feat(wiki): ingest remaining subdirectories batch (51 files)
- Others: ChinaTextbook, Obsidian笔记系列, YouTube Channel ID, TikTok PM Django - Skills: GOG CLI, Last30Days, baoyu-skills - Vibe Coding: Cursor 2.0, Trae远程开发, Vibe-Kanban+OpenCode, vibe coding经验 - 微信公众号: 养虾日记1-5, AI时代赚钱 - 跨境电商: TikTok数据抓取, 选品策略, Superset Dashboard - AI目录补充: 20个文件 Source pages: 51 Entities: TapXWorld, VibeKanban, OpenCode, Trae, SourceGrounding等 Concepts: 自举Meta生成, 5大设计原则, MD5去重, 混合搜索等
This commit is contained in:
35
wiki/sources/养虾日记4-Context调试.md
Normal file
35
wiki/sources/养虾日记4-Context调试.md
Normal file
@@ -0,0 +1,35 @@
|
||||
---
|
||||
title: "养虾日记4 Context调试"
|
||||
type: source
|
||||
tags: []
|
||||
date: 2026-04-10
|
||||
---
|
||||
|
||||
## Source File
|
||||
- [[raw/微信公众号/养虾日记4: 一次「Context Limit Exceeded」错误排查:我以为是小问题,结果踩了大坑.md]]
|
||||
|
||||
## Summary
|
||||
- 核心主题:OpenClaw Context Limit Exceeded错误排查实战
|
||||
- 问题域:AI Agent上下文溢出、模型配置问题、Gateway日志分析
|
||||
- 方法/机制:通过日志定位问题根源→发现Telegram channel绑定了小context模型→全局配置和agent配置是两回事
|
||||
- 结论/价值:不要默认认为错误信息就是表面意思,系统越复杂问题隐藏路径越深
|
||||
|
||||
## Key Claims
|
||||
- 错误表象是context满了,实际是Telegram channel绑定了deepseek-reasoner(16K context)
|
||||
- OpenClaw两层配置:全局compaction配置和agent模型配置是两码事
|
||||
- Gateway日志能清晰揭示问题根源,应养成看日志的习惯
|
||||
- 模型切换可能由API不可用、Token溢出预判、配置覆盖、负载均衡等多种原因触发
|
||||
|
||||
## Key Concepts
|
||||
- [[Gateway日志]]:问题诊断的第一手资料
|
||||
- [[模型Fallback]]:主模型不可用时自动切换备选模型
|
||||
- [[配置分层]]:Global/Agent/Channel/环境变量多层级配置
|
||||
|
||||
## Key Entities
|
||||
- [[deepseek-reasoner]]:16K context window的小模型
|
||||
|
||||
## Connections
|
||||
- [[OpenClaw]] ← debugged_by ← [[Gateway日志]]
|
||||
|
||||
## Contradictions
|
||||
- 无冲突
|
||||
Reference in New Issue
Block a user