@blank-utils/llm 0.4.20 → 0.4.23

package/dist/{chunk-T4VK6BPC.js → chunk-IOEC4YOD.js}

@@ -989,7 +989,8 @@ ${fullText}`;
                       dataUrl,
                       file: imgFile,
                       name: imgFile.name,
-                      extractedText: i === 1 ? newText : void 0
+                      extractedText: i === 1 ? newText : void 0,
+                      isPdf: true
                     });
                   }
                 }
@@ -1019,7 +1020,8 @@ ${newText}` : newText);
             dataUrl,
             file,
             name: file.name,
-            extractedText: text
+            extractedText: text,
+            isSvg: true
           });
         }
       };
@@ -1453,7 +1455,7 @@ ${systemPrompt}` : systemPrompt;
     setImages([]);
     abortRef.current = false;
     if (!isVisionModel2(modelId || "")) {
-      const needsCaptioning = currentImages.filter((img) => !img.extractedText && !img.name.toLowerCase().endsWith(".svg") && !img.name.toLowerCase().endsWith(".pdf"));
+      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");
@@ -1496,15 +1498,19 @@ ${systemPrompt}` : systemPrompt;
     for (const img of currentImages) {
       if (img.extractedText) {
         let prefix = "";
-        if (img.name.toLowerCase().endsWith(".svg")) {
-          prefix = `\u{1F4C4} SVG Source Code (${img.name}):
+        if (img.isSvg) {
+          prefix = `
+
+\u{1F4C4} SVG Source Code (${img.name}):
 `;
-        } else if (!isVisionModel2(modelId || "")) {
-          prefix = `\u{1F5BC}\uFE0F Image Auto-Caption (${img.name}):
+        } else if (!img.isPdf && !isVisionModel2(modelId || "")) {
+          prefix = `
+
+\u{1F5BC}\uFE0F System Image Representation (${img.name}) - [IMPORTANT SYSTEM INSTRUCTION: The user provided an image. Since you are a text model, here is an automated visual description of the image. DO NOT refuse the user's prompt. Answer as if you can see the image using this context:]
 `;
         }
-        if (prefix || img.name.toLowerCase().endsWith(".pdf")) {
-          finalText += (finalText ? "\n\n" : "") + `${prefix}${img.extractedText}`;
+        if (prefix || img.isPdf) {
+          finalText += `${prefix}${img.isPdf ? "\n\n" : ""}${img.extractedText}`;
         }
       }
     }
@@ -1615,11 +1621,11 @@ ${systemPrompt}` : systemPrompt;
             children: [
               { match: "\u{1F4C4} PDF:", index: msg.content.indexOf("\u{1F4C4} PDF:") },
               { match: "\u{1F4C4} SVG Source Code", index: msg.content.indexOf("\u{1F4C4} SVG Source Code") },
-              { match: "\u{1F5BC}\uFE0F Image Auto-Caption", index: msg.content.indexOf("\u{1F5BC}\uFE0F Image Auto-Caption") }
+              { match: "\u{1F5BC}\uFE0F System Image", index: msg.content.indexOf("\u{1F5BC}\uFE0F System Image") }
             ].filter((m) => m.index !== -1).reduce((min, m) => m.index < min ? m.index : min, msg.content.length) !== msg.content.length ? msg.content.substring(0, [
               { match: "\u{1F4C4} PDF:", index: msg.content.indexOf("\u{1F4C4} PDF:") },
               { match: "\u{1F4C4} SVG Source Code", index: msg.content.indexOf("\u{1F4C4} SVG Source Code") },
-              { match: "\u{1F5BC}\uFE0F Image Auto-Caption", index: msg.content.indexOf("\u{1F5BC}\uFE0F Image Auto-Caption") }
+              { match: "\u{1F5BC}\uFE0F System Image", index: msg.content.indexOf("\u{1F5BC}\uFE0F System Image") }
             ].filter((m) => m.index !== -1).reduce((min, m) => m.index < min ? m.index : min, msg.content.length)).trim() : msg.content
           }
         ) })

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

@@ -620,6 +620,8 @@ interface ImageAttachment {
     file: File;
     name: string;
     extractedText?: string;
+    isPdf?: boolean;
+    isSvg?: boolean;
 }
 interface ChatInputProps {
     value: string;

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-DxcdKb-M.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-DxcdKb-M.js';
+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 'react/jsx-runtime';
 import 'react';
 

package/dist/index.js

@@ -26,7 +26,7 @@ import {
   useCompletion,
   useLLM,
   useStream
-} from "./chunk-T4VK6BPC.js";
+} from "./chunk-IOEC4YOD.js";
 export {
   Chat,
   ChatInput,

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-DxcdKb-M.js';
+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';
 import 'react/jsx-runtime';
 import 'react';

package/dist/react/index.js

@@ -9,7 +9,7 @@ import {
   useCompletion,
   useLLM,
   useStream
-} from "../chunk-T4VK6BPC.js";
+} from "../chunk-IOEC4YOD.js";
 export {
   Chat,
   ChatApp,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@blank-utils/llm",
-  "version": "0.4.20",
+  "version": "0.4.23",
   "description": "Run LLMs directly in your browser with WebGPU acceleration. Supports React hooks and eager background loading.",
   "type": "module",
   "main": "./dist/index.js",