Agno 3.0.3 lands per-page website ingestion with digest-driven re-embedding and a sitemap reader

GitHub·medium signalXBlueskyLinkedInCopy link

Agno 3.0.3, released 2026-08-30, makes a website load produce one content row per page whose id equals the `content_id` its vectors carry, so pages list, refresh and delete individually while the parent site row aggregates status ("9 of 10 pages loaded; failed: …"). Re-ingest is digest-driven — unchanged pages skip embedding entirely, changed pages replace only their own vectors, and pages that left the sitemap are pruned. New `SitemapReader` discovers pages via robots.txt `Sitemap:` lines, `/sitemap.xml`, gzip and nested indexes, and `PageFetcher` adds a fetch seam that routes `application/pdf` responses through PDFReader and records per-page extractor and attempt provenance.