Top languages

The card reads the languages of the repositories you own or contribute to and ranks them. By default it weighs them by size — the bytes GitHub attributes to each language — which is what the progress bars show.

?username=octocat&langs_count=6

When you have more languages than langs_count, the rest are drawn as one last row, Other, so each share is a share of all your code rather than of the languages shown.

Required

Option	What it is
username	The GitHub login to read languages for

Layouts

layout changes the shape entirely, and is the option worth trying first.

compact

?username=octocat&layout=compact&langs_count=8

One bar across the top, then the languages in two columns. It fits more languages in less height, which is what makes it the usual pick for a README sidebar.

donut

?username=octocat&layout=donut&langs_count=6

donut-vertical stacks the legend under the ring instead of beside it, and pie fills the ring in.

Weighting

Two languages can rank differently depending on what you mean by “most”. size_weight and count_weight set that:

Weights	What it ranks by
size_weight=1&count_weight=0	Bytes of code — the default
size_weight=0&count_weight=1	How many repositories use the language
size_weight=1&count_weight=1	Both, multiplied together

Jupyter notebooks

GitHub counts a whole .ipynb file as Jupyter Notebook, saved outputs included, so one notebook full of plots can outweigh the language it is written in. Hiding Jupyter Notebook drops those bytes without giving the code back to Python.

?username=octocat&notebooks=code

notebooks=code reads the notebooks of each repository GitHub says has some, and counts their code cells as the kernel’s language: Python, R, Julia and so on. Markdown cells and outputs count for nothing. A cell starting with a %% magic such as %%bash counts as that language. One whose body is text, such as %%latex or %%writefile, counts for nothing, as does one in a language GitHub’s bar leaves out, such as %%sql. A kernel the card does not recognize stays Jupyter Notebook, without its outputs.

The default, notebooks=github, keeps GitHub’s count, because code costs requests on top of the card’s usual one: one per repository to list its notebooks, for at most 10 repositories, and one per notebook, for at most 25 notebooks. A repository it cannot read in full keeps GitHub’s count: one with more notebooks than are left to read, a notebook over 2 MB or one it cannot download, or a notebook GitHub leaves out of its count, such as one under docs/ or marked vendored.

Options

Option	Values	What it does
layout	normal compact donut donut-vertical pie	The shape of the card
langs_count	a number	How many languages to name before Other
hide_langs	language names	Languages to leave out
exclude_repo	owner/name or name	Repositories to leave out
size_weight	a number	Weight given to a language’s size
count_weight	a number	Weight given to its repository count
value_format	percent bytes	What the value beside each language reads
hide_progress	true false	Drop the bars and keep the names
hide_values	true false	Drop the numbers
prog_bar_bg_color	a hex color	The unfilled part of each bar
affiliation	OWNER COLLABORATOR ORGANIZATION_MEMBER	Which repositories count
notebooks	github code	How notebooks count

Plus the common options.