Support for Notebook files (.ipynb)

[DevKlim]

Hello! I am a student studying Data Science and I love working in notebook files.

I originally wanted to modify the plugin but I thought this would be a nice feature, being able to send in your notebook files for context and parsing the response into your notebook. IPYNB files are inherently JSONs files so I thought it could be possible to extract markdown or codeblocks (user-preference) and uploading the content into an LLM with the proper prompt for parsing. If we could deterministically encase the code blocks into something easier to parse and cost way less tokens, that will be amazing! Below is a structure example. Obviously, if we can parse the quotations and extract code/markdown, it will be helpful (reducing token cost).

{
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "BASE_DOWNLOAD_DIR = \"./torchaudio_datasets_download\"\n",
    "print(f\"Datasets will be downloaded to: {BASE_DOWNLOAD_DIR}\")\n",
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Helper Functions"
   ]
  },

If I have time later after finals, I might contribute this idea to the repo. Let me know your thoughts!

[robertpiosik]

This is something I would need to research. Could you send pastebins with examples? Notebook file, AI response and your instructions to reproduce.