Web  1 sites indexed in VXNA
10 articles feed address
 utodea's recent timeline updates
utodea's repos on GitHub
Rust · 84 watchers
waylog-cli
Sync and preserve your AI coding assistant history (Codex, Claude Code, Gemini CLI) into local Markdown logs
TypeScript · 69 watchers
WayLog
WayLog - Save & Export AI Chat History. A local-first extension that turns your fleeting AI conversations into a permanent, git-friendly knowledge base.
Go · 10 watchers
baton
Baton is a Go implementation of Symphony
Go · 3 watchers
llmpole
Find which LLM models can run on your hardware
TypeScript · 3 watchers
oh-my-antigravity
Oh My AntiGravity: Workflow and multi-agent orchestration layer for Antigravity CLI ! Your AntiGravity is not alone. Add hooks, agent teams, HUDs, and so much more.
0 watchers
awesome-deepseek-integration
Integrate the DeepSeek API into popular software
Python · 0 watchers
Book2Video
TypeScript · 0 watchers
boxbox
Rust · 0 watchers
boxlite
SQLite-like agent runtime: an embeddable micro-VM that stores agent state and runs agent logic safely.
0 watchers
DeepSeek-TUI
Coding agent for DeepSeek models that runs in your terminal
Ruby · 0 watchers
homebrew-tap
Sync and preserve your AI coding assistant history (Cursor, Claude Code, Gemini) into local Markdown logs.
0 watchers
rtk
CLI proxy that reduces LLM token consumption by 60-90% on common dev commands. Single Rust binary, zero dependencies
0 watchers
scoop-bucket
HTML · 0 watchers
shayne-snap.github.io
0 watchers
x-algorithm
Algorithm powering the For You feed on X
utodea

utodea

https://getroost.dev
V2EX member #238465, joined on 2017-07-03 19:49:05 +08:00
Whale 的坦白书:一个 DeepSeek Coding Agent
分享创造  •  utodea  •  Jun 27  •  Lastly replied by utodea
10
求大家荐书:有趣、接地气的书籍
阅读  •  utodea  •  Jan 1, 2025  •  Lastly replied by jciba5n4y6u
6
utodea's recent replies
https://github.com/usewhale/whale 可以试试这个,cache hit 可以到 99%
我是做了 agentic workflow 把内容都准备好、填好,然后最后浏览器里手动点击发布按钮。这应该更稳一点?
Jun 29
Replied to a topic by fakEHacker 程序员 DeepSeek 好贵啊
#38 cc 并没有对 ds 的一些常见错误做优化,而且 dsk 的 API 不支持 defer loading 。https://github.com/usewhale/whale 引入了类似于这个 tweet 里一些针对 DeepSeek 的修复机制: https://x.com/MrAhmadAwais/status/2050956678502420612 ,因为每家的模型几乎都有这或者有那的一些“问题”或者特性
Jun 29
Replied to a topic by fakEHacker 程序员 DeepSeek 好贵啊
可以试试我这个: https://github.com/usewhale/whale , 长会话 cache hit 可以达到 99%。 你这个花费感觉没道理的。
cache hit 并不特别代表什么, 会话拉长、prompt token 适当多费点,很容易拉高。 比如我这个: https://github.com/usewhale/whale 。 优化的时候 agent 开发者可以参考它, 但过低多少有点猫腻。

115 turns · $0.0841 · 98.8% cache · last prompt 94.7K · $2.8607 cache saved
@c0xt30a #9, 可以试试用 npm 或者 homebrew 来安装。
一定会崩的,熵是递增的,而且以 10x ,100x 的速度递增(人写的当然也会崩)。但你想要的答案现在估计也没人知道,更多都是通过人来约束,openclaw 就是很好的观察例子。至于文档(包括 AGENTS.md )、甚至代码等这些对 AI 上都不是强约束。
@c0xt30a 谢谢! 是指安装脚本的域名不够友好?还是 curl 安装不够友好?
@cwr31 #5 。 你的质疑是对的。1.作为第三方能做的就是引入了类似于这个 tweet 里一些针对 DeepSeek 的修复机制: https://x.com/MrAhmadAwais/status/2050956678502420612 ,因为每家的模型几乎都有这或者有那的一些“问题”或者特性 。2.这就更多是一个宣传口号... 文中也有提到如果有官方的 harness ,还是尽量用官方的 harness 。
Jun 25
Replied to a topic by IanHo 程序员 我的第一个开源项目 Star 破 700 了!
Nice 。快两个月了,也快到 700 了,https://github.com/usewhale/Whale
About   ·   Help   ·   Advertise   ·   Blog   ·   API   ·   FAQ   ·   Solana   ·   5765 Online   Highest 6679   ·     Select Language
创意工作者们的社区
World is powered by solitude
VERSION: 3.9.8.5 · 73ms · UTC 01:46 · PVG 09:46 · LAX 18:46 · JFK 21:46
♥ Do have faith in what you're doing.