Not an answer, but too long for a comment.
There does not seem to be as of mid 2022 an implementation.
Copilot uses an underlying tool, OpenAI Codex, which has a model called codex-davinci-002
, which can do code completion.
https://beta.openai.com/docs/models/codex-series-private-beta
(it's worth noting that OpenAI also have a model confusingly called text-davinci-002
which converts natural language to code, which is not what we want). It's free but has a beta waiting list. There is a jupyterlab attempt at this called jupyterlab-codex
. It may have worked at some point given the search results, eg. https://medium.com/codex/jupyter-notebook-can-do-that-b3185058fe76. So a thing to keep an eye out for is "Jupyter + Codex" in addition to "Jupyter + copilot".
Edit There are several recent GPT3-based tools for code completion in Jupyter Lab, but, as far as I have found they are not as good or even slightly eerie as Copilot. Copilot does have the annoyance that the tab key is for its accepting Copilot suggestion and PyCharm's suggestion and for adding spaces. In Jupyter Lab tab gives a tooltip with a select element, and this seem the tools Tabnine or kite cleanly use this, but will not go beyond variable completion. Say the simple case were you have written def convert_centimetres_to_inches(centimetres:float) -> float:
, in Copilot 3-4 correct lines will be suggested with docstrings too, this is not the case in the jupyterlab clones I have tried, which cannot be coerced into suggestion even the 2.6 factor.