pi-universal-view

Pi extension that converts any file to markdown via markit

Package details

← Back
extension

Install pi-universal-view from npm and Pi will load the resources declared by the package manifest.

npm report
$ pi install npm:pi-universal-view
Package
pi-universal-view
Version
0.2.1
Published
Apr 22, 2026
Downloads
351/mo · 156/wk
Author
miclivs
License
MIT
Types
extension
Size
9.4 KB
Dependencies
1 dependency · 2 peers
Pi manifest JSON
{
  "extensions": [
    ".pi/extensions/universal-view"
  ]
}

Security note

Pi packages can execute code and influence agent behavior. Review the source before installing third-party packages.

README

pi-universal-view

Universal file reading for pi. Powered by markit.

Read PDFs, DOCX, XLSX, EPUB, PowerPoint, Jupyter notebooks, audio files, and ZIPs directly in pi. The extension converts them to markdown so the LLM can work with them.

Install

pi install npm:pi-universal-view

How it works

The extension replaces pi's built-in read tool. Binary formats and URLs get converted to markdown via markit-ai. Everything else passes through to the default reader.

read("paper.pdf")                        → markit converts to markdown → LLM sees full text
read("data.xlsx")                        → markit converts to markdown tables → LLM sees rows
read("https://github.com/owner/repo")    → markit fetches README → LLM sees markdown
read("https://github.com/o/r/issues/1")  → markit fetches via API → LLM sees issue body
read("https://example.com")              → markit fetches page → LLM sees markdown
read("index.ts")                         → passes through to built-in read → unchanged

Supported formats

Category Extensions
Documents .pdf .docx .pptx .xlsx .epub
Data .csv .ipynb
Audio .mp3 .wav .ogg .flac .m4a .aac .wma
Archives .zip
Feeds .rss .atom
URLs https://github.com/* https://gist.github.com/* any web page

Text files, source code, images, and config files fall through to pi's default read.

Audio transcription

Audio files include metadata (duration, format, bitrate) by default. With an OpenAI API key, they also get transcribed via gpt-4o-mini-transcribe.

Set your key:

  1. Environment variable - export OPENAI_API_KEY=sk-...
  2. pi auth storage - the extension bridges keys from pi's auth system into the environment automatically

Without a key, you still get metadata.

Markit supports both OpenAI and Anthropic. To configure a different provider or model:

markit init
markit config set llm.provider anthropic
markit config set llm.apiKey sk-...

See markit docs for details.

How it's built

The extension registers a read tool that shadows pi's built-in:

  1. On session start - resolve the OpenAI key from pi's auth into the environment, then call createLlmFunctions() from markit
  2. On each read - check the file extension. Binary format? markit.convertFile(). Otherwise, delegate to pi's built-in reader

~70 lines of code. See markit for the conversion logic.

Credits

License

MIT