app.py

# app.py  ── 2025-06-08 适配 HuggingFace CPU Space
import os, logging, gradio as gr
from langchain_community.vectorstores import Chroma
from langchain_huggingface import HuggingFaceEmbeddings
from langchain.chains import RetrievalQA
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
from langchain.llms import HuggingFacePipeline

logging.basicConfig(level=logging.INFO)

# ========= 1. 载入本地向量库 =========
embedder = HuggingFaceEmbeddings(
    model_name="sentence-transformers/paraphrase-multilingual-mpnet-base-v2"
)
VEC_DIR = "vector_store"
if not (os.path.isdir(VEC_DIR) and os.path.isfile(f"{VEC_DIR}/chroma.sqlite3")):
    raise RuntimeError(f"❌ 未找到完整向量库 {VEC_DIR}，请先执行 build_vector_store.py")

vectordb = Chroma(persist_directory=VEC_DIR, embedding_function=embedder)

# ========= 2. 载入轻量 LLM =========
model_id = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"     # 1.1B CPU 可跑
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    device_map="auto",          # 需要 requirements 里有 accelerate
    torch_dtype="auto",
)
generator = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
    max_new_tokens=256,
    temperature=0.7,
    top_p=0.9,
)
llm = HuggingFacePipeline(pipeline=generator)

# ========= 3. 构建 RAG 问答链 =========
qa_chain = RetrievalQA.from_chain_type(
    llm=llm,
    chain_type="stuff",
    retriever=vectordb.as_retriever(search_kwargs={"k": 3}),
)

# ========= 4. 业务函数 =========
def simple_qa(user_query: str):
    if not user_query.strip():
        return "⚠️ 请输入学习问题，例如：什么是定积分？"
    try:
        return qa_chain.run(user_query)
    except Exception as e:
        logging.error(f"[QA ERROR] {e}")
        return f"⚠️ 问答失败：{e}"

def generate_outline(topic: str):
    if not topic.strip():
        return "⚠️ 请输入章节或主题", ""
    try:
        docs = vectordb.as_retriever(search_kwargs={"k": 3}).get_relevant_documents(topic)
        snippet = "\n---\n".join(d.page_content for d in docs)
        prompt = (
            f"请基于以下资料，为「{topic}」生成结构化学习大纲，格式：\n"
            f"一、章节标题\n  1. 节标题\n    （1）要点…\n\n"
            f"资料：\n{snippet}\n\n大纲："
        )
        outline = llm.invoke(prompt).strip()
        return outline, snippet
    except Exception as e:
        logging.error(f"[OUTLINE ERROR] {e}")
        return "⚠️ 生成失败", ""

def placeholder(*_):
    return "功能待开发…"

# ========= 5. Gradio UI =========
with gr.Blocks(title="智能学习助手") as demo:
    gr.Markdown("# 智能学习助手 v2.0\n💡 大学生专业课 RAG Demo")
    with gr.Tabs():
        with gr.TabItem("智能问答"):
            chatbot = gr.Chatbot(height=350)
            msg = gr.Textbox(placeholder="在此提问…")
            def chat(m, hist): 
                ans = simple_qa(m)
                hist.append((m, ans))
                return "", hist
            msg.submit(chat, [msg, chatbot], [msg, chatbot])

        with gr.TabItem("生成学习大纲"):
            topic = gr.Textbox(label="章节主题", placeholder="高等数学 第六章 定积分")
            outline = gr.Textbox(label="学习大纲", lines=12)
            debug = gr.Textbox(label="调试：检索片段", lines=6)
            gen = gr.Button("生成")
            gen.click(generate_outline, [topic], [outline, debug])

        with gr.TabItem("自动出题"):
            placeholder(label="待开发")

        with gr.TabItem("答案批改"):
            placeholder(label="待开发")

    gr.Markdown("---\nPowered by LangChain • TinyLlama • Chroma")

if __name__ == "__main__":
    demo.launch()