ocrai

frontend/bindings/changeme/quizservice.ts

@@ -183,6 +183,10 @@ export function Login(username: string, password: string): $CancellablePromise<$
     });
 }
 
+export function PerformGeminiOcr(base64Image: string): $CancellablePromise<string> {
+    return $Call.ByID(3593860722, base64Image);
+}
+
 /**
  * Register creates a new registration application
  */

frontend/src/views/QuizView.vue

@@ -209,7 +209,7 @@
                 class="textarea textarea-ghost focus:bg-base-200/30 w-full min-h-[160px] text-sm sm:text-base leading-relaxed resize-none p-2 mt-2 border-0 focus:border-0 focus:outline-none focus:ring-0"
               ></textarea>
               <div class="flex justify-between items-center mt-2 border-t border-base-200 pt-2">
-                <div class="flex gap-2">
+                <div class="flex items-center gap-3">
                   <button
                     type="button"
                     @click="triggerOcrUpload"
@@ -220,6 +220,15 @@
                     <CameraIcon :size="12" />
                     종이 답안 인식 (OCR)
                   </button>
+                  <label class="label cursor-pointer py-0 gap-1.5" title="체크 시 온디바이스 Tesseract 대신 Gemini AI를 이용해 손글씨를 고정밀도로 인식합니다">
+                    <input 
+                      type="checkbox" 
+                      v-model="useAiForOcr" 
+                      :disabled="showFeedback || ocrLoading"
+                      class="checkbox checkbox-xs checkbox-secondary" 
+                    />
+                    <span class="label-text text-[10px] sm:text-xs opacity-70 font-bold">AI 인식 사용</span>
+                  </label>
                   <input 
                     type="file" 
                     ref="ocrFileInput" 
@@ -573,6 +582,7 @@ const ocrProgress = ref(0)
 const ocrStatusText = ref('')
 const ocrResultText = ref<string | null>(null)
 const uploadedOcrImageBase64 = ref('')
+const useAiForOcr = ref(false)
 
 function triggerOcrUpload() {
   if (ocrFileInput.value) {
@@ -580,6 +590,15 @@ function triggerOcrUpload() {
   }
 }
 
+function readFileAsDataURL(file: File): Promise<string> {
+  return new Promise((resolve, reject) => {
+    const reader = new FileReader()
+    reader.onload = (e) => resolve(e.target?.result as string || '')
+    reader.onerror = (err) => reject(err)
+    reader.readAsDataURL(file)
+  })
+}
+
 async function onOcrFileChange(event: Event) {
   const target = event.target as HTMLInputElement
   const file = target.files?.[0]
@@ -587,46 +606,51 @@ async function onOcrFileChange(event: Event) {
 
   ocrLoading.value = true
   ocrProgress.value = 0
-  ocrStatusText.value = '인식 엔진 준비 중...'
+  ocrStatusText.value = '이미지 준비 중...'
   ocrResultText.value = null
   uploadedOcrImageBase64.value = ''
 
-  // Convert file to base64 synchronously
-  const reader = new FileReader()
-  reader.onload = (e) => {
-    uploadedOcrImageBase64.value = e.target?.result as string || ''
-  }
-  reader.readAsDataURL(file)
-
   try {
-    const result = await Tesseract.recognize(
+    const base64Data = await readFileAsDataURL(file)
-      file,
+    uploadedOcrImageBase64.value = base64Data
-      'kor+eng',
+
-      {
+    if (useAiForOcr.value) {
-        logger: (m) => {
+      ocrStatusText.value = 'AI 인식 진행 중 (Gemini)...'
-          if (m && m.status) {
+      ocrProgress.value = 50
-            switch (m.status) {
+      
-              case 'loading tesseract core':
+      const result = await QuizService.PerformGeminiOcr(base64Data)
-                ocrStatusText.value = '인식 엔진 불러오는 중...'
+      ocrProgress.value = 100
-                break
+      ocrResultText.value = result || ''
-              case 'initializing api':
+    } else {
-                ocrStatusText.value = '언어 데이터(한/영) 초기화 중...'
+      ocrStatusText.value = '인식 엔진 로딩 중...'
-                break
+      const result = await Tesseract.recognize(
-              case 'recognizing text':
+        file,
-                ocrStatusText.value = '글자 판독 및 텍스트 변환 중...'
+        'kor+eng',
-                break
+        {
-              default:
+          logger: (m) => {
-                ocrStatusText.value = m.status
+            if (m && m.status) {
-            }
+              switch (m.status) {
-            if (typeof m.progress === 'number') {
+                case 'loading tesseract core':
-              ocrProgress.value = Math.round(m.progress * 100)
+                  ocrStatusText.value = '인식 엔진 불러오는 중...'
+                  break
+                case 'initializing api':
+                  ocrStatusText.value = '언어 데이터(한/영) 초기화 중...'
+                  break
+                case 'recognizing text':
+                  ocrStatusText.value = '글자 판독 및 텍스트 변환 중...'
+                  break
+                default:
+                  ocrStatusText.value = m.status
+              }
+              if (typeof m.progress === 'number') {
+                ocrProgress.value = Math.round(m.progress * 100)
+              }
             }
           }
         }
-      }
+      )
-    )
+      ocrResultText.value = result.data.text || ''
-
+    }
-    ocrResultText.value = result.data.text || ''
   } catch (err: any) {
     console.error('OCR 실패:', err)
     uploadedOcrImageBase64.value = ''

quiz_service.go

@@ -768,6 +768,40 @@ func (s *QuizService) AskGemini(source string, id int, prompt string, base64Imag
 	return "started", nil
 }
 
+func (s *QuizService) PerformGeminiOcr(base64Image string) (string, error) {
+	if base64Image == "" {
+		return "", fmt.Errorf("Empty image data")
+	}
+
+	mimeType, decodedData, err := parseBase64Image(base64Image)
+	if err != nil {
+		return "", fmt.Errorf("Failed to parse base64 image: %v", err)
+	}
+
+	ctx := context.Background()
+	prompt := "이 이미지에서 손글씨 또는 인쇄된 글자(한글/영어 등)를 오직 텍스트로만 정확하게 추출해 주세요. 질문에 대한 해설이나 부가적인 설명(예: '추출한 텍스트:', 마크다운 장식 등)은 일절 제외하고, 이미지에 써져 있는 내용 그대로만 텍스트로 반환해야 합니다."
+
+	parts := []genai.Part{
+		genai.Text(prompt),
+		genai.Blob{
+			MIMEType: mimeType,
+			Data:     decodedData,
+		},
+	}
+
+	resp, err := s.executeGeminiWithRetry(ctx, parts)
+	if err != nil {
+		return "", fmt.Errorf("Gemini API call failed: %v", err)
+	}
+
+	if len(resp.Candidates) == 0 || len(resp.Candidates[0].Content.Parts) == 0 {
+		return "", fmt.Errorf("No text detected by Gemini")
+	}
+
+	resultText := fmt.Sprintf("%v", resp.Candidates[0].Content.Parts[0])
+	return strings.TrimSpace(resultText), nil
+}
+
 func (s *QuizService) attachQuestionAssets(parts []genai.Part, source string, id int) []genai.Part {
 	var questionText, choice1, choice2, choice3, choice4, choice5, passage, explanation, imageURL string
 	queryQ := `