@blank-utils/llm 0.4.12 → 0.4.14

package/dist/{chunk-CUQPPM45.js → chunk-L7XEQBE2.js}

@@ -915,6 +915,28 @@ function ChatInput({
   const textareaRef = useRef2(null);
   const fileInputRef = useRef2(null);
   const [isDragging, setIsDragging] = useState2(false);
+  const captionerRef = useRef2(null);
+  useEffect2(() => {
+    let mounted = true;
+    const initCaptioner = async () => {
+      try {
+        const { pipeline: pipeline2, 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"
+        });
+        if (mounted) captionerRef.current = captioner;
+      } catch (err) {
+        console.warn("Failed to initialize background captioner:", err);
+      }
+    };
+    initCaptioner();
+    return () => {
+      mounted = false;
+    };
+  }, []);
   useEffect2(() => {
     const textarea = textareaRef.current;
     if (!textarea) return;
@@ -1007,12 +1029,45 @@ ${newText}` : newText);
       reader2.readAsArrayBuffer(file);
       return;
     }
+    if (file.type === "image/svg+xml" || file.name.toLowerCase().endsWith(".svg")) {
+      const reader2 = new FileReader();
+      reader2.onload = (e) => {
+        const text = e.target?.result;
+        if (text) {
+          const id = Math.random().toString(36).substring(7);
+          const dataUrl = `data:image/svg+xml;utf8,${encodeURIComponent(text)}`;
+          onImageAdd?.({
+            id,
+            dataUrl,
+            file,
+            name: file.name,
+            extractedText: text
+          });
+        }
+      };
+      reader2.readAsText(file);
+      return;
+    }
     if (!file.type.startsWith("image/")) return;
     const reader = new FileReader();
-    reader.onload = (e) => {
+    reader.onload = async (e) => {
       if (e.target?.result && typeof e.target.result === "string") {
         const id = Math.random().toString(36).substring(7);
-        onImageAdd?.({ id, dataUrl: e.target.result, file, name: file.name });
+        const dataUrl = e.target.result;
+        let extractedText;
+        if (captionerRef.current) {
+          try {
+            const out = await captionerRef.current(dataUrl);
+            if (Array.isArray(out) && out[0] && out[0].generated_text) {
+              extractedText = out[0].generated_text;
+            } else if (!Array.isArray(out) && out.generated_text) {
+              extractedText = out.generated_text;
+            }
+          } catch (err) {
+            console.warn("Background captioning failed for image:", err);
+          }
+        }
+        onImageAdd?.({ id, dataUrl, file, name: file.name, extractedText });
       }
     };
     reader.readAsDataURL(file);
@@ -1432,7 +1487,17 @@ ${systemPrompt}` : systemPrompt;
     let finalText = text;
     for (const img of currentImages) {
       if (img.extractedText) {
-        finalText += (finalText ? "\n\n" : "") + img.extractedText;
+        let prefix = "";
+        if (img.name.toLowerCase().endsWith(".svg")) {
+          prefix = `\u{1F4C4} SVG Source Code (${img.name}):
+`;
+        } else if (!isVisionModel2(modelId || "")) {
+          prefix = `\u{1F5BC}\uFE0F Image Auto-Caption (${img.name}):
+`;
+        }
+        if (prefix || img.name.toLowerCase().endsWith(".pdf")) {
+          finalText += (finalText ? "\n\n" : "") + `${prefix}${img.extractedText}`;
+        }
       }
     }
     setInput("");

package/dist/index.js

@@ -26,7 +26,7 @@ import {
   useCompletion,
   useLLM,
   useStream
-} from "./chunk-CUQPPM45.js";
+} from "./chunk-L7XEQBE2.js";
 export {
   Chat,
   ChatInput,

package/dist/react/index.js

@@ -9,7 +9,7 @@ import {
   useCompletion,
   useLLM,
   useStream
-} from "../chunk-CUQPPM45.js";
+} from "../chunk-L7XEQBE2.js";
 export {
   Chat,
   ChatApp,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@blank-utils/llm",
-  "version": "0.4.12",
+  "version": "0.4.14",
   "description": "Run LLMs directly in your browser with WebGPU acceleration. Supports React hooks and eager background loading.",
   "type": "module",
   "main": "./dist/index.js",