Jupyter Notebook内核共享交互失败求助：WebSocket错误+变量未定义

目标 #

实现Python代码（CLI/脚本）与Jupyter Notebook UI共享同一kernel，达成实时双向变量交互。

预期用例 #

• 在Notebook UI中定义变量：x = 2
• 从Python脚本/CLI执行print(x)
• 预期输出2，验证内核共享生效

尝试操作 #

1. 以协作模式启动Jupyter Lab：

jupyter lab --no-browser --IdentityProvider.token=eere --ip=0.0.0.0 --port=14484 --collaborative

2. 执行以下Python脚本：

from jupyter_nbmodel_client import NbModelClient, get_jupyter_notebook_websocket_url
from jupyter_kernel_client import KernelClient

SERVER_URL = "http://localhost:14484"
TOKEN = "eere"
NOTEBOOK_PATH = "~/Documents/envs/test/kernel_test.ipynb"
cell_content = "print(x)"

kernel = KernelClient(server_url=SERVER_URL, token=TOKEN)
kernel.start()

notebook = NbModelClient(
    get_jupyter_notebook_websocket_url(server_url=SERVER_URL, token=TOKEN, path=NOTEBOOK_PATH)
)
await notebook.start()
cell_index = notebook.add_code_cell(cell_content)
notebook.execute_cell(cell_index, kernel)

错误输出 #

Websocket client stopped.
Traceback (most recent call last):
  File "~Documents/envs/test/.really_delete/lib/python3.10/site-packages/jupyter_nbmodel_client/client.py", line 86, in _listen_to_websocket
    async for message in websocket:
  File "~Documents/envs/test/.really_delete/lib/python3.10/site-packages/websockets/asyncio/connection.py", line 242, in __aiter__
    yield await self.recv()
  File "~Documents/envs/test/.really_delete/lib/python3.10/site-packages/websockets/asyncio/connection.py", line 322, in recv
    raise self.protocol.close_exc from self.recv_exc
websockets.exceptions.ConnectionClosedError: sent 1002 (protocol error) invalid status code; no close frame received
Document ws://localhost:14484/api/collaboration/room/json:notebook:54aee93a-b50f-4190-820f-c5306b218eb8?sessionId=7781e065-b341-4f85-8a56-f4caa4190de6&token=eere not yet synced.

{'execution_count': 1,
 'outputs': [{'output_type': 'error',
   'ename': 'NameError',
   'evalue': "name 'x' is not defined",
   'traceback': ['\x1b[0;31m---------------------------------------------------------------------------\x1b[0m',
    '\x1b[0;31mNameError\x1b[0m                                 Traceback (most recent call last)',
    'Cell \x1b[0;32mIn[1], line 1\x1b[0m\n\x1b[0;32m----> 1\x1b[0m \x1b[38;5;28mprint\x1b[39m(\x1b[43mx\x1b[49m)\n',
    "\x1b[0;31mNameError\x1b[0m: name 'x' is not defined"]}],
 'status': 'error'}

解决方案 #

1. 修复WebSocket连接问题 #

• 确保jupyter_nbmodel_client和jupyter_kernel_client版本与当前Jupyter Lab版本兼容，优先使用匹配的最新稳定版
• 替换Notebook路径中的~为绝对路径（如/home/username/Documents/envs/test/kernel_test.ipynb），避免路径解析错误导致同步失败
• 增加文档同步等待逻辑，确保Notebook完全同步后再执行代码：

await notebook.start()
# 等待文档同步完成
while not notebook.is_synced():
    await asyncio.sleep(0.1)

2. 确保连接Notebook已有的内核 #

当前代码中KernelClient.start()会创建新内核，而非复用Notebook正在使用的内核。需获取Notebook的内核ID再连接：

# 替换原kernel初始化代码
# 获取Notebook关联的内核信息
kernel_info = await notebook.get_kernel_info()
kernel = KernelClient(server_url=SERVER_URL, token=TOKEN, kernel_id=kernel_info['id'])
kernel.connect()

同时需确认Notebook UI已启动并连接内核，且x = 2已执行完毕。

3. 简化替代方案（无需协作模式） #

若协作模式的WebSocket问题难以排查，可直接使用官方jupyter_client库连接现有内核：

1. 在Notebook中执行%connect_info，获取内核连接信息（包含kernel_id、各端口号等）
2. 在脚本中使用以下代码连接并执行：

from jupyter_client import BlockingKernelClient

# 替换为%connect_info输出的实际连接参数
kernel_config = {
    "shell_port": 12345,
    "iopub_port": 12346,
    "stdin_port": 12347,
    "control_port": 12348,
    "hb_port": 12349,
    "ip": "127.0.0.1",
    "key": "abcdef123456",
    "transport": "tcp",
    "signature_scheme": "hmac-sha256",
    "kernel_name": ""
}

client = BlockingKernelClient()
client.load_connection_info(kernel_config)
client.start_channels()

# 执行代码并捕获输出
msg_id = client.execute("print(x)")
for msg in client.get_iopub_msg():
    if msg['parent_header']['msg_id'] == msg_id:
        if 'content' in msg and 'text' in msg['content']:
            print(msg['content']['text'])
client.stop_channels()

内容的提问来源于stack exchange，提问作者Morpheus