LLM Wiki
A pattern for building personal knowledge bases using LLMs.
This is an idea file, it is designed to be copy pasted to your own LLM Agent (e.g. OpenAI Codex, Claude Code, OpenCode / Pi, or etc.). Its goal is to communicate the high level idea, but your agent will build out the specifics in collaboration with you.
The core idea
Most people's experience with LLMs and documents looks like RAG: you upload a collection of files, the LLM retrieves relevant chunks at query time, and generates an answer. This works, but the LLM is rediscovering knowledge from scratch on every question. There's no accumulation. Ask a subtle question that requires synthesizing five documents, and the LLM has to find and piece together the relevant fragments every time. Nothing is built up. NotebookLM, ChatGPT file uploads, and most RAG systems work this way.
| type Set struct { | |
| list map[int]struct{} //empty structs occupy 0 memory | |
| } | |
| func (s *Set) Has(v int) bool { | |
| _, ok := s.list[v] | |
| return ok | |
| } |
GitHub Search Syntax for Finding API Keys/Secrets/Tokens
As a security professional, it is important to conduct a thorough reconnaissance. With the increasing use of APIs nowadays, it has become paramount to keep access tokens and other API-related secrets secure in order to prevent leaks. However, despite technological advances, human error remains a factor, and many developers still unknowingly hardcode their API secrets into source code and commit them to public repositories. GitHub, being a widely popular platform for public code repositories, may inadvertently host such leaked secrets. To help identify these vulnerabilities, I have created a comprehensive search list using powerful search syntax that enables the search of thousands of leaked keys and secrets in a single search.
Search Syntax:
(path:*.{File_extension1} OR path:*.{File_extension-N}) AND ({Keyname1} OR {Keyname-N}) AND (({Signature/pattern1} OR {Signature/pattern-N}) AND ({PlatformTag1} OR {PlatformTag-N}))
Examples:
**1.
| ; gcc -O2, x86-64, Intel syntax (via godbolt.org) | |
| ; Annotated excerpts. Full listing below. | |
| ; =========================================================== | |
| ; (1) The two comparators are byte-for-byte identical. | |
| ; One reads an int; one reads the first field of a struct. | |
| ; Nothing at the machine level distinguishes them. | |
| ; =========================================================== | |
| "cmp_int": |
| $ cargo run | |
| [3, 7, 19, 42, 88] | |
| [Point { x: 1, y: 2 }, Point { x: 2, y: 3 }, Point { x: 3, y: 1 }] | |
| Circle r=2 (rank 2) | |
| Square s=3 (rank 3) | |
| &dyn Describe = 16 bytes | |
| &Circle = 8 bytes |
在 ChatGPT 对话里做一个动态状态栏
通用教程与实现思路。作者：Dan & Joy
效果思路很简单：让模型调用一个只读 MCP 工具，工具返回 mood / place / focus / note 等结构化数据；ChatGPT 再把这份数据交给一个内嵌 HTML 组件。组件可以用 CSS 做动画，用 JavaScript 更新时间、展开面板和处理点击，也可以把用户刚刚操作后的状态带到下一轮对话。
它不是修改 ChatGPT 原生界面，也不是普通 Markdown 图片，而是一个运行在对话消息里的 MCP Apps 组件（iframe）。示例字段和文案都只是占位，每个人都可以按照自己的角色、项目或使用场景替换。
1. 架构
| #!/usr/bin/env perl | |
| ## | |
| ## Easily disable fonts that aren't for your language, or language-specific | |
| ## variants from large font families like Noto -- WITHOUT uninstalling them | |
| ## | |
| ## Authors: Kevin Ernst <ernstki -at- mail.uc.edu>, @wdoekes | |
| ## License: MIT or CC-BY-SA-4.0, at your option | |
| ## Source: https://gist.github.com/ernstki/2ae279ad89888e4099c9852bf0ba5d11 | |
| ## | |
| ## Usage: |