Introducing Files2Prompt

Last updated
Reid Barber

Files2Prompt lets you drop in files or directories and get a formatted LLM prompt copied to your clipboard.

As context windows in LLMs keep growing (Google announced 1M tokens with Gemini 1.5), we're gaining the ability to put entire codebases and books inside prompts.

Currently, there are some challenges when working with files in LLM chat interfaces like ChatGPT and Claude:

  1. Copying a file's content into a prompt window is easy, but this doesn't scale well with many files.

  2. LLMs perform better when file contents are formatted in a structured way. ChatGPT does well with markdown, while Claude does better with XML.

  3. Many LLMs currently treat files as supporting documents used for retrieval or just allow the files to be accessed by a temporary virtual machine's code interpreter environment.

  4. Most LLMs heavily limit the number of files you can upload.

Files2Prompt aims to solve this by allowing you to drop in files or directories and receive a formatted LLM prompt copied to your clipboard.

Your files never leave your browser since the entire process is done client-side.

Current features

  • Switching between prompt structures (Markdown, XML, or JSON)
  • Adding files nested in directories
  • Automatically unarchiving zip files
  • Parsing text from PDFs
  • Parsing Excel files to comma-separated text
  • Parsing text from images with OCR
  • Reordering dropped files for the prompt
  • Removing individual files or the entire prompt
  • Previewing individual files or the entire prompt
  • Auto-copying when the file list or settings are changed
  • Viewing the token count for individual files and the entire prompt

Upcoming features

  • Support for custom prompt structures
  • Ability to paste text or files directly from the clipboard
  • Support for ignoring files in gitignore
  • Ability to edit the text content of dropped files
  • Downloading the prompt as a file
  • Support for more file types
  • More