ByteCode.NewsSubmit
Use quotes for an exact word.

Subscribe

RSSPodcast
Log in

Chat Logs

  1. jreicherI'm a bit surprised to hear you say your name with a soft `g' though. Wouldn't have thought that would be it.
  2. jreicherI'm sure I'm missing something obvious, because I haven't had time to learn about PUDL properly and I might have missed mention here, but why is there both a PUDL and non-PUDL version of BCN?
  3. dreamrealjreicher: https://pudl.bytecode.news/posts/2026/10/updates-for-bytecode-news
  4. dreamrealthere's also a nextjs UI and two basic UIs
  5. * skillbot quit (Remote host closed the connection)
  6. * skillnews quit (Remote host closed the connection)
  7. * skillbot joined #primate
  8. * skillnews joined #primate
  9. bluedreamreal: good narration there
  10. blueone the thing I appreciate about this narration is that it's not a 1:1 rendition of the text; it's an understanding that the spoken word needs to carry a bit more familiarity and flair than a 'dry' written text
  11. blueone thing*
  12. blueany way, do more of these, you have a wonderful voice dreamreal!
  13. blueside note: I've been annoying Z (my wife) to no end, prononuced lever 'LEEver' (which is the accepted UK way); she's sure it's an absolute abomination, and I agree; but I also say `DEBri` instead of `debRI`, for debris, and that's a uk thing as well
  14. jreicherThe oral version elaborates on the written version. Wonder where that idea comes from...
  15. blueprobably by jreicher.
  16. * lioloc joined #primate
  17. liolocblue: hey! is nub a package manager you made?
  18. liolocoh nvm found it, never heard of it
  19. bluelioloc: hi. no. but also, we're moving to flog, which *is* a package manager I'm making
  20. bluehttps://repopack.com/project/flog
  21. bluehttps://flog.run/
  22. nevetflog — predictable JavaScript package management
  23. bluethe installer won't work yet because repopack doesn't have release primitives yet, but primate 0.43 will move to flog
  24. blueand anyway repopack will have release primitives soon
  25. liolocwhat's the advantage of flog over nub? Seems like nub was kind of similar to your philosophy of not being bound by one specific library
  26. bluenub didn't work as expected, published prelease versions, using otp flows for publishing instead of a web flow, and was generally pretty broken...
  27. liolocah i see
  28. blueand flog adds things I've been missing in a package manager for nearly forever
  29. bluelike this https://flog.run/docs/release-sets
  30. bluesomehow package manager authors think it beneath them to include this feature :P
  31. bluebut this has caused me to publish broken versions in the past, more than once
  32. bluebecause the workspace protocol has no concept of versioning, that's the problem described there
  33. liolocah i see
  34. bluethe workspace protocol is a "trust-me-bro" situation. so flog checks for you that you're not publishing a package in a way that would break stuff
  35. blueanyway, 0.43 will be delayed a bit while we're moving to flog and finishing a few things flog needs in repopack for that, like releases
  36. bluelioloc: btw, repopack can do deploy task branches
  37. bluecan now*
  38. liolocso if you're testing a feature you can deploy it to a specific url?
  39. blueyes
  40. liolocthat's neat
  41. bluefor example, I can deploy your #151 point
  42. bluehttps://pewtxngij-task-151.repopack.app/
  43. nevetPrimate — The Universal Web Framework
  44. blueit will contain your changes to docs
  45. liolocthat's very cool
  46. bluehttps://pewtxngij-task-151.repopack.app/docs/backend/python#desktop-targets
  47. nevetPrimate — Python
  48. bluethis feature is currently limited to apps which do not use secrets or a fixture. but it will be extended to secrets/fixtures
  49. blueso in the future you'll be able to deploy a full app from a branch, seeded with fixtures and secrets
  50. liolocnice
  51. liolocwhat's an example of a fixture that would be added to the deployment?
  52. bluea database import
  53. liolocah yes neat, so you can use a test database for task branches
  54. liolocsomething like that?
  55. blueyes
  56. lioloccool
  57. bluefixtures are actually going to get their repo per project. so they'll be versioned
  58. blueand then when deploying a preview, you'll be able to provide a sha for the fixtures, or just get master
  59. liolocwait a second, i'm getting some interesting ideas
  60. lioloclike in the case of a cms, you could use a versioned fixture as your database no?
  61. blueyes
  62. liolocwow
  63. liolocand you'd be able to rollback to a specific sha
  64. lioloc?
  65. blueya of course. task apps are considered throwaways. but fixtures will be usable by normal apps as well, and you'll be able to deploy with any sha
  66. liolocwow
  67. liolocvery cool
  68. blueya
  69. bluereleases, packages and fixtures are a few of the remaining primitives missing from repopack to be able to cover most use cases
  70. blueand anyway I can already do with repopack now, what I couldn't do with fly before, and that is get a preview of my primate task branches
  71. liolocwhat about hosting costs ? are you using aws btw?
  72. blueusing my own vps
  73. blueon contabo
  74. bluecontabo has an api, so the network will grow mostly automatically, once I have installed the ability to do so
  75. liolocah kk
  76. bluecontabo will not back the only resource provider, could be I use aws in the future too, but reselling margins on aws are brutal
  77. bluewill not be the only*
  78. liolocis contabo better than hetzner?
  79. blueit's much easier to resell resources if you go lower, like vps or better dedis
  80. blueit's probably similar to hetzner, I never used hetzner but contabo has been very reliable
  81. bluerepopack is also going to offer an agentbox feature, which is getting a box for your `rp agent` and opencode/claude code/codex to run on
  82. blueconnected to the chat on repopack.com, of course
  83. blueit would be an alternative to having rp agent and the agent run on your computer
  84. bluebetter isolation, and your computer/server doesn't have to be on all the time
  85. liolochmm wouldn't it be better to let users use their own models with rp agent? I'm asking cause I'm wondering if agentboxes would strain your infrastructure
  86. blueya, rp agent is model agnostic. rp agent is just a bridge to connect your repopack.com chat to your agent
  87. blueyour agent controls what models you use
  88. liolocoh I see, but then what is an agentbox?
  89. bluea computer where rp agent runs on, alongside your opencode/claude code/codex
  90. liolocso its like a docker container for your agent?
  91. blueyes
  92. bluemore precisely, every task gets a docker container
  93. blueand is isolated from anything else
  94. blueit also gets the branch's checkout and instructions via an RP Tool
  95. bluewell, that's the theory, anyway
  96. liolocbut how does a person configure the agentbox? they would need to create a file or something? Like with docker containers?
  97. blueit's all done for you in the background, you buy the agentbox from repopack.com. or you buy your own vps independently and install rp agent and your agent on it
  98. blueyou can either. run it on your comp, on your vps, or via an agentbox from repopack.com
  99. liolocso you have 3 choices for rp agent, 1. buy from repopack.com, 2. buy vps and instsall rp agent on it 3. use your local computer?
  100. blueyes
  101. lioloci feel like most people would just use their own computer, unless you're an enterprise you wouldn't do it
  102. liolocwhich makes sense, since enterprises are better paying customers anyway
  103. bluethat's ok, the agentbox isn't the only source of income anyway, resellgin resources is
  104. blueagentbox is just going to be the least hassle. you add your codex api key to repopack.com and you're done
  105. liolocmaybe agentbox is not the best name, agentcontainer would be better imo
  106. blueya, too long though :P
  107. lioloccause everyone knows what a container is for coding purposes
  108. liolocyeah but the marketing spend you would have to do to tell people what a box is will be quite high
  109. bluecould be just `hosted agents` I think
  110. liolocyeah that's good too
  111. lioloceven better
  112. blueyup
  113. bluebbiab
  114. * lioloc quit (Read error: Connection reset by peer)
  115. dreamrealI need to do better music, too. That's very much a trial run
  116. dreamrealpodcast music needs to be flourish and not theme, and that's an actual song of mine
  117. dreamrealand my son said it sounded "very egyptian" :D
  118. blueyou need to do a podcast, with that radiophonic voice
  119. dreamrealoh my voice sucks, and my enunciation is awful
  120. blueI don't know, I like it. you also have a cool accent, slightly southern
  121. bluedreamreal: thoughts on the fixture subrepo?
  122. bluemy only concern is space, if it's used too much for binary stuff, then technically ostree is the superior solution, but I don't know
  123. bluethen again, db dumps are not binary. but db dumps aren't hte only type of fixture people use
  124. dreamrealcould use OCI images for it, they are naturally compressable
  125. blueyou want to use OCI images for *data*?
  126. blueso layered data?
  127. dreamrealI don't WANT anything, but if you're talking data, why not?
  128. blueI'm not against it, I just associatd OCI images with structure/build instructions, not data. but I guess why not
  129. bluethe important part though is that it's easily versionable, that's where git shines
  130. dreamrealeasily managed, easily versioned, easily compressed, easily MOUNTED, cohesive unit...
  131. dreamrealI dunno, maybe that's a dumb idea, I just woke up and I'm trying to come up with new music for the podcast intro, what do I know
  132. dreamrealjust sent you what I was thinking of
  133. dreamrealin discord
  134. bluehow is OCI easily versioned?
  135. * lioloc joined #primate
  136. dreamrealOCI images are innately versioned, although I don't know what you're asking
  137. blueI'll think about it
  138. dreamrealblue: listen on discord, dang it!
  139. bluedreamreal: scrum week, sorry
  140. bluesoon
  141. * phaleth joined #primate
  142. bluehi phaleth
  143. bluephaleth: flog can now fully build primate :)
  144. phalethhi blue
  145. phalethI failed to type backticks https://repopack.com/project/flog/tasks/40?
  146. phalethand I noticed, yeah
  147. bluecool
  148. * phaleth quit (Quit: phaleth)
  149. * phaleth joined #primate
  150. phalethanyway, surprising that flog works this well
  151. blueya, it's amaze sauce
  152. blueand we're just getting started
  153. blue"Clean Primate install improved from approximately 71s to 18s on this host, with byte-identical lockfiles."
  154. blue#40 soon done
  155. blue18s is still too long though
  156. phalethI've tried the trust me bro installer https://repopack.com/project/flog/tasks/42
  157. bluephaleth: yes, this is a known limitation
  158. bluethe flog project needs repopack release artifacts for that
  159. bluehttps://repopack.com/project/repopack/tasks/283
  160. bluethe last comment is by the flog agent
  161. blueand shows the way forward
  162. phalethwall of text
  163. phalethok, I will just put back the artifact url
  164. blueya
  165. blue#40 is solved
  166. phalethtwo jobs at once running
  167. blueya
  168. phalethit got a bit faster, nice
  169. phalethwould it be difficult to publish flog to npm?
  170. phalethI don't think the trust me bro script would work during image build
  171. blueprobably not
  172. blue(about publishing to npm)
  173. phalethok, then how to figure out the url to latest build? https://repopack.com/api/v0/jobs/1/runs/28/artifacts?project=flog
  174. phalethwithout knowing that the number 28 is latest
  175. bluechecking
  176. blueI'm afraid you'll have to wait for #283 to land for that
  177. blueI'll priortise it
  178. phalethok, no hurry
  179. bluelioloc: https://pewtxngij-task-213.repopack.app/blog/primate-043#native-desktop-launchers-and-sidecars
  180. nevetPrimate — Primate 0.43: diamond migrations, deferred navigation and direct SQL
  181. bluephaleth: I seriously hate how those task apps look. I'm going to make short_id electable
  182. bluemaybe `t-primate-213
  183. bluewhere primate is the short_id
  184. bluet-primate-213.repopack.app looks good; and is also a deterministic way to prevent people from using anything that starts with `t-` for normal apps
  185. phalethyeah, but at least they are not scrapeable
  186. phalethlike imagine somebody wanting to share the link publicly and then somebody else guessing the name of their project and trying out all task numbers
  187. blueyes, but almost all task apps will be private
  188. blueyou'll need to explicitly share them with a one-time link, or use --public (with special permissions)
  189. bluethe default mechanism won't be generally accessibl
  190. bluee
  191. phalethone-time link, but what if they wanna navigate the page or just accidentally close browser?
  192. phalethI think it's totally fine the way it is
  193. blueI think exposing task apps to -world- has all kinds of bad implications
  194. blueand those urls look like some chinese spam website
  195. blueanyway, task apps are almost exclusively for developers wanting to preview things
  196. phalethyeah, but the project maintainer can always create a named app where they can deploy whatever
  197. blueyes
  198. bluethat's exactly what I was going to say
  199. blueif you just want a system where you have some people preview features, then create a normal app for it
  200. bluethe task apps are just a way to preview a task without having to download and deploy it all locally, including setting up a db and whatever
  201. phalethok, that means you can somehow hide them behind login, but they are not on the same domain as repopack.com
  202. phalethand traefik just routes trafic straight to containers
  203. blueya, I think that can work
  204. blueanyway they are also shortlived phaleth
  205. bluethey currently expire after 24h, unless you redeploy
  206. blueI'm thinking of reducing that massively
  207. bluemaybe 8h
  208. bluemaybe less
  209. phalethcan I take this one? https://repopack.com/project/primate/tasks/214
  210. phalethok
  211. phalethI'll just do that before clanker takes it
  212. phalethyeah, I cannot take it
  213. * skillbot quit (Remote host closed the connection)
  214. * skillnews quit (Remote host closed the connection)
  215. phalethwell, assign works
  216. * skillbot joined #primate
  217. * skillnews joined #primate
  218. blueyes
  219. bluejuts selfassign
  220. bluephaleth: https://repopack.com/project/flog/tasks/43
  221. blueaccording to clanker:
  222. blue- Resolution: 13.33s → 4.50–5.11s
  223. blue- Full warm-cache install: ~18s → 11.06s
  224. blueand he's not done yet, this is incremental
  225. phalethok, I can revise 214 afterwards
  226. phalethit invented the word packuments
  227. bluewhat's that?
  228. phalethdunno, nobody reads the crap it comes up with
  229. bluelol
  230. phalethelectron-to-chromium@1.5.454 was published during review :)
  231. phalethit looked through those files just to find that, crazy
  232. * phaleth quit (Quit: phaleth)
  233. * nevet joined #primate
  234. * blue!~blue@user/blue changed the topic to: https://primate.run | this channel is publicly logged at https://bytecode.news/logs/irc/libera/primate and mirrored to Primate discord, channel #irc
  235. bluedreamreal: BCN intro is GOOD
  236. blueforgive my preposterous delay
  237. dreamrealheh
  238. bluedreamreal: another thing I thought about: you reading the article aloud is gonna shut down good all those trolls on HN that say the website is completely ai-generated
  239. dreamrealYeah, imagine someone actually writing decent diction
  240. dreamrealbut I'm not worried about the trolls
  241. dreamrealreading it aloud also shows some inversions of thought process, which may feed into the posts themselves
  242. dreamrealLike, as a writer I know I'm going to get there, but as a speaker, you close some gaps sooner than you might otherwise have
  243. bluemhm
  244. * lioloc quit (Read error: Connection reset by peer)
  245. * jreicher quit (Quit: In transit)
  246. * jreicher joined #primate
  247. bluedreamreal: jreicher: https://www.youtube.com/watch?v=L01w2jJJGek
  248. nevetYouTube: John Butler Trio - Media | Because Music
  249. * jreicher is now known as Guest8319
  250. * jreicher joined #primate
  251. * Guest8319 quit (Ping timeout: 246 seconds)
  252. * jreicher is now known as Guest8204
  253. * Guest8204 quit (Killed (manganese.libera.chat (Nickname regained by services)))
  254. * jreicher joined #primate
  255. * jreicher is now known as Guest5791
  256. * Guest5791 quit (Killed (niobium.libera.chat (Nickname regained by services)))
  257. * jreicher joined #primate
  258. * jreicher is now known as Guest4711
  259. * Guest4711 quit (Killed (manganese.libera.chat (Nickname regained by services)))
  260. * jreicher joined #primate
  261. * jreicher quit (Killed (niobium.libera.chat (Nickname regained by services)))
  262. * jreicher joined #primate
  263. * nevet joined #primate
  264. * blue!~blue@user/blue changed the topic to: https://primate.run | this channel is publicly logged at https://bytecode.news/logs/irc/libera/primate and mirrored to Primate discord, channel #irc