and a missing cache-mark that will make skills & project-claude.md cachemiss every time too https://github.com/anthropics/claude-code/issues/47098

  TLDR: for now launch using `CLAUDE_CODE_DISABLE_GIT_INSTRUCTIONS=1 claude "Hello"`
  https://news.ycombinator.com/item?id=47754795

otherwise you will only ever hit on the tools-block, and your first follow-up-question

(Note: setting includeGitInstructions=false in settings.json is an option to and likely the better thing anyhow since that git-status will always be noise - but since one needs the "Hello" remembering one alias is simple same deal ;D)

Currently the whole query is build from 3 cache-able blocks:

  {tools | claude-version}, 
  {system-prompt | ~/.claude/claude.md | git-status},
  {skills | ./claude.md | user-prompt}

the "Hello" allows skills & ./claude.md to get cached again https://github.com/anthropics/claude-code/issues/47098

Before:

  claude
  > alive?
  > -> Yes, alive and ready. What do you need? 
  > /exit
  # (11k cache read, 6k cache write)

  claude
  > are you here?
  > -> Yes, I'm here. What can I help you with? 
  > /exit
  # (11k cache read, 6k cache write)
  ^^ nothing changed, still cashed 6k write tokens

  claude
  > alive?
  > -> Yes, I'm here. What can I help you with?  
  > /exit
  # (18k cache read)
  ^^ the "exact same question" DOES hit the cache
  
  git commit --allow-empty -m "Dummy"
  claude
  > alive?
  > -> Yes, alive and well. What do you need?
  > /exit
  # (11k cache read, 6k cache write)
  ^^ same question, but now git busts the cache....

  $ CLAUDE_CODE_DISABLE_GIT_INSTRUCTIONS=1 claude "Hello"
      Claude Code v2.1.104
   Hello
  ● Hello! How can I help you today?
   Alive?
  ● Yes, alive and ready to help! What are you working on?
  # (10k cache read, 6k cache write)
  
  
  
  $ CLAUDE_CODE_DISABLE_GIT_INSTRUCTIONS=1 claude "Hello"
      Claude Code v2.1.104
   Hello
  ● Hello! How can I help you today?
   are you alive?
  ● I'm an AI, so not alive in the biological sense — but I'm here and ready to help. What can I do for you?
  # (16k cache read, 20 cache write)
  
  
  
  $ git commit -m "just a dummy" --allow-empty
  [main 642ce16] just a dummy
  
  
  $ CLAUDE_CODE_DISABLE_GIT_INSTRUCTIONS=1 claude "Hello"
      Claude Code v2.1.104
   Hello
  ● Hello! How can I help you today?
   is this alive?
  ● Yes, I'm alive and running! Is there something I can help you with?
  # (16k cache read, 18 cache write)

- https://news.ycombinator.com/item?id=47739260 (Pro Max 5x quota exhausted in 1.5 hours despite moderate usage)

- https://news.ycombinator.com/item?id=47736476 (Anthropic downgraded cache TTL on March 6th)

But there has been some news on this too: https://github.com/anthropics/claude-code/issues/46829#issuecomment-4237689126

  At this point we have rolled out 1h prompt cache by default in a number of places for subscribers ...
  We also are not defaulting API customers to 1h yet -- this needs more testing to make sure it's a net improvement on average.
  We will soon be changing the client side default to 1h for a few queries

Comments URL: https://news.ycombinator.com/item?id=47754795

Points: 3

# Comments: 1

(think i picked up some line about not being aware how to use it in offline mode / calling this out as an "error" while it was the expected configuration at the time)

and the toilet problem being as banal as "well, you need to wait for the green ready light to turn on..." (and this just taking a bit longer than expected on the first go)

(Note: setting includeGitInstructions=false in settings.json is an option to and likely the better thing anyhow since that git-status will always be noise....)

Currently the whole query is build from 3 cache-able blocks:

  {tools | claude-version}, 
  {system-prompt | ~/.claude/claude.md | git-status},
  {skills | ./claude.md | user-prompt}

the "Hello" allows skills & ./claude.md to get cached again https://github.com/anthropics/claude-code/issues/47098

otherwise you will only ever hit on the tools-block, and your first follow-up-question

Before:

  claude
  > alive?
  > -> Yes, alive and ready. What do you need? 
  > /exit
  # (11k cache read, 6k cache write)

  claude
  > are you here?
  > -> Yes, I'm here. What can I help you with? 
  > /exit
  # (11k cache read, 6k cache write)
  ^^ nothing changed, still cashed 6k write tokens

  claude
  > alive?
  > -> Yes, I'm here. What can I help you with?  
  > /exit
  # (18k cache read)
  ^^ the "exact same question" DOES hit the cache
  
  git commit --allow-empty -m "Dummy"
  claude
  > alive?
  > -> Yes, alive and well. What do you need?
  > /exit
  # (11k cache read, 6k cache write)
  ^^ same question, but now git busts the cache....

  $ CLAUDE_CODE_DISABLE_GIT_INSTRUCTIONS=1 claude "Hello"
      Claude Code v2.1.104
   Hello
  ● Hello! How can I help you today?
   Alive?
  ● Yes, alive and ready to help! What are you working on?
  # (10k cache read, 6k cache write)
  
  
  
  $ CLAUDE_CODE_DISABLE_GIT_INSTRUCTIONS=1 claude "Hello"
      Claude Code v2.1.104
   Hello
  ● Hello! How can I help you today?
   are you alive?
  ● I'm an AI, so not alive in the biological sense — but I'm here and ready to help. What can I do for you?
  # (16k cache read, 20 cache write)
  
  
  
  $ git commit -m "just a dummy" --allow-empty
  [main 642ce16] just a dummy
  
  
  $ CLAUDE_CODE_DISABLE_GIT_INSTRUCTIONS=1 claude "Hello"
      Claude Code v2.1.104
   Hello
  ● Hello! How can I help you today?
   is this alive?
  ● Yes, I'm alive and running! Is there something I can help you with?
  # (16k cache read, 18 cache write)

- https://news.ycombinator.com/item?id=47739260 (Pro Max 5x quota exhausted in 1.5 hours despite moderate usage)

- https://news.ycombinator.com/item?id=47736476 (Anthropic downgraded cache TTL on March 6th)

Comments URL: https://news.ycombinator.com/item?id=47747209

Points: 2

# Comments: 0

launching with `CLAUDE_CODE_DISABLE_GIT_INSTRUCTIONS=1 claude "Hello"` till those are fixed seems to be th way

there is env.ENABLE_PROMPT_CACHING_1H_BEDROCK - but that is - as the name says "when using Bedrock"

for the raw API the docs are also clear -> "ttl": "1h" https://platform.claude.com/docs/en/build-with-claude/prompt...

but how to make claude-code send that when paying by API-key? or when using a custom ANTHROPIC_BASE_URL? (requests will contain cache_control, but no ttl!)

its all "cache_control": { "type": "ephemeral" } there is no "ttl" anywhere.

// edit: cc_version=2.1.104.f27

in particular i´d like to call claude-models - in openai-schema hosted by a reseller - with some proxy that offers anthropic format to my claude --- but it seems like nothing gets to fully line things up (double-translated tool names for example)

reseller is abacus.ai - tried BerriAI/litellm, musistudio/claude-code-router, ziozzang/claude2openai-proxy, 1rgs/claude-code-proxy, fuergaosi233/claude-code-proxy,

"my assumption" would have been some x/y/angle/zoom params on the url, maybe even with some of the settings encoded in... (autorotate, render-radius or terrain height would be first to come to mind)

my "not getting any tiles" must have been 2 or 3 days ago - and from the last time i had to mess with tiles i do remember that to be fairly ugly in regards to performance/scaling so i feel you ;)

also addon after trying the "direct link to an airport": - this still loads at the random-US-airport first. with some tweaking there is likely an entire tilefetch you can save^

thx!

also "copy this view" does nothing (neither location nor any settings, gives just the bare link)

pretty sure your old behavior was the broken one tho - i vaguely remember fugling with this to "fullscreen correctly" for a claude-in-docker-in-cygwin-via-MSYS2 a while ago

but `--host-resolver-rules` may work, https://chromium.googlesource.com/chromium/src/+/main/net/ba... @terry_hc

"closing the sidebar" seems to grant access to another menu that is not accessible elsewhere? (appears to be of the actual canvas? has "export as svg" options like the context-menu, but also has "redo & revert" for example)

also appears there is a bit of an a "dynamic pen" (similar to excalidraw´s thickness?) - this could be a bit more pronounced i think or/and maybe needs some smoothing (see https://news.ycombinator.com/item?id=44915897 and some of the other demos linked in there)

a onbeforeunload (or whatever the current go to) to "warn before closing the tab" would be neat too (possibly gated behind Incognito-Mode detection - or only triggerd if the user himself did a import at the beginning or such?...)

very neat tho

what shines combined with a push-subcommand when doing "active development" in the downstream repo - since it can shove the fix you made back into the upstream so you can commit and release it there

("push" may then also have a --with-commit option that 1) makes a commit in the local upstream 2) updates downstreams `.git-remote-files` commit-key)

on --no-commit idk... while i very much like the idea of it being able to auto-commit, default-enabled might be a little overreaching but not 100% certain on this... - maybe your `--commit ` could become `--ref ` instead to free up --(no-)commit? (ref might even be more git-nomenclature correct ;))

readme: no thanks, just on a throwaway anyhow ;)

for the "push", I think my idea was mostly about "local-remotes", think "I have both cloned locally, with both IDEs open going back&forth"

one injection there would be `../someupstream/file` vs. `../someupstream/.git/refs/HEAD:file`... aka "pick the file as is" (potentially marked as "${HEAD}-dirty" vs. "only committed things are truth" (and if just so one doesn't need a extra `cp` command ;))

`just placed in the directory for you to do as you please.` could open a "--auto-commit" option -> based on a template similar to the dry run? (ideally overridable in .git-remote-files)

--dry-run

a "push" subcommand? (especially in combination with 'overwrite to local repository-path" mentioned below, for remotes rater useless sure ;))

also your readme leaves "kinda open" what happens with the actual file(s), `.git-remote-files` is mentioned "should be committed", but the file it cloned?

also a little unclear how `--save` plays into that (since the .git-remote-files example shows only a commit no branch) (and when would one ever run it without save?)

cli-arg/secondary `.git-remote-files`-File (possibly as secondary `.local.git-remote-files` or such that can also override repository-URLs) for local/private repos?

option (autodetect?) to also write gitattributes to mark those picked files binary (what may also be done into the repo, or local-only into the .git/ dir of the repo...)

since its called `git-fetch-file` and not `.git-remote-files` a overall comment may be nice as refence when first generating the file ;)

but by now i´m just rambling, looking forward to actually try it when home ;) thanks in advance