Turn files into a portable, embedding-free .skill knowledgebase. Everything
runs in your browser — no upload, no model, no network. Drop files, download the skill, install it in any
agent that supports skills.
Text formats (.txt .md .html .rst .csv …) and .pdf (parsed locally;
pdf.js loads from a CDN on first PDF). Other binaries (DOCX, RTF) — convert to text first.
Add extensions below to ingest more.
The packer chunks your files, builds a BM25 inverted index, and zips it with a query protocol and two
searchers (search.js / search.py) into a <name>.skill. There is no
embedding model: the agent you install the skill into supplies the semantic layer by expanding each query into
search terms. Install the downloaded skill, then ask the agent questions about your corpus.