Update app.py

app.py

@@ -1,89 +1,90 @@
-import gradio as gr
-from huggingface_hub import InferenceClient
-
-import gradio as gr
-from huggingface_hub import InferenceClient
 import os
-import zipfile  
-import nbformat
-from nbconvert.preprocessors import ExecutePreprocessor
-
-# zip 파일 자동 해제 코드
+import zipfile
+import pandas as pd
+import gradio as gr
+from langchain_community.vectorstores import FAISS
+from langchain_community.embeddings import HuggingFaceEmbeddings
+from langchain.chains import RetrievalQA
+from langchain.prompts import PromptTemplate
+from langchain_community.llms import HuggingFacePipeline
+from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
+import torch
+
+# ====== ZIP 자동 해제 ======
 zip_path = "solo_leveling_faiss_ko.zip"
 extract_dir = "solo_leveling_faiss_ko"
-
 if os.path.exists(zip_path) and not os.path.exists(extract_dir):
     with zipfile.ZipFile(zip_path, 'r') as zip_ref:
         zip_ref.extractall(extract_dir)
     print(f"[INFO] 압축 해제 완료 → {extract_dir}")
 
-
-
-"""
-For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-"""
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
-
-def run_notebook(path):
-    with open(path, encoding='utf-8') as f:
-        nb = nbformat.read(f, as_version=4)
-    ep = ExecutePreprocessor(timeout=600, kernel_name='python3')
-    ep.preprocess(nb, {'metadata': {'path': './'}})  # 현재 경로에서 실행
-    print(f"[INFO] Notebook {path} 실행 완료")
-
-def respond(
-    message,
-    history: list[tuple[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-):
-    messages = [{"role": "system", "content": system_message}]
-
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
-
-    messages.append({"role": "user", "content": message})
-
-    response = ""
-
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
-
-        response += token
-        yield response
-
-
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
-demo = gr.ChatInterface(
-    respond,
-    additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
-    ],
+# ====== 데이터 로드 ======
+df = pd.read_csv("sl_webtoon_full_data_sequential.tsv", sep="\t")
+df['row_id'] = df.index
+df['text'] = df.apply(
+    lambda x: f"[{x['에피소드']}] #{x['row_id']} {x['type']} {x['scene_text']}",
+    axis=1
+)
+texts = df['text'].tolist()
+
+# ====== FAISS 로드 ======
+embedding_model = HuggingFaceEmbeddings(model_name='jhgan/ko-sroberta-multitask')
+vectorstore = FAISS.load_local(extract_dir, embedding_model, allow_dangerous_deserialization=True)
+
+# ====== 모델 로드 (CPU 전용) ======
+model_name = "kakaocorp/kanana-nano-2.1b-instruct"
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.float32).to("cpu")
+llm_pipeline = pipeline("text-generation", model=model, tokenizer=tokenizer, max_new_tokens=200)
+llm = HuggingFacePipeline(pipeline=llm_pipeline)
+
+# ====== QA 체인 ======
+custom_prompt = PromptTemplate(
+    input_variables=["context", "question"],
+    template="다음 문맥을 참고하여 질문에 답하세요.\n\n문맥:\n{context}\n\n질문:\n{question}\n\n답변:"
+)
+qa_chain = RetrievalQA.from_chain_type(
+    llm=llm,
+    retriever=vectorstore.as_retriever(search_kwargs={"k": 5}),
+    chain_type="stuff",
+    return_source_documents=True,
+    chain_type_kwargs={"prompt": custom_prompt}
 )
 
+# ====== 선택지 ======
+choices = [
+    "황동석 무리를 모두 처치한다.",
+    "진호를 포함한 황동석 무리를 모두 처치한다.",
+    "전부 기절 시키고 살려둔다.",
+    "시스템을 거부하고 그냥 도망친다."
+]
+
+# ====== Gradio 함수 ======
+def run_episode(selection):
+    user_choice = choices[int(selection) - 1]
+    result = qa_chain({"query": user_choice})
+    retrieved_context = "\n".join([doc.page_content for doc in result["source_documents"]])
+
+    prompt = f"""
+    당신은 웹툰 '나 혼자만 레벨업'의 성진우입니다.
+    현재 상황:
+    {retrieved_context}
+    사용자 선택: {user_choice}
+    성진우의 말투로 간결하고 자연스러운 ���사를 1~2문장 생성하세요.
+    중복된 내용이나 비슷한 문장은 만들지 마세요.
+    """
+
+    response = llm_pipeline(prompt)[0]["generated_text"]
+    return f"[성진우 응답]\n{response}"
+
+# ====== Gradio UI ======
+demo = gr.Interface(
+    fn=run_episode,
+    inputs=gr.Dropdown(choices=["1", "2", "3", "4"], label="선택 번호", type="value"),
+    outputs="text",
+    title="성진우 선택 시뮬레이션",
+    description="번호를 선택하면 성진우의 응답이 생성됩니다."
+)
 
 if __name__ == "__main__":
-    import r_story_test
     demo.launch()