GitP/A/Log/26/10/04 Sun
Properties
2026-10-04 Sun
- My Notes
DOING Ghost Gardener please Follow up on Front of House vs Back of House
- This morning, I'm following up on A Codex agent submitted Draft September 24 episode from garden and Ableton evidence by codekiln · Pull Request #9 · codekiln/gitpa to publish 2026-09-24 Thu's Making/Music/Log/26/09/24 Thu - GitP Reboot with Novation Launchpad.
- After reading through Draft September 24 episode from garden and Ableton evidence by codekiln · Pull Request #9 · codekiln/gitpa, I realized that I am setting both myself and my agents up for confusion by not articulating a clear design boundary between the purpose of Person/codekiln/GitHub/gitpa and the parts of Person/codekiln/GitHub/logseq-encode-garden that are for items related to the podcast.
- Going forwards, I plan for all GitP/House/Back to occur in Person/codekiln/GitHub/logseq-encode-garden, and GitP/House/Front to be in Person/codekiln/GitHub/gitpa.
- I articulated the vision for this split in Ghost in the Patch - Front vs Back House Responsibilities.
- Ghost Gardener please take a look at all assets related to GitP and make a plan in GitP/A/Log/26/10/04 Sun/Front v Back Split/Plan for how to transform these responsibilities and get started.
- This morning, I'm following up on A Codex agent submitted Draft September 24 episode from garden and Ableton evidence by codekiln · Pull Request #9 · codekiln/gitpa to publish 2026-09-24 Thu's Making/Music/Log/26/09/24 Thu - GitP Reboot with Novation Launchpad.
DOING Ghost Gardener please ensure that remote agents can manage podcast assets while I'm remote coding, and prep usage of dvc
- Following up on GitP/A/Log/26/09/26 Sat - RSS Project
- After yesterday, I realized that 1Password/Environments have a major downside for my workflow: they effectively make it impossible for me to use AI/Agent/Remote coding, as in order to publish or manage assets that are published to Backblaze/B2, they require me to use my Touch ID directly on the Macbook. As a result, I think that I want to take an approach using fnox; the main question is, which one, as there are several possible approaches I could take.
- Specifically, within the
logseq-encode-garden, I'd like my remote coding agents to be able to CRUD assets for that garden, and to make local proxies inassets/.remote. - It's very important that whatever situation we come up with be applicable to all my logseq-based knowledge gardens.
- perhaps using fnox/Daemon with fnox/How To/Set Up a Project with age and SSH Keys could be a practical use case for this, although I'm leaning towards the fnox/Golden Path with a caching setup for the ssh integration with fnox.
- Ghost Gardener please create GitP/A/Log/26/10/04 Sun/Fnox/Plan you can use fnox in ghq locally and update it, too
- Private garden storage PR #1 · unattended private B2 access supplies the setup task and verified encrypted cache.
- Following up on GitP/A/Log/26/09/26 Sat - RSS Project
DOING Ghost Gardener please prototype dvc as a possible replacement for git/lfs for larger assets in my various Knowledge Gardens, including as the primary storage mechanism for larger files in My/Knowledge/Garden/logseq-encode-garden and in powering GitP
- Following up on GitP/A/Log/26/09/26 Sat - RSS Project
- I'm leaning towards using dvc, partly in order to get some familiarity with that technology, and also partly in order to ensure that we have a standard method of taking checksums of files and storing them in git, so as to ensure what's what, what's backed up, etc. Also, I sense that there's going to be some benefit here in the long term in terms of data transformation pipelines related to the podcast (for example,
.wav->.mp3or video equivalents). - I'm picturing that
assets/.remotewould be a proxy area for the remote b2 storage, which actually had selected assets (gitignored). It would be a way to bring items down from and push items up to the bucket namespace associated with each garden. - In the garden, I'd likely end up preferring the remote URL version of binary assets, as it will be more portable over time.
- Any one computer that has cloned the garden may or may not have the full representation of the bucket, but it should be possible to proxy that remote bucket locally so the filepaths can be inspected and worked on locally.
- In particular, I'm interested in eventually migrating recording assets currently in
~/Documents/ableton/GitPinto some b2 bucket, though probably notlogseq-encode-garden, as they are too rough draft. I'd prefer to keep the very rough draft assets "in"logseq-garden's bucket (uncreated). - create GitP/A/Log/26/10/04 Sun/Dvc for Knowledge Gardens/Plan
- I'm leaning towards using dvc, partly in order to get some familiarity with that technology, and also partly in order to ensure that we have a standard method of taking checksums of files and storing them in git, so as to ensure what's what, what's backed up, etc. Also, I sense that there's going to be some benefit here in the long term in terms of data transformation pipelines related to the podcast (for example,
- Garden PR #179 · page-named DVC asset commands and WAV-to-MP3 stages includes verified restore, push, status, and conversion checks.
- Following up on GitP/A/Log/26/09/26 Sat - RSS Project
Ghost Gardener - please come up with a Backblaze/B2 entity description and a convention for how logseq assets should be related to the pages they were uploaded for
- Public episode audio contains this:
The September 24 podcast MP3 is published at
gitpa/episodes/2026-09-24/GitP.26.09.24.mp3. Listeners and RSS use that stable URL.
- I need for us to finally document in the entities system what my conventions are for filesystem assets in logseq both locally and in terms of remote backblaze B2 assets. In general, I want there to be a page in the garden for each uploaded asset, and then the filename of the asset should correspond to the place in the garden. This needs to be based on articulated principles and preferences. Looking at the image assets for Microfreak/UG and the image assets for Launchpad/UG and the associated entity descriptions, and thinking about My/Principle/Simplify/Create Uniform Interfaces and My/Principle/Make Illegal States Unrepresentable/Discussion, but considering that I probably want something like GitP/A/Session/26/09/24-Thu/Asset/Synth/Full/mp3 to be the page that has the reference to the mp3, then the asset would be something like
https://f005.backblazeb2.com/file/logseq-encode-garden/GitP___26___09___24___24-Thu___Asset___Synth___Full.mp3, and then that page would contain - PR #177 · garden asset pages and matching local/B2 paths defines the ownership and storage convention, with a September 24 MP3 asset page.
- Ghost Gardener read https://github.com/codekiln/logseq-encode-garden/pull/177/changes#r4178643489 and update the whole pr
- PR #177 · simplified asset pages and flat filenames updates the garden convention and recording link; Gitpa PR #10 · episode audio player embeds the matching MP3.
- Ghost Gardener read https://github.com/codekiln/logseq-encode-garden/pull/177/changes#r4178643489 and update the whole pr
- Public episode audio contains this:
DOING follow My/Principle/Simplify/Don't Repeat Yourself DRY and My/AI/Rule/Prune useless commandments with respect to every episode related to GitP that's distributed across Person/codekiln/GitHub/gitpa and Person/codekiln/GitHub/logseq-encode-garden
- Right now there are tons of pieces of information that are not needed and/or duplicated in multiple places.
- I want to radically simplify this process. Again, I have no listeners, all options are on the table. It matters more that we get it right than that we have backwards compatibility.
- Maybe episode pages in Person/codekiln/GitHub/gitpa could be simple logseq proxies of session pages like GitP/A/Session/26/09/24-Thu without any need for intermediate files.
- you could open prs in both repos which would make it so the two published episodes are represented in a uniform way in logseq-encode-garden and then sync'd with logseq proxy entity definitions to the other garden, updating the logseq queries appropriately so that they still show up on the "home page" for ghost in the patch alpha.
- Eventually, we should completely rewrite Person/codekiln/GitHub/gitpa so it just is a podcast website static site and RSS that uses the info in logseq-encode-garden as its data source. I've been wanting to make progress on Person/codekiln/GitHub/logseq-gardener as a possible source for a static site builder; maybe that's a good idea. But that's too far out.
- What do you think? What's the most economical, best way to make this less abhorrant to my DRY sensibilities today? Take action and submit prs.
- Garden PR #178 · canonical sessions and B2 asset pages and Gitpa publication PR #12 · publish four sessions from proxies are ready. Merge the garden PR first.
- My Notes