Skip to content

Latest commit

 

History

History
268 lines (184 loc) · 9.49 KB

README.md

File metadata and controls

268 lines (184 loc) · 9.49 KB
Raw

What is this project?

This project allows you to download and view the contents of a GitHub repository as a single text file. It also supports uploading local ZIP files containing text files and presenting their contents as a single text file.

Why was this tool created?

This tool was created to make it easy to share context with LLM chatbots like ChatGPT, Claude, and Gemini. By downloading a GitHub repository or uploading a ZIP file as a single text file, you can easily provide the chatbot with the necessary context to understand and discuss the contents of the repository.

Here are a few examples of how this tool can be useful when interacting with chatbots:

  • Share the source code of a project with the chatbot to discuss implementation details or get suggestions for improvements.

  • Provide the chatbot with a repository containing documentation or research papers to facilitate a conversation about the topics covered.

  • Use the downloaded text file to give the chatbot context about a specific topic or domain, enabling more informed and relevant responses.

Somewhat interesting stuff

Here's some stuff that is at least somewhat interesting about the project.

Token counting

We provide a token count for the text file. The interesting thing about this is that the token count is done in-browser via a packaging of the python tiktoken library via WASM.

Counting tokens isn't exactly a trivial task. By doing it on the client we take a load off the server. The tradeoff is that it's a bit more than 4MB to download.

ZIP files

To download a given repo, we just slap /archive/master.zip onto the end of the URL. This is basically the same as clicking the "Download ZIP" button on the GitHub page. See this issue for improvements someone could make to this.

In-memory

We don't write the downloaded files to disk, nor do we extract the ZIP file to disk. Everything is kept in memory for a couple of reasons:

  • No static/media files to worry about. No storage needed. No cleaning up storage needed.
  • Security implications of unzipping files from the internet. We don't have to think about handling relative paths or any other shenanigans that could be in the ZIP file.

A side effect of this is that we do a bit of non-standard handling of the POST/GET of the form. We do the downloading and processing of the repo in the view that you get redirected to after the form is submitted. The weird part is that the downloading and processing might lead to some errors that are best displayed on the form page. So we store the appropriate errors in the session and then redirect back to the form page which reads the errors from the session and sends them to the template for display.

Data URI

As a side effect of doing everything in-memory, we don't have an URL to download the concatenated file from. Instead, we embed the file in the HTML on the results page as a data URI. I was at first concerned about the size of the data URI, but it seems that you can have a lot!

According to MDN:

Firefox version 97 and newer supports data URLs of up to 32MB (before 97 the limit was close to 256MB). Chromium objects to URLs over 512MB, and Webkit (Safari) to URLs over 2048MB.

Since we also have a "copy to clipboard" button, we just read the data URI in Javascript, decode it, and copy it to the clipboard.

Resource management

We do need to worry about downloading excessively large repos or delivering excessively large text files.

  • downloader.repo_utils.download_repo restricts the largest size of ZIP file we download,
  • downloader.file_utils.extract_text_files limits the total number of files we'll process.
  • downloader.file_utils.extract_text_files also limits the total size of text we'll deliver.

Text files only (hopefully)

Since we're constructing data to give to a text-based LLM, we only want text files. The first thought I had was to have a hard-coded list of file extensions that would be considered text, but that's not very flexible and requires maintenance.

Instead, I use the algorithm the zlib library uses. You can see it in downloader.file_utils.is_plain_text_file. The algorithm identifies a file as plain text if it contains at least one byte from an allowed list (including typical text characters like TAB, LF, CR, and the range SPACE to 255) and no bytes from a block list (comprising mainly control characters outside common text usage, such as NUL to 6, and 14 to 31), offering a straightforward method for text detection without necessitating the analysis of the entire file or considering ambiguous byte values.

File encoding

Once we've determined a file is a text file, we check it for encoding declarations like XML's <?xml version="1.0" encoding="UTF-8"?> or Python's # -*- coding: utf-8 -*- and use whatever we find there. If we don't find anything we just assume "UTF-8". See more here: downloader.file_utils.detect_internal_encoding.

Configurations

  • The maximum size of repository that will be downloaded is in settings.MAX_REPO_SIZE.
  • The maximum size of ZIP file that can be uploaded is in settings.MAX_REPO_SIZE.
  • We'll stop after processing settings.MAX_FILE_COUNT files from the repo or ZIP file.
  • We'll only deliver up to settings.MAX_TEXT_SIZE of text.

How to Contribute

Thank you for your interest in contributing! Here are the steps to get started with development:

Step 1: Set Up Your Development Environment

Prerequisites

  • Python 3.12 or later
  • Node.js 20 or later

Fork and Clone the Repository

  1. Fork the Repository: Begin by forking the project to your GitHub account.
  2. Clone Your Fork: Clone your fork to your local environment: 
    git clone https://github.com/your-username/gh_repo_download.git

Install Dependencies

For managing Python dependencies, you can choose between uv and pip-tools. uv is much faster for installing dependencies, but you may be more comfortable with pip-tools. uv installation instructions here.

Use a virtualenv!

Select the one that best fits your workflow:

  • Using uv:

    • Install uv if you haven't already.
    • Compile and sync dependencies: 
      uv pip compile requirements.in dev-requirements.in -o requirements.txt
uv pip sync requirements.txt

  • Using pip-tools:

    • Install pip-tools: 
      python3 -m pip install pip-tools
    • Compile and sync dependencies: 
      pip-compile requirements.in dev-requirements.in -o requirements.txt
pip-sync requirements.txt

For Node.js dependencies:

npm install

Step 2: Run the local server(s)

Django

  1. Run the Server:

     python manage.py runserver

Vite

  1. Run the Server:

     npm run dev

Step 2: Make Your Changes

  1. Create a new branch for your changes:

    git checkout -b feature/your-feature-name

  2. Implement your changes, adhering to the project's code style and guidelines.

  3. Test your changes thoroughly.

  4. Commit and push your changes:

    git commit -am 'Add some feature'
git push origin feature/your-feature-name

Step 4: Submit Your Changes

  1. Open a Pull Request (PR) from your branch to the main repository. Provide a detailed description of your changes and the impact they have.

Development Guidelines

  • Adhere to the coding standards (PEP 8 for Python; black for code formatting).
  • Ensure new features or fixes are accompanied by tests.
  • Update the documentation as necessary to reflect your changes.

Getting Help

If you need assistance or have questions, feel free to open/continue an issue on GitHub.

How to use

Downloading a GitHub repository:

  1. Enter the URL of a GitHub repository in the form.
  2. Click the "Submit" button.
  3. The site will process the repository and generate a downloadable text file.
  4. On the next page, click the "Download file" link to save the text file locally.

Uploading a ZIP file:

  1. Select a ZIP file containing the desired text files using the file input field.
  2. Click the "Submit" button to upload and process the ZIP file.
  3. The site will extract the text files from the ZIP and generate a downloadable text file.
  4. On the next page, click the "Download file" link to save the text file locally.

Limitations and Considerations

  • The maximum size of repositories that can be downloaded is 10.0 MB.
  • The maximum size of ZIP files that can be uploaded is 10.0 MB.
  • Only text files from the repository or ZIP file will be included in the downloaded file.
  • The downloaded file is a concatenation of the text files from the repository or ZIP file.
  • We'll stop after processing 1000 files from the repo or ZIP file.
  • We'll only deliver up to 10.0 MB of text.

FAQ

Q: What types of files are included in the download?

A: Only text files from the repository or ZIP file are included in the downloaded file.

Q: What is the size limit for the repositories that can be downloaded?

A: The maximum size limit for repositories is 10.0 MB. Repositories exceeding this limit cannot be downloaded.

Q: What is the size limit for the ZIP files that can be uploaded?

A: The maximum size limit for ZIP files is 10.0 MB. ZIP files exceeding this limit cannot be uploaded.