pi-lookup

Web search, page fetch, library docs (Context7), and GitHub repo Q&A (DeepWiki) for Pi Agent. No API keys for search/fetch/docs (DeepWiki is public-repo only).

Install

pi install npm:@xaccefy/pi-lookup

Tools

web_search

Queries engines via the local open-websearch daemon (no API key).

• query (required)
• limit (optional, default 10)
• engines (optional, e.g. duckduckgo, brave, bing)

web_fetch

Fetches a URL as clean text/markdown (GitHub READMEs use a dedicated path).

• url (required, http/https)

SPA / JS-rendered pages: if the daemon only returns a thin HTML shell, web_fetch re-renders with system Chromium (--headless --dump-dom) when available and swaps in the richer text.

| Variable | Purpose | |----------|---------| | PI_CHROMIUM_PATH | Absolute path to chromium/chrome binary | | PI_WEBSEARCH_PORT | Daemon port (default 3210) |

Fallback binaries checked: /usr/bin/chromium, /usr/sbin/chromium, Chrome stable paths, etc. If Chromium is missing, static extraction is returned as-is.

context7

Up-to-date library docs + examples.

• libraryName (required, e.g. react)
• topic (optional, e.g. hooks)
• maxTokens (optional, default 10000)

deepwiki

Natural-language Q&A over a public GitHub repo.

• repo (required, owner/name)
• question (required)

Lifecycle

• session_start: best-effort warm-up of the open-websearch daemon (non-blocking).
• session_shutdown: stops the daemon process started by this extension.

Development

bun test packages/pi-lookup
bun run typecheck