v0.5.0 — Sitemap ingestion + 2-tier cache

What's new

dompruner_sitemap — full-site ingestion

Fetch every page in a sitemap.xml and return DOM-pruned Markdown for each, with 90%+ token reduction applied site-wide.

{ "tool": "dompruner_sitemap", "sitemap_url": "https://docs.example.com/sitemap.xml", "query": "authentication" }

Options: filter_urls, max_pages (default 20, max 100), concurrency (default 8), ignore_errors (default true).

2-tier page-fault cache (L1/L2)

• L1 (result cache): keyed by url + query, LRU 256, TTL 5 min — skips the full pipeline on repeated calls
• L2 (fetch cache): keyed by url, LRU 64, TTL 2 min — skips HTTP fetch when the same URL is queried with a different query

Per-URL semaphore prevents redundant concurrent fetches.

---

Migrating from v0.3.x

No breaking changes. Existing configs work without modification.

To pick up new features automatically, remove the pinned version:

Before

{ "command": "npx", "args": ["-y", "dompruner-mcp@0.3.1"] }

After

{ "command": "npx", "args": ["-y", "dompruner-mcp"] }

Or use the Remote HTTP endpoint — no local install, always up to date:

{ "url": "https://dompruner-mcp.vercel.app/api/mcp" }