Mixed documents
Code, Markdown, HTML, JSON/CSV, PDF, Office, notebooks, archives — extracted into one dump.
Pulp walks a tree in parallel, pulls text out of mixed documents, and writes one file you can hand a model. Paper pulp. Juice. Disposable reading.
Code, Markdown, HTML, JSON/CSV, PDF, Office, notebooks, archives — extracted into one dump.
Plain text, Markdown, or Claude-style XML. Layout follows -f or the -o extension.
Skips node_modules, target, venvs, secrets, and other noisy trees unless you say otherwise.
pulp ui opens a localhost mill at 127.0.0.1:8747. Nothing is uploaded.
cargo install --git https://github.com/BeeGass/pulp --locked
Requires Rust 1.85+. The GitHub repo must be public (or you need access) for --git install to work.
pulp -o dump.xml .
pulp ui
In-browser mill: /mill. Same privacy story as localhost: packing runs in WebAssembly inside your tab. No server-side upload/packer.
MVP extracts text/code/JSON/HTML; use local pulp for PDF/Office. Design notes: docs/wasm-mill.md.