[2.18.3] — 2026-09-29
Fixed
- Web captures whose URL has a query string ingest again, in a profile or not. Every chunk of a capture carries a
url:<canonical url> tag, and @monoes/memory refused tags containing ?, =, & or %, so any such page (every YouTube video) failed with all chunk stores failed and indexed nothing. url: and parent: tags now follow the same rule as src: (anything but control characters), a URL too long for a tag is left off the chunk rather than failing it, and a refused store now says why in the ingest error. Also: the extra documents a mono-agent capture carries beside readable.md (transcript.md, summary.md) no longer supersede each other under their shared URL, so all of them stay searchable and a repeat ingest is a no-op; and doc cite / doc related with --scope profile:<id> read that profile's store instead of the project's.
npm: monomind@2.18.3, @monoes/monomindcli@2.18.3
