note_type	tags	publish
sharing	jupyter_notebook	true

Jupyter Notebook to Markdown (and PDF) directly to Obsidian Vault

little background of why i'm doing this:

I'm working with python script and jupyter notebook a lot, and most of the time, i need to write record and write some report based on what's inside the notebook. i find it very exhausting that i have to manually copy and paste everything from notebook to my obsidian vault, which where i record literally everything.

I take nbconvert ftw. so here i am.

fyi:

• im using windows
• im using python 3.10
• and none of the obsidian plugins for this workflow
• most of my plots in jupyternotebook is .png format

so, this workflow is based on

1. my main folder structure in obsidian fault
2. how nbconvert handles notebook to markdown file, that:
   1. every image in cell output will be exported and located in a new sub-folder regardless of the target directory
   2. the markdown link of the image is not a relative path neither the image name (this is a problem since obsidian recognizes it as a broken link)

folder structure in my obsidian look like this

How i use this

here, i run this on my terminal in the same directory where my jupyter notebook i want to convert located. you can run it wherever you want as long as you provide a correct path for the input notebook

python converter.py <notebook_name.ipynb> <obsidian_folder_target>

after run, it would show something like this

notice that the last argument "8 - Works" is one of my main folder in my vault. after i run that python script, it automatically creates a new folder called JupyterNB (u can name it whtvr you want by changing 1 line in the converten.py)

inside JupyterNB folder, nbconvert would automatically creates a folder with the same name as the input notebook name. inside that folder, there are

1. a sub-folder (name appended with _files) to hold every figure existed in the notebook.
2. converted markdown, and
3. converted pdf

more about the converter.py in section this section

here's the before afters

Jupyter Notebook	Markdown	Pdf

Dependencies

Python 3.10

y'all can go here to install jupyter-notebook and the nbconvert or by run these commands on your terminal

python -m pip install jupyter
python -m pip install notebook
python -m pip install notebook

for brevity, bellow are the main package i use:

• Jinja2==3.1.4
• jupyter==1.0.0
• nbconvert==7.16.4
• notebook==6.4.12

y'all can find (probably) all related python package in py_reqs.txt

Obsidian

nothing, yeah, you heard me.

The Scriptingss (jinja and python)

I tweaked 2 files of jinja template in nbconvert configuration, null.j2 and index.md.j2 and i

Jinja template

null.j2

y'all can find the null.j2 file in below path (mine on windows btw) << wherever y'all python installed >>\share\jupyter\nbconvert\templates\base\null.j2

I add obsidian-note's frontmatters in this template, there are some restriction of what you can pull out from this jinja kinda thing, i've tried several but im not very familiar with jinja so this is the best i can try.

basically i just add these in the beginning of the file, (not at the very beginning, after the commented lines)

---
kernel: {{ nb.metadata.language_info.name }}
notebook: {{ nb.metadata.kernelspec.display_name }}
tags: codings
source_type: jupyter_notebook
reviewed: false
---

so the modified null.j2 would look like this:

for more detail of the format, please go to The Notebook file format — nbformat 5.10 documentation

index.md.j2

<< wherever y'all python installed >>\share\jupyter\nbconvert\templates\markdown\index.md.j2

in index.md.j2, my concern was on how nbconvert handle the path of figures cell's output that appear to be a broken link in obsidian.

nbconvert returns a string of full path, (not the relative path or the filename) of the image and formats it using ![[ ... ]]

as you can see in the code snipet below, the output.metadata is a key-value pair. as i said before, the filepath is a string of so

i tweak it by split the string, and take the last item. note that im using \\ since i'm working on WindowsOS.

...
{% block data_png %}
{% if "filenames" in output.metadata %}
{% set file_name = output.metadata['filenames']['image/png'].split('\\')[-1] %}
![[{{ file_name }}]]
{% else %}
![[ data:image/png;base64,{{ output.data['image/png'] | path2url }}]]
{% endif %}
{% endblock data_png %}
...

so the modified index.md.j2 would look like this:

Python -converter.py

as i meant before, i want my converted notebook be in a separate folder called JupyterNB regardless the target directory that were passed before. It is my preference as i can easily tell that whatever inside that folder, is generated from a jupyter notebook. feel free to explore tho, i don't want to dictate how people should take their notes in obsidian, im here to share :D. anyways, here's the snippet of the code

in general, you only have to change VAULTS_DIR variables in order to directly insert the converted notebook to your vaults, other than that, feel free to modify the code to match your workflow/preference(s). i specify the JupyterNB as folder name in function create_output_directory

I'm planning on

1. explore the jinja and nbconvert format
2. play with jupyter-notebook cell's tags
3. read more about jinja thing 😔