Upload 2 files

app (2).py

@@ -0,0 +1,136 @@
+import os
+import time
+import fitz  # PyMuPDF
+import gradio as gr
+from groq import Groq
+from langchain_core.documents import Document
+from langchain_huggingface import HuggingFaceEmbeddings  # ✅ 수정된 임포트
+from langchain_community.vectorstores import FAISS
+from langchain_community.vectorstores.utils import DistanceStrategy
+from langchain_groq import ChatGroq
+from langchain.chains import RetrievalQA
+from langchain.prompts import PromptTemplate
+
+# ✅ 1. GROQ API Key 환경변수에서 불러오기
+client = Groq(api_key=os.environ.get("GROQ_API_KEY"))
+
+# ✅ 2. PDF 파일 로딩 및 텍스트 추출
+all_documents = []
+
+def load_and_extract(file):
+    global all_documents
+    pdf_texts = []
+
+    with fitz.open(file.name) as doc:
+        text = "".join(page.get_text() for page in doc)
+        pdf_texts.append({"filename": file.name, "text": text})
+
+    all_documents = [
+        Document(page_content=doc["text"], metadata={"source": doc["filename"]})
+        for doc in pdf_texts
+    ]
+
+# ✅ 3. 임베딩 모델 설정
+embedding_model = HuggingFaceEmbeddings(
+    model_name="jhgan/ko-sbert-nli",
+    model_kwargs={"device": "cpu"},
+    encode_kwargs={"normalize_embeddings": True}
+)
+
+# ✅ 4. 문서 필터링
+def filter_documents_by_keyword(docs, keyword):
+    keyword_lower = keyword.lower()
+    return [doc for doc in docs if keyword_lower in doc.page_content.lower()]
+
+# ✅ 5. QA 체인 빌더
+def build_qa_chain(filtered_docs):
+    if not filtered_docs:
+        return None
+
+    vectorstore = FAISS.from_documents(
+        documents=filtered_docs,
+        embedding=embedding_model,
+        distance_strategy=DistanceStrategy.COSINE
+    )
+
+    retriever = vectorstore.as_retriever(
+        search_type="mmr",
+        search_kwargs={"k": 5, "lambda_mult": 0.2}
+    )
+
+    llm = ChatGroq(model_name="llama3-8b-8192", temperature=0.1)
+
+    prompt = PromptTemplate(
+        input_variables=["context", "question"],
+        template="""
+당신은 문화 프로그램에 대해 친절하고 정확하게 설명하는 한국어 도우미입니다.
+
+문서 내용:
+{context}
+
+질문: {question}
+
+지침:
+- 반드시 한국어로 답변해주세요
+- 문서에 없으면 "죄송하지만 해당 정보는 찾을 수 없습니다"라고 답변하세요
+"""
+    )
+
+    return RetrievalQA.from_chain_type(
+        llm=llm,
+        chain_type="stuff",
+        retriever=retriever,
+        chain_type_kwargs={"prompt": prompt},
+        return_source_documents=False
+    )
+
+# ✅ 6. Gradio 인터페이스
+chat_history = []
+current_chain = None
+current_keyword = ""
+
+def handle_chat(message, keyword):
+    global current_chain, current_keyword
+
+    if not all_documents:
+        return "", [("❗ PDF 파일을 먼저 업로드해주세요.", "")]
+
+    if not keyword.strip():
+        return "", [("❗ 키워드를 입력해주세요.", "")]
+
+    if keyword != current_keyword:
+        filtered = filter_documents_by_keyword(all_documents, keyword)
+        current_chain = build_qa_chain(filtered)
+        current_keyword = keyword
+
+    if not current_chain:
+        return "", [(f"'{keyword}' 관련 문서를 찾을 수 없습니다.", "")]
+
+    response = current_chain({"query": message})
+    answer = response["result"]
+    chat_history.append((f"🙋‍♂️ {message}", f"🤖 {answer}"))
+    return "", chat_history
+
+def clear_history():
+    global chat_history
+    chat_history = []
+    return chat_history
+
+with gr.Blocks(title="오아시스 챗봇 Musesis") as demo:
+    gr.Markdown("### 📚 오아시스 PDF 기반 문화 Q&A 챗봇 (Musesis)")
+
+    file_upload = gr.File(label="📎 PDF 업로드", file_types=[".pdf"], type="file")
+    chatbot = gr.Chatbot(label="대화", height=400)
+    keyword_input = gr.Textbox(label="🔍 키워드", placeholder="예: 단오축제, 문화학교")
+    question_input = gr.Textbox(label="✉️ 질문", placeholder="질문을 입력하세요", lines=2)
+
+    with gr.Row():
+        submit_btn = gr.Button("질문하기 💬")
+        clear_btn = gr.Button("대화 초기화 🧹")
+
+    file_upload.change(fn=load_and_extract, inputs=file_upload)
+    submit_btn.click(fn=handle_chat, inputs=[question_input, keyword_input], outputs=[question_input, chatbot])
+    question_input.submit(fn=handle_chat, inputs=[question_input, keyword_input], outputs=[question_input, chatbot])
+    clear_btn.click(fn=clear_history, outputs=chatbot)
+
+demo.launch()

requirements (2).txt

@@ -0,0 +1,9 @@
+gradio
+groq
+PyMuPDF
+langchain
+langchain-community
+langchain-groq
+langchain-huggingface
+faiss-cpu
+sentence-transformers