rtk

Oh the verbosity and the cryptic words that it uses. The other day, all of a sudden it used an acronym "DoD". I had no idea what it was and made me feel dumb. It's "Definition of Done". I don't care how widely used this acronym is, you just can't throw it in there. I've now installed quite a number of tools to combat this. Just in the last few days I've installed - https://www.codewithbullet.com - https://maki.sh - https://github.com/rtk-ai/rtk Has it helped? Somewhat.

> I think the issues is, it is going against a very well established pattern. I have a tool that wraps ripgrep so that search results always includes context and from time to time, the agent will use ripgrep by itself and when I ask why, it would go "yeah I should have done that" There's a limit of how many simultaneous instructions an agent can follow (the exact number depends on the specific model so instructions that are fine for one model may overwhelm another). If this keeps happening, consider trimming your instructions or even better, solving it at the harness level (like intercepting and rewriting ripgrep calls to use your thing, like rtk [0] does in agents that supports this) Overall, never leave to an agent an instruction that must be followed at all times. For example, doing things in a git hook beats a multi-command workflow e

My criteria is "do they measure performance, or at least even try to?". Caveman [1], RTK [2] and more recently ponytail [3] don't or use a few trivial tests. Those projects don't measure performance on widely used benchmarks (like SWE Pro and stuff), that have their issues but at least it would give some indication. They also don't measure "big model + caveman vs smaller model". I've had a few times where removing all custom instructions that I started using with model N-2 made model N perform way better, so I'm very suspicious of everything that changes how the model works, it's easy to get degraded performance silently and suddenly you're paying latest Opus costs for 6 months old Sonnet performance. [1]: https://github.com/JuliusBrussee/caveman [2]: https://github.com/rtk-ai/rtk [3]: https://github.com/DietrichGebert/ponytail

I was looking through the issues as investigation. Some issues that caught my attention are looking quite bad https://github.com/rtk-ai/rtk/issues/2494 https://github.com/rtk-ai/rtk/issues/2462 https://github.com/rtk-ai/rtk/issues/2395

I might have picked better title, but they are literally called rtk https://github.com/rtk-ai/rtk and it stands for Rust Token Killer

100% with the author on that one, albeit the performance decay seems to depend on the type of task for me. Simple plumbing tasks seem to run okay with longer running contexts. Also, some colleagues were playing around with RTK (https://github.com/rtk-ai/rtk), which decreases the amount of token used by tool calls and, although it seems an interesting idea, I am pretty sure there are many caveats. Although, I believe if these type of tools prove to be efficient enough, perhaps harnesses will have them natively.

Haven't benchmarked it, but for agentic coding e.g. Claude Code I found this useful: https://github.com/rtk-ai/rtk

I can see that and I don't know your setup, but there are people pushing >70t/s with MTP on a single 3090, with big contexts still >50t/s. 64k is not a lot for agentic coding, and IIRC 128k with turboquant and the likes should be possible for you. r/LocalLLM/ and r/LocalLLaMA/ are worth a visit IMO. EDIT: just found this recipe repo, may wanna give it a go: https://github.com/noonghunna/club-3090 EDIT-2: this can also shave off a lot of context need for tool calling -> https://github.com/rtk-ai/rtk

Since this site seems to have an allergy to information density and usability, here's the GH: https://github.com/mksglu/context-mode Some discussion about this from a couple of months ago: https://news.ycombinator.com/item?id=47193064 Similar tools in this space: RTK:CLI proxy that reduces LLM token consumption: https://github.com/rtk-ai/rtk 8v: One CLI for you and your AI agent. Posted here a few days ago: https://news.ycombinator.com/item?id=47914963 Headroom:The Context Optimization Layer for LLM: https://github.com/chopratejas/headroom Of these, it looks like RTK and 8v are somewhat equivalent, while headroom is complementary. Although this does perhaps go further than rtk/v8.

Similar tools in the space: RTK which I guess is a close equivalent. https://github.com/rtk-ai/rtk Headroom which is complementary. https://github.com/chopratejas/headroom I have tried both of these and I have to say, I didn't notice any difference. That's not to say there wasn't a difference (rtk logs were telling me enthusiastically how many tokens I'd saved), it's just that pretty much all context is coming from reading data from files and neither of these tools can do anything about that. As I switched from Claude Code to Pi recently, and tweaked my workflow, I've seen a much bigger/more obvious reduction in context use from that and I don't use either of these tools right now. I'm open to trying them again though, has anyone had a different experience than mine?

This smells heavily of astroturfing. Particularly because Headroom is a paid product, and that fact is not mentioned here or in the GitHub README. Here was my experience… I download and run the Mac application, which starts installing a bunch of things. Then the following happens without advance notice: - Adds background item(s) from "Idiosyncratocracy BV" - Downloads over 2 GB of files - Pollutes home with ~/.headroom directory - Adds hook(s) to ~/.claude/hooks/ - Modifies your ~/.claude/settings.json to add above hook(s) … and then I see something in the settings that talks about creating an account. That's when I realized that this is a paid product, after all of the above has happened. Headroom seems to use https://github.com/rtk-ai/rtk under the hood. What does Headroom offer over the actually-free RTK? Who knows. At this point I ha

What about some thing like https://github.com/rtk-ai/rtk

To reduce token count on command outputs you can also use RTK [0] [0]: https://github.com/rtk-ai/rtk

Also https://github.com/rtk-ai/rtk but some people see that changing how commands output stuff can confuse some models