Skip to content

Index & watch commands

webx index <domain> — crawl → local SQLite FTS5 index (sitemap-driven or BFS).

Terminal window
webx index go.dev
webx index docs.example.com --collection eng --concurrency 8
webx index example.com --from-cc latest # bootstrap from Common Crawl — zero host requests
Flag Effect
--from-cc latest Common Crawl bootstrap — zero requests to the host
--stale Skip pages already fresh (re-crawl mode)
--respect-robots Honor robots.txt
--include-paths / --exclude-paths <regex> Scope to URL paths
--concurrency N Parallel fetches (WEBX_CRAWL_CONCURRENCY, default 4)
--collection <name> Named corpus → ~/.webx/index-<name>.db
--gc --older-than 90d --vacuum Garbage-collect stale pages + orphan embeddings

With WEBX_EMBED_MODEL, pages embed per-chunk (heading-path segments) — WEBX_INDEX_EMBED=page restores whole-page vectors.

webx query <q> — index-only search. Offline, private, ~ms:

Terminal window
webx query "error handling"
webx query --semantic "graceful shutdown" # bm25 + chunk-embedding cosine via RRF
webx query --collection eng "auth flow"

--semantic needs WEBX_EMBED_MODEL (Ollama works) and returns best-chunk snippets — see Private index & semantic search.

webx seed — index the curated list of ~25 top docs domains. Slow, polite crawl — run once for a useful default corpus.

webx similar <url> — find pages like this one via term-signature match. Local index by default; --web searches live providers (source URL excluded). --limit controls count.

webx diff <url> — line-diff live page vs indexed version. Sends If-Modified-Since, so unchanged pages cost a cheap 304. --update-index reindexes only on change.

webx watch <url> — monitor a page for content changes via conditional-GET polls against the index; reports +added −removed lines.

Terminal window
webx watch <url> --every 30m
webx watch <url> --once --exec "./notify.sh" # cron-friendly
webx watch <url> --semantic # LLM judges if a diff is meaningful (skips timestamp churn)

--exec hooks receive WEBX_CHANGED_URL, WEBX_ADDED, WEBX_REMOVED.

webx llms <domain> — emit a whole-site llms.txt from the local index, grouped by path section with one-line blurbs. --limit caps pages. Hosted vendors sell this as a feature; here it’s a JOIN.