Update app.py and requirements.txt with OCR support

app.py

@@ -1,30 +1,27 @@
 
 import os
-from langchain_community.document_loaders import UnstructuredPDFLoader, CSVLoader  # deprecated 해결
+from langchain_community.document_loaders import UnstructuredPDFLoader, CSVLoader
 from langchain.text_splitter import RecursiveCharacterTextSplitter
-from langchain_community.embeddings import HuggingFaceEmbeddings  # deprecated 해결
-from langchain_community.vectorstores import FAISS  # deprecated 해결
-from langchain_huggingface import HuggingFaceEndpoint  # deprecated 해결 (HuggingFaceHub 대체)
+from langchain_community.embeddings import HuggingFaceEmbeddings
+from langchain_community.vectorstores import FAISS
+from langchain_huggingface import HuggingFaceEndpoint
 from langchain.chains import RetrievalQA
 import gradio as gr
 from PIL import Image
 import pytesseract
 
-# 1. 환경 변수 설정 (HF Space Secrets에서 토큰 로드)
 if "HUGGINGFACEHUB_API_TOKEN" not in os.environ:
     raise ValueError("HUGGINGFACEHUB_API_TOKEN 환경 변수가 설정되지 않았습니다. HF Space Settings > Secrets에서 추가하세요.")
 os.environ["HUGGINGFACEHUB_API_TOKEN"] = os.getenv("HUGGINGFACEHUB_API_TOKEN")
 
-# 2. 모델 설정 (Llama 3.2-3B-Instruct, 텍스트 태스크)
-repo_id = "meta-llama/Llama-3.2-3B-Instruct"  # Vision 태스크 에러 해결, 무료 티어 호환
+repo_id = "meta-llama/Llama-3.2-3B-Instruct"
 llm = HuggingFaceEndpoint(
     repo_id=repo_id,
     huggingfacehub_api_token=os.getenv("HUGGINGFACEHUB_API_TOKEN"),
     temperature=0.7,
-    task="text-generation"  # 태스크 명시
+    task="text-generation"
 )
 
-# 3. 문서 로드 함수 (PDF/CSV/이미지)
 def load_documents(files):
     docs = []
     for file_path in files:
@@ -36,7 +33,7 @@ def load_documents(files):
             docs.extend(loader.load())
         elif file_path.endswith((".jpg", ".png")):
             img = Image.open(file_path)
-            text = pytesseract.image_to_string(img, lang="kor+eng")  # 한국어+영어 OCR
+            text = pytesseract.image_to_string(img, lang="kor+eng")
             docs.append(text)
         else:
             continue
@@ -46,7 +43,6 @@ def load_documents(files):
     vectorstore = FAISS.from_documents(texts, embeddings)
     return vectorstore
 
-# 4. RAG 체인 생성
 def create_rag_chain(vectorstore):
     qa_chain = RetrievalQA.from_chain_type(
         llm=llm,
@@ -55,7 +51,6 @@ def create_rag_chain(vectorstore):
     )
     return qa_chain
 
-# 5. 챗봇 함수
 def chatbot(query, files):
     if not files:
         return "파일 업로드 필요 (PDF/CSV/이미지/PPTX)."
@@ -67,7 +62,6 @@ def chatbot(query, files):
     except Exception as e:
         return f"오류: {str(e)}."
 
-# 6. Gradio UI
 with gr.Blocks(title="Ericsson 장비 분석 챗봇") as demo:
     gr.Markdown("# 🚀 3G/LTE/5G 장비 불량/불요파 분석 챗봇")
     gr.Markdown("PDF/CSV/이미지 업로드 후 질문: e.g., 'Spurious Emission 통계?'")
@@ -78,4 +72,4 @@ with gr.Blocks(title="Ericsson 장비 분석 챗봇") as demo:
     btn.click(chatbot, inputs=[query, files], outputs=output)
 
 if __name__ == "__main__":
-    demo.launch(server_name="0.0.0.0", server_port=7860)  # HF Space 호환
+    demo.launch(server_name="0.0.0.0", server_port=7860)