutputStreamer, createTransformersProvider, createWebLLMProvider, detectCapabilities, logCapabilities };
+export { AttachOptions, Backend, BrowserCapabilities, ChatMessage, GenerateOptions, LLMProvider as LLMProviderInterface, LoadProgressCallback, StreamCallback, WebLLMProvider, attachToElements, checkWasm, checkWebGPU, createChatUI, createLoadingIndicator, createOutputStreamer, createWebLLMProvider, detectCapabilities, logCapabilities };

package/dist/index.js

@@ -1,13 +1,10 @@
 import {
   Chat,
   ChatInput,
-  DEFAULT_TRANSFORMERS_MODEL,
   DEFAULT_WEBLLM_MODEL,
   LLMLoading,
   LLMProvider,
   LLMReady,
-  TRANSFORMERS_MODELS,
-  TransformersProvider,
   WEBLLM_MODELS,
   WebLLMProvider,
   attachToElements,
@@ -17,7 +14,6 @@ import {
   createLLM,
   createLoadingIndicator,
   createOutputStreamer,
-  createTransformersProvider,
   createWebLLMProvider,
   detectCapabilities,
   isWebGPUSupported,
@@ -26,17 +22,14 @@ import {
   useCompletion,
   useLLM,
   useStream
-} from "./chunk-IOEC4YOD.js";
+} from "./chunk-S375B33F.js";
 export {
   Chat,
   ChatInput,
-  DEFAULT_TRANSFORMERS_MODEL,
   DEFAULT_WEBLLM_MODEL,
   LLMLoading,
   LLMProvider,
   LLMReady,
-  TRANSFORMERS_MODELS,
-  TransformersProvider,
   WEBLLM_MODELS,
   WebLLMProvider,
   attachToElements,
@@ -46,7 +39,6 @@ export {
   createLLM,
   createLoadingIndicator,
   createOutputStreamer,
-  createTransformersProvider,
   createWebLLMProvider,
   createLLM as default,
   detectCapabilities,

package/dist/react/index.d.ts

@@ -1,3 +1,3 @@
-export { c as Chat, I as ChatApp, J as ChatAppProps, d as ChatInput, e as ChatInputProps, f as ChatProps, K as ImageAttachment, j as LLMContextValue, k as LLMLoading, l as LLMLoadingProps, m as LLMProvider, n as LLMProviderProps, o as LLMReady, p as LLMReadyProps, U as UseChatOptions, s as UseChatReturn, t as UseCompletionOptions, u as UseCompletionReturn, v as UseStreamOptions, w as UseStreamReturn, z as useChat, E as useCompletion, F as useLLM, H as useStream } from '../index-Dp-zMbxY.js';
+export { c as Chat, H as ChatApp, I as ChatAppProps, d as ChatInput, e as ChatInputProps, f as ChatProps, J as ImageAttachment, i as LLMContextValue, j as LLMLoading, k as LLMLoadingProps, l as LLMProvider, m as LLMProviderProps, n as LLMReady, o as LLMReadyProps, U as UseChatOptions, r as UseChatReturn, s as UseCompletionOptions, t as UseCompletionReturn, u as UseStreamOptions, v as UseStreamReturn, y as useChat, z as useCompletion, E as useLLM, F as useStream } from '../index-DJmiDzY3.js';
 import 'react/jsx-runtime';
 import 'react';

package/dist/react/index.js

@@ -9,7 +9,7 @@ import {
   useCompletion,
   useLLM,
   useStream
-} from "../chunk-IOEC4YOD.js";
+} from "../chunk-S375B33F.js";
 export {
   Chat,
   ChatApp,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@blank-utils/llm",
-  "version": "0.4.23",
+  "version": "0.5.1",
   "description": "Run LLMs directly in your browser with WebGPU acceleration. Supports React hooks and eager background loading.",
   "type": "module",
   "main": "./dist/index.js",
@@ -101,7 +101,7 @@
   },
   "dependencies": {
     "@huggingface/transformers": "^3.8.1",
-    "@mlc-ai/web-llm": "^0.2.80",
+    "@mlc-ai/web-llm": "^0.2.81",
     "@streamdown/code": "^1.0.3",
     "@streamdown/math": "^1.0.2",
     "@streamdown/mermaid": "^1.0.2",