import os, pdfplumber, gradio as gr
from huggingface_hub import hf_hub_download
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_community.vectorstores import FAISS
from langchain_huggingface import HuggingFaceEmbeddings, HuggingFaceEndpoint
from langchain.chains import RetrievalQA

token = os.getenv("HUGGINGFACEHUB_API_TOKEN")
if not token:
    raise ValueError("❌ Hugging Face Token 누락 (Spaces Secrets에 HUGGINGFACEHUB_API_TOKEN 설정 필요)")

llm = HuggingFaceEndpoint(repo_id="meta-llama/Llama-3.2-3B-Instruct", huggingfacehub_api_token=token, temperature=0.7, task="text-generation")

pdf_files = [
    "(20220324) L2 Switch 운용 매뉴얼_Innovation TF_Ver3.1_OCR.pdf",
    "(20230504) 23년 기술교육 교재 1 (LTE)_가치혁신팀_OCR.pdf",
    "(20230531) 23년 기술교육 교재 2 (5G)_가치혁신팀_OCR.pdf",
    "차단기 종류 및 용도_OCR.pdf"
]

docs = []
for name in pdf_files:
    try:
        pdf_path = hf_hub_download(repo_id="dgmos/ericsson-manuals", filename=name, repo_type="dataset", token=token)
        with pdfplumber.open(pdf_path) as pdf:
            text_pages = [p.extract_text().replace("\n", " ") for p in pdf.pages if p.extract_text()]
        text = "\\n".join(text_pages)
        if text.strip():
            docs.append({"page_content": text})
    except Exception as e:
        print(f"🚨 오류: {name} - {e}")

if not docs:
    raise ValueError("❌ PDF 추출 실패 (docs 비어있음)")

splitter = RecursiveCharacterTextSplitter(chunk_size=800, chunk_overlap=100)
texts = splitter.split_documents(docs)
embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/multi-qa-mpnet-base-dot-v1")
vectorstore = FAISS.from_documents(texts, embeddings)
qa_chain = RetrievalQA.from_chain_type(llm=llm, retriever=vectorstore.as_retriever(search_kwargs={"k":3}), chain_type="stuff")

def chatbot(q):
    try:
        return qa_chain.run(q)
    except Exception as e:
        return f"❌ 오류: {str(e)}"

with gr.Blocks(title="Ericsson RAG 챗봇") as demo:
    gr.Markdown("# ⚙️ Ericsson LTE/5G 매뉴얼 챗봇")
    query = gr.Textbox(label="질문 입력", placeholder="예: LTE 스퓨리어스 원인은?")
    output = gr.Textbox(label="응답", lines=8)
    btn = gr.Button("분석 시작")
    btn.click(chatbot, inputs=query, outputs=output)

if __name__ == "__main__":
    demo.launch(server_name="0.0.0.0", server_port=7860)