GitP/A/Log/26/10/06 Tue - gitpa proxy follow-ups
Properties
2026-10-06 Tue
- AI Notes by Anthropic/Model/Claude/5/5/Opus
Merged
- codekiln/logseq-encode-garden#178 garden session pages as GitP episode sources: the four session pages and their asset pages are on main, with asset URLs on the public S3 endpoint. The
episode/draft,episode/testandepisode/transcribetask pages are gone. - codekiln/logseq-encode-garden#179 page-named DVC asset tasks and codekiln/logseq-encode-garden#180 cached Fnox B2 uploads: the asset tasks GitP sessions use for recordings.
- codekiln/logseq-encode-garden#190 proxy companion tasks: proxying a page now brings its entity definitions and Logseq/Entity/Proxy/Page/mise/Task/sync along. See Logseq/Entity/Log/26/10/05 Mon - Mise Tasks Accompany Logseq Proxies.
- codekiln/logseq-encode-garden#192 relative graph path in proxy manifest: the sync task had written an absolute home-directory path into gitpa's
.logseq-proxy/manifest.json, which gitpa's secretlint rule rejects.
- codekiln/logseq-encode-garden#178 garden session pages as GitP episode sources: the four session pages and their asset pages are on main, with asset URLs on the public S3 endpoint. The
Ready to merge
- codekiln/gitpa#12 publish four GitP sessions from garden proxies: every review thread is answered and resolved. Merging it publishes to listeners.
- The session page shows a heading, description, audio player, GIF and preset download links. The property block is hidden on the published site and still editable in the desktop app.
- All 14 media URLs point at the public
logseq-encode-gardenbucket and answer anonymous range requests. - The homepage query had two faults: an invalid binding, and Logseq rewriting
(pull ?b [*])to a fixed attribute list without:block/name. The query now pulls?p, andmise run site:query:checkapplies the same rewrite as Logseq 0.10.6, so it fails on the old query. The homepage lists 4 episodes, newest first. - All 21 pages gitpa proxies exist in this garden. Re-syncing them with
logseq:entity:proxy:page:syncchanged no page body, and a second run wrote nothing.
- codekiln/gitpa#12 publish four GitP sessions from garden proxies: every review thread is answered and resolved. Merging it publishes to listeners.
Questions
- Should a property added to a source page after the first copy reach the proxy on re-sync?
- The sync keeps the destination's property lines, so later source properties never arrive.
entity-tasks::on Logseq/Entity/Proxy/Page was missing from gitpa for this reason and was added by hand. - My Notes
- Yes.
- In time, I think the entities system may consider having a SemVer versioning system, and perhaps there may even be a Changelog and something akin to a Database/Schema/Migration system tied to that versioning system so downstream gardens can do the right thing - though I'm picturing that some of the changes will be deterministic updates, and others may require AI's help to adapt the new entity's changes to the local context (I'm anticipating that may be necessary in the future, but it's not something I want to design around right now).
- codekiln/logseq-encode-garden#194 re-sync proxy properties from the source: the source owns properties on re-sync; the destination keeps
tags::, the proxy keys and keys an entity definition lists inentity-proxy-destination-properties::. Any other destination-only property is removed, and the preview lists it first.
- The sync keeps the destination's property lines, so later source properties never arrive.
- Should gitpa retire its own
episode:syncnow that Logseq/Entity/Proxy/Page/mise/Task/sync produces the same page bodies?episode:syncfollows embeds and syncs a session with its assets in one batch. The garden task syncs one page at a time.- My Notes
- codekiln/logseq-encode-garden#194 follow embeds in proxy sync:
--follow-embedssyncs a session with its embedded asset pages in one batch. - codekiln/gitpa#15 retire episode:sync: deletes
episode:syncand re-vendors the garden task; it targets the branch of gitpa#12 and merges after garden#194.
- Should a property added to a source page after the first copy reach the proxy on re-sync?
- AI Notes by Anthropic/Model/Claude/5/5/Opus