Codex接CPA,触发自动压缩上下文时请求 /responses/compact 直接 404 了。 让AI看了下源码说CPA这个路径没实现。 我打开方式不对吗?或者我应该关掉Codex的自动压缩? 2 个帖子 - 2 位参与者 阅读完整话题
Error running remote compact task: unexpected status 401 Unauthorized: Encountered invalidated oauth token for user, failing request, url: https://chatgpt.com/backend-api/codex/responses/compact , cf-ray: a087a5e7bc81039f-HKG, request id: 08529d0b-65f5-453a-9a20-81078d4da905, auth error: 401, auth error code: token_revoked 修正一下 aef.claudeai.life 注册的两个号都挂了,一个用了一部分 一个都没登录,都挂了 另外一个team 还能用 不知道能蹬多久 10 个帖子 - 9 位参与者 阅读完整话题
突然发现我好像从来没有主动使用过这个命令。我一般是先出文档,再开新窗口开始vibe,期间基本上没管过了。 13 个帖子 - 9 位参与者 阅读完整话题
一直对这方面有一些疑惑,如果是要运行一个长期的任务例如autoresearch难道就是设定goal之后让codex一直跑然后上下文也让他自己来compact,但是这样会不会因为信息损失导致方向越来越偏等等。还有现在agent都在宣传长时间的自主运行,实际使用的时候应该如何能让他长时间自主运行的效果更好呢? 8 个帖子 - 8 位参与者 阅读完整话题
有佬友遇到过吗,基本没有压缩成功过 1 个帖子 - 1 位参与者 阅读完整话题
感觉有1-2 周左右了, 无论是使用 glm 还是 claude 自己 都经常出现,上下文无限堆积 就是不会 自动compact,尝试 让claude 自己解决,但是无解 /model 切换 可以确认不是使用的 4.8 1M上下文模式 1 个帖子 - 1 位参与者 阅读完整话题
Error running remote compact task: stream disconnected before completion: error sending request for url ( https://chatgpt.com/backend-api/codex/responses/compact ) 基本上过一小段时间会话就失败了,不是中转,是直接登录的pro,求一个解决方案 1 个帖子 - 1 位参与者 阅读完整话题
LumenTale https://lumentale.org/ is the kind of game where early choices can shape how comfortable the rest of a run feels. Players may want to explore freely, but they also need a way to check starter options, Animon details, affinity information, and evolution paths when the systems begin to overlap. A good reference should support discovery instead of replacing it. That is why a compact resource hub can be useful for Memories of Trey players. When guide material is scattered across different posts, it is easy to lose track of which page explains starters, which one covers items, and which one has map information. A single organized reference makes it simpler to move from a question to an answer. For beginners, the first helpful step is usually understanding the basic shape of the game. Starter picks, skills, item uses, and boss preparation all become easier when the information is grouped clearly. The goal is not to tell every player exactly what to do. It is to make the first few decisions less confusing. LumenTale works well as that kind of companion because it brings together Animon, starters, evolutions, affinity notes, Fountain recipes, items, boss fights, and the Talea map in one place. A player can check one topic quickly, then return to playing without turning the whole experience into a checklist. That balance matters. Games are more memorable when players still make their own discoveries, but most people appreciate a clear reference after they get stuck. A lightweight guide can keep the fun moving while still leaving space for experimentation. For anyone starting LumenTale or returning after a break, it is worth keeping a resource like this nearby. It helps with planning, comparison, and quick reminders, especially when a run starts to involve multiple systems at once.
前面刚刚/compact,之后第一条对话正常回复中文,第二条我还是中文提问,结果它开始飙日语了 5 个帖子 - 5 位参与者 阅读完整话题
Error running remote compact task: unexpected status 503 Service Unavailable: No available channel for model gpt-5.5-openai-compact under group default (distributor) (request id: 202606030214337343574988268d9d6GXCSdaxv) 不知道佬友遇到过没有,今天突然遇到了,codex一直都是有更新就更新,不知道是不是codex的问题,还是api问题。有没有解决之道? 2 个帖子 - 2 位参与者 阅读完整话题
Error running remote compact task: stream disconnected before completion: error sending request for url ( https://la1api.top/responses/compact ) 是我配置有啥问题,还是这中转不行?
Error running remote compact task: stream disconnected before completion: error sending request for url ( https://la1api.top/responses/compact ) 是我配置有啥问题,还是这中转不行?
Error running remote compact task: stream disconnected before completion: error sending request for url ( https://la1api.top/responses/compact ) 是我配置有啥问题,还是这中转不行?
如题,gpt-5.5-openai-compact跟gpt-5.5有什么区别吗? 7 个帖子 - 5 位参与者 阅读完整话题
cc降级到154,配置也按大佬们说的改了,但还是无法恢复session,/compact也是卡在95%,于是我新开一个session 然后cc帮我截断有效信息,重进session再/compact就可以继续了,赶紧让它把内容保存到/memory里 1 个帖子 - 1 位参与者 阅读完整话题
如何处理cc在compact的时候超出上限。。? 我给用满了,结果在compact跑了一大半之后compact的上下文窗口也满了导致无法compact。 1 个帖子 - 1 位参与者 阅读完整话题
为啥any的opus1m在60k的时候就会返回524,然后compact也是卡在95%,但是新对话能正常使用 2 个帖子 - 2 位参与者 阅读完整话题
Error running remote compact task: stream disconnected before completion: missing field output at line 1 column 42679 上下文达到80左右的时候就开始一直报错这个,并且压缩上下文一直失败,不知道什么原因, 只要重新打开一个对话立马解决. 但是新开对话会导致上下文丢失. 试过手动压缩也是不行的. n次 使用到第三方api 非官方的订阅, 我个人感觉是不是第三方api的账户额度用完了切换账号然后再去压缩的时候就会导致这个问题. 1 个帖子 - 1 位参与者 阅读完整话题
Error running remote compact task: stream disconnected before completion: missing field `output` at line 1 column 34172 6 个帖子 - 3 位参与者 阅读完整话题
不知道触发了什么奇奇怪怪的 bug 。 搁那自己“内耗”了两个钟头屁事没干。耗费了我一堆 token 。 切回 gpt-5.5 干活了。