jupyter-nbmodel-client 0.7.0

No project description provided

Jupyter NbModel Client

Client to interact with a Jupyter Notebook model.

To install the library, run the following command.

pip install jupyter_nbmodel_client

[!WARNING] This package requires temporary a dev version of pycrdt. Therefore you will need to install a Rust compiler to install it. Once down, execute pip install maturin[patchelf] and then pip install "pycrdt@git+https://github.com/fcollonval/pycrdt.git@dev".

Usage

1. Ensure you have the needed packages in your environment to run the example here after.

pip install jupyterlab jupyter-collaboration ipykernel matplotlib

2. Start a JupyterLab server, setting a port and a token to be reused by the agent, and create a notebook test.ipynb.

jupyter lab --port 8888 --IdentityProvider.token MY_TOKEN

3. Open a Python REPL and execute the following snippet to add a cell.

from jupyter_nbmodel_client import NbModelClient, get_jupyter_notebook_websocket_url

with NbModelClient(
    get_jupyter_notebook_websocket_url(
        server_url="http://localhost:8888",
        token="MY_TOKEN",
        path="test.ipynb"
    )
) as notebook:
    notebook.add_code_cell("print('hello world')")

Check test.ipynb in JupyterLab.

5. The previous example does not involve kernels. Put that now in the picture, adding a cell and executing within a kernel process.

from jupyter_kernel_client import KernelClient
from jupyter_nbmodel_client import NbModelClient, get_jupyter_notebook_websocket_url

with KernelClient(server_url="http://localhost:8888", token="MY_TOKEN") as kernel:
    with NbModelClient(
        get_jupyter_notebook_websocket_url(
            server_url="http://localhost:8888",
            token="MY_TOKEN",
            path="test.ipynb"
        )
    ) as notebook:
        cell_index = notebook.add_code_cell("print('hello world')")
        results = notebook.execute_cell(cell_index, kernel)

        assert results["status"] == "ok"
        assert len(results["outputs"]) > 0

Check test.ipynb in JupyterLab.

You can go further and create a plot with Matplotlib.

from jupyter_kernel_client import KernelClient
from jupyter_nbmodel_client import NbModelClient, get_jupyter_notebook_websocket_url

CODE = """import matplotlib.pyplot as plt

fig, ax = plt.subplots()

fruits = ['apple', 'blueberry', 'cherry', 'orange']
counts = [40, 100, 30, 55]
bar_labels = ['red', 'blue', '_red', 'orange']
bar_colors = ['tab:red', 'tab:blue', 'tab:red', 'tab:orange']

ax.bar(fruits, counts, label=bar_labels, color=bar_colors)

ax.set_ylabel('fruit supply')
ax.set_title('Fruit supply by kind and color')
ax.legend(title='Fruit color')

plt.show()
"""

with KernelClient(server_url="http://localhost:8888", token="MY_TOKEN") as kernel:
    with NbModelClient(
        get_jupyter_notebook_websocket_url(
            server_url="http://localhost:8888",
            token="MY_TOKEN",
            path="test.ipynb"
        )
    ) as notebook:
        cell_index = notebook.add_code_cell(CODE)
        results = notebook.execute_cell(cell_index, kernel)

        assert results["status"] == "ok"
        assert len(results["outputs"]) > 0

Check test.ipynb in JupyterLab.

[!NOTE]

Instead of using the clients as context manager, you can call the start() and stop() methods.

from jupyter_nbmodel_client import NbModelClient, get_jupyter_notebook_websocket_url

kernel = KernelClient(server_url="http://localhost:8888", token="MY_TOKEN")
kernel.start()

try:
    notebook = NbModelClient(
        get_jupyter_notebook_websocket_url(
            server_url="http://localhost:8888",
            token="MY_TOKEN",
            path="test.ipynb"
        )
    )
    notebook.start()
    try:
        cell_index = notebook.add_code_cell("print('hello world')")
        results = notebook.execute_cell(cell_index, kernel)
    finally:
        notebook.stop()
finally:
    kernel.stop()

Uninstall

To remove the library, run the following.

pip uninstall jupyter_nbmodel_client

Contributing

Data models

The following json schema describe the data model used in cells and notebook metadata to communicate between user clients and the ai agent.

{
  "datalayer": {
    "type": "object",
    "properties": {
      "ai": {
        "type": "object",
        "properties": {
          "prompts": {
            "type": "array",
            "items": {
              "type": "object",
              "properties": {
                "id": {
                  "title": "Prompt unique identifier",
                  "type": "string"
                },
                "prompt": {
                  "title": "User prompt",
                  "type": "string"
                },
                "username": {
                  "title": "Unique identifier of the user making the prompt.",
                  "type": "string"
                },
                "timestamp": {
                  "title": "Number of milliseconds elapsed since the epoch; i.e. January 1st, 1970 at midnight UTC.",
                  "type": "integer"
                }
              },
              "required": ["id", "prompt"]
            }
          },
          "messages": {
            "type": "array",
            "items": {
              "type": "object",
              "properties": {
                "parent_id": {
                  "title": "Prompt unique identifier",
                  "type": "string"
                },
                "message": {
                  "title": "AI reply",
                  "type": "string"
                },
                "type": {
                  "title": "Type message",
                  "enum": [0, 1, 2]
                },
                "timestamp": {
                  "title": "Number of milliseconds elapsed since the epoch; i.e. January 1st, 1970 at midnight UTC.",
                  "type": "integer"
                }
              },
              "required": ["id", "prompt"]
            }
          }
        }
      }
    }
  }
}

Development install

# Clone the repo to your local environment
# Change directory to the jupyter_nbmodel_client directory
# Install package in development mode - will automatically enable
# The server extension.
pip install -e ".[test,lint,typing]"

Running Tests

Install dependencies:

pip install -e ".[test]"

To run the python tests, use:

pytest

Development uninstall

pip uninstall jupyter_nbmodel_client

Packaging the library

See RELEASE