Installation

Metanno runs in JupyterLab 3 or 4, with Python 3.10 or later. For a first try, install it in the Python environment that runs JupyterLab:

pip install metanno
python -m metanno.viewer.explorer_kernel --install

The second command registers the metanno-explorer kernel, which reads your databases for the Metanno sidebar. Refresh JupyterLab: the Metanno icon appears in the left sidebar, and you can start with Annotate a table.

In a project

To keep Metanno with the other dependencies of your project, add it to your pyproject.toml or requirements.txt, then install them in a new environment:

[project]
name = "my-annotation-project"
version = "0.1.0"
requires-python = ">=3.10"
dependencies = [
  "metanno>=2.0.0-beta.8",
  "edsnlp>=0.21",  # used to pre-annotate the documents
]

mamba create -n annotation python=3.11 -y
mamba activate annotation
pip install .

metanno>=2.0.0-beta.8
edsnlp>=0.21  # used to pre-annotate the documents

mamba create -n annotation python=3.11 -y
mamba activate annotation
pip install -r requirements.txt

When JupyterLab runs in another environment, register this one as a notebook kernel, then select it in the top-right corner of your notebooks:

pip install ipykernel
python -m ipykernel install --user --name annotation

JupyterLab then needs the sidebar and tabs of Metanno, installed in this environment: run this command in it once, then refresh JupyterLab in your browser.

pret update-jupyter-config --apply

It links the front-end files of Metanno and Pret into your Jupyter data folder (~/.local/share/jupyter/labextensions), which JupyterLab reads whatever its own environment.

Check your installation

Run this cell in a notebook:

from pret import component
from pret_joy import Button


@component
def SanityCheckApp():
    return Button("Metanno is installed ✅", sx={"m": 1})


SanityCheckApp()

A button appears below the cell when JupyterLab and your kernel are ready. Nothing shows up? Check that the kernel of the notebook is the environment where you installed Metanno: import sys; print(sys.executable) prints its path.

Databases and the explorer kernel

The metanno-explorer kernel reads the databases listed in metanno-home/metanno.yaml, as shown in Browse a database. Install it, with the Ibis driver of each database, in the environment that can reach these databases:

pip install metanno 'ibis-framework[duckdb]'
python -m metanno.viewer.explorer_kernel --install

Install each driver through its Ibis extra, eg 'ibis-framework[pyspark]' for Spark, even when the driver itself comes from another package index: the extra also installs the packages Ibis needs to read its results.

Registering records the Python executable that ran the command, so this environment can differ from the ones of JupyterLab and of your notebooks, which do not need the explorer kernel to open notebook data.