@blank-utils/llm 0.4.14 → 0.4.16

package/dist/{chunk-L7XEQBE2.js → chunk-YGT6L4HX.js}

@@ -915,24 +915,27 @@ function ChatInput({
   const textareaRef = useRef2(null);
   const fileInputRef = useRef2(null);
   const [isDragging, setIsDragging] = useState2(false);
-  const captionerRef = useRef2(null);
+  const captionerPromiseRef = useRef2(null);
   useEffect2(() => {
     let mounted = true;
+    if (captionerPromiseRef.current) return;
     const initCaptioner = async () => {
       try {
+        console.log("[ImagePipeline] Initializing Transformers.js background captioner...");
         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"
+          device: "wasm"
         });
-        if (mounted) captionerRef.current = captioner;
+        if (mounted) console.log("[ImagePipeline] Captioner loaded successfully!");
+        return captioner;
       } catch (err) {
         console.warn("Failed to initialize background captioner:", err);
+        return null;
       }
     };
-    initCaptioner();
+    captionerPromiseRef.current = initCaptioner();
     return () => {
       mounted = false;
     };
@@ -1055,17 +1058,27 @@ ${newText}` : newText);
         const id = Math.random().toString(36).substring(7);
         const dataUrl = e.target.result;
         let extractedText;
-        if (captionerRef.current) {
+        if (captionerPromiseRef.current) {
+          console.log(`[ImagePipeline] Generating caption for ${file.name} (Waiting for captioner)...`);
           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;
+            const captioner = await captionerPromiseRef.current;
+            if (captioner) {
+              const out = await captioner(dataUrl);
+              console.log("[ImagePipeline] Raw captioner output:", out);
+              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;
+              }
+              console.log("[ImagePipeline] Extracted caption text:", extractedText);
+            } else {
+              console.log("[ImagePipeline] Captioner initialized to null, skipping caption generation.");
             }
           } catch (err) {
-            console.warn("Background captioning failed for image:", err);
+            console.warn("[ImagePipeline] Background captioning failed for image:", err);
           }
+        } else {
+          console.log("[ImagePipeline] Captioner promise ref is null, skipping caption generation.");
         }
         onImageAdd?.({ id, dataUrl, file, name: file.name, extractedText });
       }

package/dist/index.js

@@ -26,7 +26,7 @@ import {
   useCompletion,
   useLLM,
   useStream
-} from "./chunk-L7XEQBE2.js";
+} from "./chunk-YGT6L4HX.js";
 export {
   Chat,
   ChatInput,

package/dist/react/index.js

@@ -9,7 +9,7 @@ import {
   useCompletion,
   useLLM,
   useStream
-} from "../chunk-L7XEQBE2.js";
+} from "../chunk-YGT6L4HX.js";
 export {
   Chat,
   ChatApp,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@blank-utils/llm",
-  "version": "0.4.14",
+  "version": "0.4.16",
   "description": "Run LLMs directly in your browser with WebGPU acceleration. Supports React hooks and eager background loading.",
   "type": "module",
   "main": "./dist/index.js",