颠覆传统RAG！微软发布GraphRAG革新AI检索！打造你的AI助手：GraphRAG+Chainlit实现跨文档智能检索分析，效果惊人！#graphrag #rag #ai #AI检索 #知识图谱

🔥🔥🔥本篇笔记所对应的YouTube视频 https://youtu.be/_UU4cffEmKI

安装

pip3 install graphrag

mkdir -p ./ragtest/input
curl https://raw.githubusercontent.com/win4r/mytest/main/book.txt > ./ragtest/input/book.txt
python3 -m graphrag.index --init --root ./ragtest
python3 -m graphrag.index --root ./ragtest

python3 -m graphrag.query \
--root ./ragtest \
--method global \
"show me some Prompts about Interpretable Soft Prompts."

python3 -m graphrag.query \
--root ./ragtest \
--method local \
"show me some Prompts about Knowledge Generation."

chainlit

#pip3 install chainlit

import chainlit as cl
import subprocess
import shlex

@cl.on_chat_start
def start():
    cl.user_session.set("history", [])

@cl.on_message
async def main(message: cl.Message):
    history = cl.user_session.get("history")
    # 从 Message 对象中提取文本内容
    query = message.content
    # 构建命令
    cmd = [
        "python3", "-m", "graphrag.query",
        "--root", "./ragtest",
        "--method", "local",
    ]
    # 安全地添加查询到命令中
    cmd.append(shlex.quote(query))
    # 运行命令并捕获输出
    try:
        result = subprocess.run(cmd, capture_output=True, text=True, check=True)
        output = result.stdout
        # 提取 "SUCCESS: Local Search Response:" 之后的内容
        response = output.split("SUCCESS: Local Search Response:", 1)[-1].strip()
        history.append((query, response))
        cl.user_session.set("history", history)
        await cl.Message(content=response).send()
    except subprocess.CalledProcessError as e:
        error_message = f"An error occurred: {e.stderr}"
        await cl.Message(content=error_message).send()

if __name__ == "__main__":
    cl.run()

👉👉👉如有问题请联系我的徽信 stoeng

🔥🔥🔥本项目代码由AI超元域频道制作，观看更多大模型微调视频请访问我的频道⬇

👉👉👉我的哔哩哔哩频道

👉👉👉我的YouTube频道

👉👉👉我的开源项目 https://github.com/win4r/AISuperDomain