Chen Yulin's Blog

Posted 2024-10-17Updated 2025-05-13Note2 minutes read (About 321 words)

Objective

I use a hyprland plugin called hyprscroller to emulate the scroll functionality of PaperWM.
However, I can’t get whether I am in row mode or column mode, so I want to show the mode status on my waybar.

Implementation

I wrote two shell script under ~/.config/waybar/

scroller_mode_listener.sh, responsible for listening the IPC message of hyprscroller and store that information in a tmp file. This script can be executed in hyprland config file using exec-once

#!/bin/bash

MODE_FILE="/tmp/hyprland_scroller_mode"

function handle {
  if [[ ${1:0:8} == "scroller" ]]; then
    if [[ ${1:10:9} == "mode, row" ]]; then
        echo "Row" > "$MODE_FILE"
    elif [[ ${1:10:12} == "mode, column" ]]; then
        echo "Column" > "$MODE_FILE"
    #else
        #echo "" > "$MODE_FILE"
    fi
    hyprctl dispatch submap reset && pkill -SIGRTMIN+8 waybar # update the widget on waybar
  fi
}

# Initialize the mode file
echo "" > "$MODE_FILE"

socat - "UNIX-CONNECT:$XDG_RUNTIME_DIR/hypr/$HYPRLAND_INSTANCE_SIGNATURE/.socket2.sock" | while read -r line; do handle "$line"; done

scroller_mode_reader.sh, responsible for reading the tmp file and return the json value to waybar to change the shown icon.

#!/bin/bash

MODE_FILE="/tmp/hyprland_scroller_mode"

current_mode=$(cat "$MODE_FILE")

if [[ "$current_mode" == "Row" ]]; then
    icon="Row"
    percent=0
    class="mode-row"
elif [[ "$current_mode" == "Column" ]]; then
    icon="Column"
    percent=100
    class="mode-column"
else
    icon="Row"
    percent=0
    class=""
fi

echo "{\"text\":\"$icon\", \"tooltip\":\"Scroller Mode: $current_mode\", \"class\":\"$class\",\"percentage\": $percent}"

Then add the custom module to waybar in config file (we use signal to update the module):

"custom/scroller-mode": {
    "exec": "~/.config/waybar/scroller_mode_reader.sh",
    "return-type": "json",
    "interval": "once",
    "signal": 8,
    "on-click": "hyprctl dispatch submap reset && pkill -SIGRTMIN+8 waybar",
    "format": "{icon}",
    "format-icons": ["", ""],
}

Also remember to add the module to the bar:

1	"modules-right": ["custom/scroller-mode", "cpu", "memory", "temperature","battery","clock","custom/power"],

Then everything is done

Posted 2024-10-17Updated 2025-05-13读读噜5 minutes read (About 816 words)

《学做工：工人阶级子弟为何继承父业》读书会p11

特别值得一提的是，反学校文化与工人阶级的车间文化有很多根本的相似之处——大部分反学校文化成员注定成为工人阶级的一分子。尽管我们必须考虑地区和职业的差异，但工人阶级车间文化的核心是：无论条件多么艰苦，上级指导多么苛刻，人们总是在寻求意义，并为之搭建参照体系。他们运用自己的才能，即使在最受他人控制的活动中也能寻求到乐趣。矛盾的是，他们能从死板的工作经历中活出一种生动的文化，这绝不是对失败的简单反映。在反学校文化中，我们亦能看到这种在令人疏离的环境中立足的根本理念，同时，他们都试图在枯燥的制度下编织出属于自己的兴趣和娱乐。这些文化不只是人类与不快之间层层叠叠的填充物。这些文化自有其逻辑，使用各种技巧、动作和行为，以达到特定目的。

车间文化和反学校文化一样，基于相同的组织性团体。非正式群体奠定了所有其他可能的文化要素。正是这个团体生成并传播着那些与官方权威争夺符号控制和真实空间的策略。这个非正式组织无处不在，并将车间文化与中产阶级工作文化区分开来。

家伙们”也从根本上排斥文凭这个概念。对他们而言，
正如制度界定的那样，“文凭”就是知识权力的爪牙。因为他
们反抗知识，所以他们也必然抵制和怀疑文凭并使其失去信
用。正如其他情况那样，使正式标准失去信用的主要方式是用
非正式的模式“看透这些标准，了解它是如何发挥作用的”。
他们靠经验，至少是听来的经验与正式定义周旋。在一定程度
上，他们真的认为自己懂的更多。不做功课、没有文凭照样能
过活，因为真正有用的是“了解点世界”“脑子清楚”、必要
的时候“撒手”。当然，成人工人阶级世界中，英国车间里盛
行的实用氛围及对理论的不信任都大大强化了他们的这种观
点。这一基本文化资源在学校的分化过程中被磨炼得更为坚
固。
“家伙们”从心底里认为肯定有更简单的方法。文凭对他
们来说似乎是对直接行动的偏离或替换。他们觉得他们总能
“在岗位上”展示必要的能力，而且做一件事总比描述它或在
考试中陈述它容易，或者比正式说明它容易。

Posted 2024-10-12Updated 2025-05-13Note2 minutes read (About 291 words)

Basic Neovim Plugin Development Tutorial

Basic Plugin File

Just take my ColorfulDiff.nvim as an example.

$ tree ColorfulDiff.nvim

├── LICENSE
├── lua
│   └── tint.lua
├── plugin
│   └── hello.lua
└── README.md

The lua scripts in plugin/ will be executed automatically and the scripts in lua/ will be called when required

To let nvim load the plugin, use lazy.nvim.
Add this line in your lazy.nvim configuration:

1	{ dir = "/home/cyl/nvim_plugins/ColorfulDiff.nvim/" },

Then this plugin will be loaded on start.

How To Use API

Use :Telescope help_tags or short hand <Ctrl-h> to open the help search panel.
If I want to know the API regarding the highlight:

the function with `nvim_` as prefix are the nvim api. Press `Enter` to see the detailed description:

 *nvim_buf_add_highlight()*
nvim_buf_add_highlight({buffer}, {ns_id}, {hl_group}, {line},{col_start},{col_end})
    Adds a highlight to buffer.

    Useful for plugins that dynamically generate highlights to a buffer (like
    a semantic highlighter or linter). The function adds a single highlight to a buffer. Unlike |matchaddpos()| highlights follow changes to line
    numbering (as lines are inserted/removed above the highlighted line), like signs and marks do.
    ...

To use this api:

1	vim.api.nvim_buf_add_highlight(...)

Test

Create a new folder called test
And create the test lua file

describe("colorfulDiff", function()
	it("can be required", function()
		require("colorful_diff")
	end)
	it("...", function()
		...
	end)
	...
end)

然后在该文件中按<leader>pt运行测试

## Reference

1 2	:h lua-plugin :h write-plugin

https://www.youtube.com/watch?v=n4Lp4cV8YR0&t=3960s

Posted 2024-10-11Updated 2025-05-13Note5 minutes read (About 677 words)

VG501 Description of Experimental Method and Details

Background Introduction to the Methodology: Motion Planning with Transformers

Motion planning is a critical aspect of robotics, enabling robots to navigate complex environments while avoiding obstacles and achieving specific goals. Traditional motion planning methods, such as sampling-based planners, often struggle with high-dimensional tasks and long-term planning due to their reliance on random sampling and tree expansion techniques. Recent advancements in deep learning have led to the development of neural motion planners, which leverage neural networks to learn from demonstrations and generate efficient paths. Among these, the Motion Planning Transformer (MP-Former) framework stands out by treating motion planning as a sequence-to-sequence generation task, utilizing the transformer architecture to capture long-term dependencies and improve path quality.

Experimental Method: Evaluation of MP-Former

In this study, we conducted a series of experiments to evaluate the performance of the MP-Former framework in generating collision-free, near-optimal paths for robotic manipulation tasks. The experiments were designed to compare MP-Former against traditional sampling-based planners and previous neural motion planners in terms of efficiency, path quality, and robustness.

Experimental Details

Sample Information

The experiments utilized a synthetic dataset comprising various complex environments represented as point clouds. Each environment included multiple obstacles and varying configurations to challenge the motion planning capabilities of the algorithms.

Procedures

Dataset Preparation: A synthetic dataset was generated, consisting of 100 unique environments with diverse obstacle arrangements. Each environment was represented as a point cloud, simulating real-world scenarios.
Algorithm Implementation: The MP-Former framework was implemented alongside traditional planners (e.g., RRT and A*) and previous neural planners (e.g., MPNet). Each algorithm was configured to operate under the same conditions for a fair comparison.
Path Generation: For each environment, the algorithms were tasked with generating paths from a predefined start configuration to a goal configuration. The MP-Former generated paths using its transformer-based architecture, while the traditional planners utilized their respective sampling and tree expansion methods.
Data Recording: The following metrics were recorded for each generated path:

Path length (in meters)
Number of collisions encountered
Time taken to generate the path (in seconds)
Success rate (percentage of successful path generations without collisions)

Instruments and Equipment

Computational Resources: Experiments were conducted on a high-performance computing cluster equipped with NVIDIA GPUs to facilitate the training and evaluation of the MP-Former framework.
Software Tools: The implementation of the algorithms was carried out using Python with libraries such as TensorFlow for neural network training and Open Motion Planning Library (OMPL) for traditional planners.

Repetitions and Parallel Experiments: Each algorithm was executed 50 times per environment to ensure statistical significance. The results were averaged across these repetitions to provide a comprehensive evaluation of each method’s performance.

Results and Analysis

The results of the experiments were compiled into tables and graphs to illustrate the performance differences between the MP-Former and the other planners.

Table 1: Performance Metrics Comparison

Algorithm	Average Path Length (m)	Average Time (s)	Success Rate (%)	Average Collisions
MP-Former	5.2	1.5	95	0.2
RRT	6.8	3.2	80	1.5
A*	6.5	2.8	85	1.0
MPNet	5.9	2.0	90	0.5

Justification of Methodology

The choice of using the MP-Former framework for motion planning tasks is justified by its ability to efficiently generate high-quality paths while maintaining robustness in complex environments. By treating motion planning as a sequence-to-sequence generation task, MP-Former effectively captures long-term dependencies, reducing the risk of falling into equilibrium points that can hinder traditional planners. The experimental results demonstrate that MP-Former outperforms traditional methods in terms of path quality, efficiency, and success rate, making it a valuable approach for robotic motion planning.

Reference

Li, B., Wang, R., Chen, Y., Feng, B., Zhou, Q., & Bi, Y. (2025). Motion Planning Transformers (MP-Former): Treat Motion Planning as Sequence-to-Sequence Generation. IEEE International Conference on Robotics and Automation (ICRA).
Karaman, S., & Frazzoli, E. (2011). Sampling-based algorithms for optimal motion planning. The International Journal of Robotics Research, 30(7), 846-894.
Gammell, J. D., Srinivasa, S. S., & Barfoot, T. D. (2015). Batch informed trees (BIT*): Sampling-based optimal planning via the heuristically guided search of implicit random geometric graphs. IEEE International Conference on Robotics and Automation (ICRA).

Posted 2024-10-09Updated 2025-05-13Reference3 minutes read (About 444 words)

Guideline For My Nvim Configuration

For what

nvim用了两年了，插件和自己的配置真搞了不少，但是很容易出现：

我记得这个功能好像通过什么插件实现了来着，插件叫什么名字来着？（对着一堆名字牛头不对马嘴的插件望洋兴叹）对应的指令是啥来着？快捷键是啥来着？

装了若干个实现类似功能的插件。。。

这个插件更新了？不会用辽！

我当时为啥装这个插件来着？

我刚一不小心触发了个啥玩意儿的快捷键？

……

综上，决定按照功能分块梳理一下自己的插件和快捷键配置。顺带整理一下插件系统，把不要的统统删咯。

Guide

页面

Bufferline

左右循环移动bufferCtrl+h/l
选择buffer关闭Leader+bc

Ctrl+w（重新映射了一些C-W指令）

关闭当前分屏sc
关闭所有其他分屏so
垂直分屏sv
分屏sh
移动指针处于的窗口Alt+h/j/k/l
窗口比例调整Ctrl+上/下/左/右
一键等比例s=

Terminal

打开浮动终端Leader+t
垂直分屏打开终端Leader+vt
退出: 终端输入模式下Esc

搜索

搜索文件Ctrl+p
搜索文件内容Ctrl+f
搜索帮助文档Ctrl+h

编辑

指针移动/跳转/导航

跳回上一个位置Ctrl+o
重做跳回Ctrl+i

文件管理

打开文件树F2
重命名文件/文件夹r
新建文件/文件夹a
删除文件/文件夹d
折叠所有文件夹W
搜索S

LSP

显示效果增强

插件开发

运行_spec测试<leader>pt

Posted 2024-10-08Updated 2025-05-13Note6 minutes read (About 967 words)

Hexo AI Assistant

参见 https://dingfen.github.io/2024/07/21/2024-7-21-ai_assist/

阿里云的大模型之前在搞wechat bot的时候接触过，这次再加上一个云函数的服务，就搞定了。
LLM:https://bailian.console.aliyun.com/?spm=a2c4g.11186623.0.0.2db33048PGTSE6#/app-center
云函数:https://fcnext.console.aliyun.com/applications?spm=5176.fcnext.0.0.65f378c8HmZpOP

在本地主要需要做的是在hexo中注入代码
添加文件scripts/injector.js

hexo.extend.injector.register('body_end',`
<link rel="stylesheet" crossorigin href="https://g.alicdn.com/aliyun-documentation/web-chatbot-ui/0.0.11/index.css" />
<script type="module" crossorigin src="https://g.alicdn.com/aliyun-documentation/web-chatbot-ui/0.0.11/index.js"></script>
<script>
  window.CHATBOT_CONFIG = {
    endpoint: "/chat", // 可以替换为 https://{your-fc-http-trigger-domain}/chat
    displayByDefault: false, // 默认不展示 AI 助手聊天框
    aiChatOptions: { // aiChatOptions 中 options 会传递 aiChat 组件，自定义取值参考：https://docs.nlkit.com/nlux/reference/ui/ai-chat
      conversationOptions: { // 自定义取值参考：https://docs.nlkit.com/nlux/reference/ui/ai-chat#conversation-options
        conversationStarters: [
          {prompt: '哪款手机续航最长？'},
          {prompt: '你们有哪些手机型号？'},
          {prompt: '有折叠屏手机吗?'},
        ]
      },
      displayOptions: { // 自定义取值参考：https://docs.nlkit.com/nlux/reference/ui/ai-chat#display-options
        height: 600,
      },
      personaOptions: { // 自定义取值参考：https://docs.nlkit.com/nlux/reference/ui/ai-chat#chat-personas
        assistant: {
          name: '你好，我是你的 AI 助手',
          // AI 助手的图标
          avatar: 'https://img.alicdn.com/imgextra/i2/O1CN01Pda9nq1YDV0mnZ31H_!!6000000003025-54-tps-120-120.apng',
          tagline: '您可以尝试点击下方的快捷入口开启体验！',
        }
      }
    }
  };
</script>
<style>
  :root {
    /* webchat 工具栏的颜色 */
    --webchat-toolbar-background-color: #1464E4;
    /* webchat 工具栏文字和按钮的颜色 */
    --webchat-toolbar-text-color: #FFF;
  }
  /* webchat 对话框如果被遮挡，可以尝试通过 z-index、bottom、right 等设置来调整位置 */
  .webchat-container {
    z-index: 100;
    bottom: 10px;
    right: 10px;
  }
  /* webchat 的唤起按钮如果被遮挡，可以尝试通过 z-index、bottom、right 等设置来调整位置 */
  .webchat-bubble-tip {
    z-index: 99;
    bottom: 20px;
    right: 20px;
  }
</style>`
);

主要要更改的是window.CHATBOT_CONFIG.endpoint, 改为触发器的公网访问地址

然后更改avatar，添加浮动的对话泡图标balabala

hexo.extend.injector.register('body_end',`
<link rel="stylesheet" crossorigin href="https://g.alicdn.com/aliyun-documentation/web-chatbot-ui/0.0.11/index.css" />
<script type="module" crossorigin src="https://g.alicdn.com/aliyun-documentation/web-chatbot-ui/0.0.11/index.js"></script>
<script>
  window.CHATBOT_CONFIG = {
    endpoint: "https://webchat-bot-iqu-knzhgrvznd.cn-hangzhou.fcapp.run/chat", // 可以替换为 https://{your-fc-http-trigger-domain}/chat
    displayByDefault: false, // 默认不展示 AI 助手聊天框
    aiChatOptions: { // aiChatOptions 中 options 会传递 aiChat 组件，自定义取值参考：https://docs.nlkit.com/nlux/reference/ui/ai-chat
      conversationOptions: { // 自定义取值参考：https://docs.nlkit.com/nlux/reference/ui/ai-chat#conversation-options
        conversationStarters: [
          {prompt: '你是谁？'},
          {prompt: '博主又是谁？'},
          {prompt: '博主喜欢的人是？'},
          {prompt: '想要博主联系方式！'},
        ]
      },
      displayOptions: { // 自定义取值参考：https://docs.nlkit.com/nlux/reference/ui/ai-chat#display-options
        height: 600,
      },
      personaOptions: { // 自定义取值参考：https://docs.nlkit.com/nlux/reference/ui/ai-chat#chat-personas
        assistant: {          name: '博主的AI助手，十四行诗参上！',
          // AI 助手的图标
          avatar: 'https://chen-yulin.github.io/thumb/14.png',
          tagline: '要不要试试问下面的问题呢？',
        }
      }
    }
  };
</script>
<style>
  :root {
    /* webchat 工具栏的颜色 */
    --webchat-toolbar-background-color: #1464E4;
    /* webchat 工具栏文字和按钮的颜色 */
    --webchat-toolbar-text-color: #FFF;
  }
  /* webchat 对话框如果被遮挡，可以尝试通过 z-index、bottom、left 等设置来调整位置 */
  .webchat-container {
    z-index: 100;
    bottom: 10px;
    right: 10px;
  }
  /* webchat 的唤起按钮如果被遮挡，可以尝试通过 z-index、bottom、left 等设置来调整位置 */
  .webchat-bubble-tip {
    z-index: 99;
    bottom: 20px;
    right: 20px;
  }
  .webchat-bubble-tip {
    overflow: visible !important;
  }
  @keyframes float {
    0% {
      transform: translateY(0px) translateX(-50%);
    }
    50% {
      transform: translateY(-10px) translateX(-50%);
    }
    100% {
      transform: translateY(0px) translateX(-50%);
    }
  }

  .webchat-bubble-tip::before {
    content: '';
    position: absolute;
    top: -25px;
    left: 70%;
    width: 40px;
    height: 40px;
    background-image: url("data:image/svg+xml,%3Csvg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 24 24' fill='white'%3E%3Cpath d='M20 2H4c-1.1 0-2 .9-2 2v18l4-4h14c1.1 0 2-.9 2-2V4c0-1.1-.9-2-2-2zm0 14H6l-2 2V4h16v12z'/%3E%3C/svg%3E");
    background-repeat: no-repeat;
    background-position: center;
    background-size: contain;
    filter: drop-shadow(0 4px 6px rgba(0, 0, 0, 0.5));
    animation: float 3s ease-in-out infinite;
  }
</style>
);

最终效果：

将网站文件同步到数据库：阿里云数据管理
并且在知识索引中进行添加

Posted 2024-10-07Updated 2025-05-13Note11 minutes read (About 1651 words)

Neovim Cursor (avante.nvim)

n仓库地址: https://github.com/yetone/avante.nvim

Accounts

虚拟手机号: https://sms-activate.io/en
（学校邮箱，密码需要大写）

注册Claude-chat
授权Claude-api(https://console.anthropic.com/)

Apply Claude API

Installation

我使用lazy.nvim插件管理器所以非常方便，只需要在lazy_setup.lua里加上

{
  "yetone/avante.nvim",
  event = "VeryLazy",
  lazy = false,
  version = false, -- set this if you want to always pull the latest change
  opts = {
    -- add any opts here
  },
  -- if you want to build from source then do `make BUILD_FROM_SOURCE=true`
  build = "make",
  -- build = "powershell -ExecutionPolicy Bypass -File Build.ps1 -BuildFromSource false" -- for windows
  dependencies = {
    "nvim-treesitter/nvim-treesitter",
    "stevearc/dressing.nvim",
    "nvim-lua/plenary.nvim",
    "MunifTanjim/nui.nvim",
    --- The below dependencies are optional,
    "nvim-tree/nvim-web-devicons", -- or echasnovski/mini.icons
    "zbirenbaum/copilot.lua", -- for providers='copilot'
    {
      -- support for image pasting
      "HakonHarnes/img-clip.nvim",
      event = "VeryLazy",
      opts = {
        -- recommended settings
        default = {
          embed_image_as_base64 = false,
          prompt_for_file_name = false,
          drag_and_drop = {
            insert_mode = true,
          },
          -- required for Windows users
          use_absolute_path = true,
        },
      },
    },
    {
      -- Make sure to set this up properly if you have lazy=true
      'MeanderingProgrammer/render-markdown.nvim',
      opts = {
        file_types = { "markdown", "Avante" },
      },
      ft = { "markdown", "Avante" },
    },
  },
}

lazy安装完毕之后会有报错：
https://github.com/yetone/avante.nvim/issues/612
解决方案为

1 2	cd ~/.local/share/nvim/lazy/avante.nvim make BUILD_FROM_SOURCE=true

配置

添加~/.config/nvim/lua/avante.lua

---NOTE: user will be merged with defaults and
---we add a default var_accessor for this table to config values.

local Utils = require("avante.utils")

---@class avante.CoreConfig: avante.Config
local M = {}

---@class avante.Config
M.defaults = {
	debug = false,
	---@alias Provider "claude" | "openai" | "azure" | "gemini" | "cohere" | "copilot" | [string]
	provider = "claude", -- Only recommend using Claude
	auto_suggestions_provider = "claude",
	---@alias Tokenizer "tiktoken" | "hf"
	-- Used for counting tokens and encoding text.
	-- By default, we will use tiktoken.
	-- For most providers that we support we will determine this automatically.
	-- If you wish to use a given implementation, then you can override it here.
	tokenizer = "tiktoken",
	---@alias AvanteSystemPrompt string
	-- Default system prompt. Users can override this with their own prompt
	-- You can use `require('avante.config').override({system_prompt = "MY_SYSTEM_PROMPT"}) conditionally
	-- in your own autocmds to do it per directory, or that fit your needs.
	system_prompt = [[
Act as an expert software developer.
Always use best practices when coding.
Respect and use existing conventions, libraries, etc that are already present in the code base.
]],
	---@type AvanteSupportedProvider
	openai = {
		endpoint = "https://api.openai.com/v1",
		model = "gpt-4o",
		timeout = 30000, -- Timeout in milliseconds
		temperature = 0,
		max_tokens = 4096,
		["local"] = false,
	},
	---@type AvanteSupportedProvider
	copilot = {
		endpoint = "https://api.githubcopilot.com",
		model = "gpt-4o-2024-05-13",
		proxy = nil, -- [protocol://]host[:port] Use this proxy
		allow_insecure = false, -- Allow insecure server connections
		timeout = 30000, -- Timeout in milliseconds
		temperature = 0,
		max_tokens = 4096,
	},
	---@type AvanteAzureProvider
	azure = {
		endpoint = "", -- example: "https://<your-resource-name>.openai.azure.com"
		deployment = "", -- Azure deployment name (e.g., "gpt-4o", "my-gpt-4o-deployment")
		api_version = "2024-06-01",
		timeout = 30000, -- Timeout in milliseconds
		temperature = 0,
		max_tokens = 4096,
		["local"] = false,
	},
	---@type AvanteSupportedProvider
	claude = {
		endpoint = "https://api.anthropic.com",
		model = "claude-3-5-sonnet-20240620",
		timeout = 30000, -- Timeout in milliseconds
		temperature = 0,
		max_tokens = 8000,
		["local"] = false,
	},
	---@type AvanteSupportedProvider
	gemini = {
		endpoint = "https://generativelanguage.googleapis.com/v1beta/models",
		model = "gemini-1.5-flash-latest",
		timeout = 30000, -- Timeout in milliseconds
		temperature = 0,
		max_tokens = 4096,
		["local"] = false,
	},
	---@type AvanteSupportedProvider
	cohere = {
		endpoint = "https://api.cohere.com/v1",
		model = "command-r-plus-08-2024",
		timeout = 30000, -- Timeout in milliseconds
		temperature = 0,
		max_tokens = 4096,
		["local"] = false,
	},
	---To add support for custom provider, follow the format below
	---See https://github.com/yetone/avante.nvim/README.md#custom-providers for more details
	---@type {[string]: AvanteProvider}
	vendors = {},
	---Specify the behaviour of avante.nvim
	---1. auto_apply_diff_after_generation: Whether to automatically apply diff after LLM response.
	---                                     This would simulate similar behaviour to cursor. Default to false.
	---2. auto_set_keymaps                : Whether to automatically set the keymap for the current line. Default to true.
	---                                     Note that avante will safely set these keymap. See https://github.com/yetone/avante.nvim/wiki#keymaps-and-api-i-guess for more details.
	---3. auto_set_highlight_group        : Whether to automatically set the highlight group for the current line. Default to true.
	---4. support_paste_from_clipboard    : Whether to support pasting image from clipboard. This will be determined automatically based whether img-clip is available or not.
	behaviour = {
		auto_suggestions = false, -- Experimental stage
		auto_set_highlight_group = true,
		auto_set_keymaps = true,
		auto_apply_diff_after_generation = false,
		support_paste_from_clipboard = false,
	},
	history = {
		storage_path = vim.fn.stdpath("state") .. "/avante",
		paste = {
			extension = "png",
			filename = "pasted-%Y-%m-%d-%H-%M-%S",
		},
	},
	highlights = {
		---@type AvanteConflictHighlights
		diff = {
			current = "DiffText",
			incoming = "DiffAdd",
		},
	},
	mappings = {
		---@class AvanteConflictMappings
		diff = {
			ours = "co",
			theirs = "ct",
			all_theirs = "ca",
			both = "cb",
			cursor = "cc",
			next = "]x",
			prev = "[x",
		},
		suggestion = {
			accept = "<M-l>",
			next = "<M-]>",
			prev = "<M-[>",
			dismiss = "<C-]>",
		},
		jump = {
			next = "]]",
			prev = "[[",
		},
		submit = {
			normal = "<CR>",
			insert = "<C-s>",
		},
		-- NOTE: The following will be safely set by avante.nvim
		ask = "<leader>aa",
		edit = "<leader>ae",
		refresh = "<leader>ar",
		toggle = {
			default = "<leader>at",
			debug = "<leader>ad",
			hint = "<leader>ah",
			suggestion = "<leader>as",
		},
		sidebar = {
			switch_windows = "<Tab>",
			reverse_switch_windows = "<S-Tab>",
		},
	},
	windows = {
		---@alias AvantePosition "right" | "left" | "top" | "bottom"
		position = "right",
		wrap = true, -- similar to vim.o.wrap
		width = 30, -- default % based on available width in vertical layout
		height = 30, -- default % based on available height in horizontal layout
		sidebar_header = {
			align = "center", -- left, center, right for title
			rounded = true,
		},
		input = {
			prefix = "> ",
		},
		edit = {
			border = "rounded",
		},
	},
	--- @class AvanteConflictConfig
	diff = {
		autojump = true,
	},
	--- @class AvanteHintsConfig
	hints = {
		enabled = true,
	},
}

---@type avante.Config
M.options = {}

---@class avante.ConflictConfig: AvanteConflictConfig
---@field mappings AvanteConflictMappings
---@field highlights AvanteConflictHighlights
M.diff = {}

---@type Provider[]
M.providers = {}

---@param opts? avante.Config
function M.setup(opts)
	vim.validate({ opts = { opts, "table", true } })

	M.options = vim.tbl_deep_extend(
		"force",
		M.defaults,
		opts or {},
		---@type avante.Config
		{
			behaviour = {
				support_paste_from_clipboard = M.support_paste_image(),
			},
		}
	)
	M.providers = vim.iter(M.defaults)
		:filter(function(_, value)
			return type(value) == "table" and value.endpoint ~= nil
		end)
		:fold({}, function(acc, k)
			acc = vim.list_extend({}, acc)
			acc = vim.list_extend(acc, { k })
			return acc
		end)

	vim.validate({ provider = { M.options.provider, "string", false } })

	M.diff = vim.tbl_deep_extend(
		"force",
		{},
		M.options.diff,
		{ mappings = M.options.mappings.diff, highlights = M.options.highlights.diff }
	)

	if next(M.options.vendors) ~= nil then
		for k, v in pairs(M.options.vendors) do
			M.options.vendors[k] = type(v) == "function" and v() or v
		end
		vim.validate({ vendors = { M.options.vendors, "table", true } })
		M.providers = vim.list_extend(M.providers, vim.tbl_keys(M.options.vendors))
	end
end

---@param opts? avante.Config
function M.override(opts)
	vim.validate({ opts = { opts, "table", true } })

	M.options = vim.tbl_deep_extend("force", M.options, opts or {})
	M.diff = vim.tbl_deep_extend(
		"force",
		{},
		M.options.diff,
		{ mappings = M.options.mappings.diff, highlights = M.options.highlights.diff }
	)

	if next(M.options.vendors) ~= nil then
		for k, v in pairs(M.options.vendors) do
			M.options.vendors[k] = type(v) == "function" and v() or v
			if not vim.tbl_contains(M.providers, k) then
				M.providers = vim.list_extend(M.providers, { k })
			end
		end
		vim.validate({ vendors = { M.options.vendors, "table", true } })
	end
end

M = setmetatable(M, {
	__index = function(_, k)
		if M.options[k] then
			return M.options[k]
		end
	end,
})

M.support_paste_image = function()
	return Utils.has("img-clip.nvim") or Utils.has("img-clip")
end

M.get_window_width = function()
	return math.ceil(vim.o.columns * (M.windows.width / 100))
end

---@param provider Provider
---@return boolean
M.has_provider = function(provider)
	return M.options[provider] ~= nil or M.vendors[provider] ~= nil
end

---get supported providers
---@param provider Provider
---@return AvanteProviderFunctor
M.get_provider = function(provider)
	if M.options[provider] ~= nil then
		return vim.deepcopy(M.options[provider], true)
	elseif M.vendors[provider] ~= nil then
		return vim.deepcopy(M.vendors[provider], true)
	else
		error("Failed to find provider: " .. provider, 2)
	end
end

M.BASE_PROVIDER_KEYS = {
	"endpoint",
	"model",
	"deployment",
	"api_version",
	"proxy",
	"allow_insecure",
	"api_key_name",
	"timeout",
	-- internal
	"local",
	"_shellenv",
	"tokenizer_id",
	"use_xml_format",
}

return M

~/.config/nvim/init.lua中添加

1	require("plugin-config.avante").setup()

在~/.bashrc添加api key

1	export ANTHROPIC_API_KEY=*** #码了

Fix Error

Error #1

运行时出现这个报错

参见 https://github.com/yetone/avante.nvim/pull/560/files

由于这个PR没有被merge所以决定fork一下这个仓库再应用更改

push到远程仓库后记得release一下

然后将该plugin的仓库地址改为我自己的

成功消除了这个报错 > 现已切换回原仓库的main branch

Error #2

没有回复访问 https://console.anthropic.com/settings/plans 发现是没有Funds（说好的注册送5刀的额度的呢。。。）

使用WildCard服务 https://bewildcard.com/card ，具体账密见密码簿

Error #3

发现服务器无响应

设置终端proxy export https_proxy=127.0.0.1:7897
尝试输入

curl https://api.anthropic.com/v1/messages \                                                                                                   
  -H "x-api-key: <api-key>" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{
    "model": "claude-3-sonnet-20240229",
    "max_tokens": 1024,
    "messages": [
      {
        "role": "user",
        "content": "What is the capital of France?"
      }
    ]
  }'

sudo conda update curl to solve curl(77) error
export CURL_CA_BUNDLE=/etc/ssl/certs/ca-certificates.crt

Error #4

更新编译报错

ISSUE: https://github.com/yetone/avante.nvim/issues/612#issuecomment-2375729928

cd ~/.local/share/nvim/lazy/avante.nvim
make # or BUILD_FROM_SOURCE=true make
ls ./build/
avante_html2md.so  avante_repo_map.so  avante_templates.so  avante_tokenizers.so

Posted 2024-10-06Updated 2025-05-13Referencea minute read (About 130 words)

64型自测

此为MBTI64型的自测表格，使用方法为先在功能表内通过描述确定自己的理性和非理性功能（可以参考功能机制说明），然后在思维回路表中根据功能组合确定类型。
表格来自微博的大师@毛球好吃，是从网上的各种信息归纳总结的，感恩！

Posted 2024-09-29Updated 2025-05-13Reviewa minute read (About 185 words)

LSTM

背景

LSTM主要是用于解决递归网络中梯度指数级消失或者梯度爆炸的问题

https://www.youtube.com/watch?v=YCzL96nL7j0&t=267s
LSTM和RNN主要的区别就在于：LSTM有两条记忆链，一条短期记忆，一条长期记忆。

主要分成三个模块 - Forget Gate: 决定遗忘多少长期记忆 - Input Gate: 决定将多少当前输入存入长期记忆 - Output Gate: 基于短期记忆和输入决定输出的百分比，乘上长期记忆激活后的值，获得新的短期记忆，也就是输出。

这里gate的概念启发了grConv[[On the Properties of Neural Machine Translation= Encoder–Decoder Approaches]]

Posted 2024-09-27Updated 2025-05-13Review3 minutes read (About 376 words)

On the Properties of Neural Machine Translation= Encoder–Decoder Approaches

概要

对比了 RNN Encoder-Decoder 和 GRU(new proposed)之间的翻译能力，发现GRU更具优势且能够理解语法。

背景

RNN Encoder–Decoder

因为会把要翻译的语句映射到固定长度的vector所以训练需要的内存空间是固定的且很小，500M和几十G形成对比。
但也有问题：

As this approach is relatively new, there has not been much work on analyzing the properties and behavior of these models. For instance: What are the properties of sentences on which this approach performs better? How does the choice of source/target vocabulary affect the performance? In which cases does the neural machine translation fail?

不够Fancy的地方：

随着源句长度的增加，神经机器翻译模型的性能迅速下降。
词汇量的大小对翻译效果有很大的影响。

Encoder For Variable-Length Sequences

RNN

递归神经网络(RNN)在变长序列x = ( x1 , x2, … , xT)上通过保持隐藏状态h随时间变化而工作

grConv

这是本文提出的用于替换RNN Encoder-Decoder 中的Encoder的一种新的神经网络，文中称为：gated recursive convolutional neural network (grConv)

如图a为Recursive convolutional NN (这是啥？) #question 图b为grConv grConv则是让隐藏层通过训练w参数可以从三个输入中挑选：

其中 $\omega_c+\omega_l+\omega_r=1$ 由此便获得了如图c,d所示的自主学习语法结构的能力。

非常直观的图 #paradigm