relay_url was being destructured as .. and thrown away, so
update_relay_last_event was called with an empty string and
matched no rows. Capture relay_url from the notification and
pass it through so each relay row gets its last_event updated.
NIP-35:
- Accept 64-char SHA256 info hashes (BitTorrent v2) in addition to 40-char SHA1 (v1)
- Parse `size` tag as explicit total size; file-size sum is now a fallback only
- Parse URL hint from third element of `x` tag as blossom_url fallback
- Parse `magnet` tag on ingest and extract trackers from it
- Emit `magnet` tag in published NIP-35 events
NIP-56:
- Handle `e` tags in kind 1984 reports: look up the torrent event's
publisher and penalise them the same as a direct `p` tag report
- BlossomClient: unauthenticated fetch, kind 24242 signed upload
- Parser: capture `url` tag from NIP-35 events as blossom_url
- Reader: spawn background fetch + cache of blob when url tag present
- Writer: include `url` tag in published NIP-35 events when blob uploaded
- Watcher: upload .torrent file to Blossom before publishing, attach URL
- Migration 007: adds blossom_url + torrent_blob columns to torrents
- Route GET /torrent/<info_hash>.torrent serves cached blobs
- Config: blossom.enabled + blossom.server (disabled by default)
insert_torrent now upserts into publishers so torrent authors appear
in the publishers table immediately, making the name/nip05 join work
on the indexed and dashboard pages. Migration 006 backfills existing
torrent pubkeys on fresh DB opens.
Fetch kind 0 metadata events for known publishers in background
batches (150 at a time, hourly). Store name, nip05, picture, about
in new columns (migration 005). UI now shows avatar + display name
instead of raw pubkeys across indexed, dashboard, and publishers views.
Adds the full writer/publisher stack: NIP-35 event signing and relay
delivery, qBittorrent polling with publish-delay queue, lava_torrent
.torrent file parsing, TMDB inline lookup before publish, and
kindexr-cli identity/publish subcommands.
- Migration 003: add wot_level/muted/report_count to publishers;
create curation_items and reports tables
- WoT builder (wot/follows.rs): fetches operator kind-3 contact list
to build depth-1 and depth-2 trust graph; ingests kind-10000 mute
list; runs at startup and refreshes every 24h
- Trust scoring (wot/trust.rs): recomputes trust from wot_level and
report_count after each WoT build
- Reader: subscribes to kind-30004 (curation sets) from configured
pubkeys; subscribes to kind-1984 (reports) from anyone; WoT/block/
mute check gates kind-2003 ingest when wot_only=true; auto-block
publishers that hit report threshold
- Search: curated events (referenced by kind-30004) sort first in
both FTS and full-scan paths
- Publisher CLI: list, info, block, unblock, trust, mute subcommands
- Fix .gitignore: /bin/ not bin/ (was shadowing src/bin/)
- tvsearch/movie/music/book endpoints with ID filters (tvdbid, imdbid,
tmdbid), season/ep, and type-specific default category restrictions
- Title parser (enrich/parser.rs) extracts season, episode, quality,
source from scene-format names; fills DB fields not supplied by tags
- TMDB enricher (enrich/tmdb.rs) runs as background task after insert
when no TMDB ID in event; uses reqwest against TMDB v3 API directly
- Adult content exclusion at ingest via curation.exclude_categories
- Simplified server.rs routing: t= param now part of SearchQuery
- Remove unused tmdb-api alpha crate; use reqwest directly
enki