braineo/fate-emacs

Table of Contents

Get Started
External tools
- ripgrep
- fd
- enchant
- Python Tools
- Go Tools
- JavaScript/JSON Tools
- Shell Tools
- nerd-icons
Customize
- Custom.el
- EditorConfig (Code style)
Install and Update Language Servers
Vterm
GPT
- Install CUDA
- llama.cpp
  - Build
  - Run
- Ollama
Useful links

Get Started

Only tested with >= Emacs 28.2 in Debian and macOS

Backup

backup your .emacs and .emacs.d, rename them to something else because emacs loads .emacs first if it exists

Install

git clone https://github.com/braineo/fate-emacs.git ~/.emacs.d

Try it in docker

docker pull braineo/fate-emacs
# run in terminal
docker run -it --rm braineo/fate-emacs
# run with GUI
xhost +local:root # WARN: this comes with security issues
docker run -it --rm -e DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix -v /your/workspace/:/mnt/workspace braineo/fate-emacs

External tools

ripgrep

ripgrep is basically a fast grep written in Rust.

Download ripgrep release from https://github.com/BurntSushi/ripgrep/releases

# example
wget https://github.com/BurntSushi/ripgrep/releases/download/0.10.0/ripgrep_0.10.0_amd64.deb && dpkg -i ripgrep_0.10.0_amd64.deb
# or
apt install ripgrep

for masOS

brew install ripgrep

fd

fd is a replacement for find

sudo apt install fd-find

enchant

enchant is a library for spell checking

## Ubuntu
sudo apt install libenchant-2-dev pkg-config

### mac
brew install enchant pkg-config

Python Tools

M-x fate/python-install-tools

Go Tools

M-x fate/go-install-tools

JavaScript/JSON Tools

Name	Installation command	Description
prettier	npm i -g prettier	Formater for JavaScript, TypeScript, CSS, JSON, and more
linter LSP	npm i -g vscode-langservers-extracted	Language server of eslint, HTML, CSS
TypeScript LSP	npm i -g typescript-language-server

or M-x fate/js-install-tools

Shell Tools

Name	Installation command	Description
shellcheck	apt install shellcheck	A great teacher helping you write better shell script
shfmt	go install mvdan.cc/sh/v3/cmd/shfmt@latest	A shell script formatter

nerd-icons

Install fonts for mode-line

M-x nerd-icons-install-fonts

Customize

Custom.el

The first time you open Emacs, it will create a custom.el for you automatically from fate-custom-template.el. There you can made custom configuration does not go with Emacs for different computers (like your personal ones and working ones). For example, PATH variable.

Especially for macOS users, because exec-path-from-shell is slow that I did not include it and you might want to edit ENV variables and exec-path

exec-path

(setq exec-path (append exec-path '("/path/to/bin")))

Setting ENV variables

(setenv "LD_LIBRARY_PATH"
  (let ((current (getenv "LD_LIBRARY_PATH"))
        (new "/path/to/lib/"))
    (if current (concat new ":" current) new)))

Custom variable for packages

(setq lsp-python-ms-extra-paths
  '("/path/to/a/site-packages"
    "/path/to/b/site-packages"))

EditorConfig (Code style)

EditorConfig helps maintain consistent coding styles for multiple developers working on the same project across various editors and IDEs. The EditorConfig project consists of a file format for defining coding styles and a collection of text editor plugins that enable editors to read the file format and adhere to defined styles. EditorConfig files are easily readable and they work nicely with version control systems.

wget https://github.com/braineo/configs/blob/master/editorconfig/.editorconfig ~/.editorconfig

Optional configuration, take a look at my configs if interested

Install and Update Language Servers

Previously lsp-mode was used as LSP client. The configuration now migrates to lsp-bridge, a client emphasizes performance. It has not covered all the features in lsp-mode, but it is really fast.

# install dependencies for lsp-bridge
pip3 install epc orjson sexpdata six setuptools paramiko rapidfuzz

For each language server, refer to the configuration guide in lsp-bridge

Language servers implemented in NodeJS

Language servers implemented in NodeJS can obtain directly by doing lsp-install-server.

Language	Installation command
TypeScript/JavaScript	npm i -g typescript-language-server; npm i -g typescript
JSON	npm i -g vscode-json-languageserver
Dockerfile	npm install -g dockerfile-language-server-nodejs

Python

Microsoft Python Language Server (depreciated)

lsp-python-ms now downloads language server binary automatically, or you can build latest Microsoft Python Language Server by yourself

Install dotnet-sdk
Clone and install python-language-server

git clone https://github.com/Microsoft/python-language-server.git
cd python-language-server/src/LanguageServer/Impl
dotnet build -c Release
dotnet publish -c Release -r linux-x64 # linux
dotnet publish -c Release -r osx-x64   # mac

change the value of the -r flag depending on your architecture and operating system. See Microsoft’s Runtime ID Catalog for the right value for your system.

Then, link the executable to somewhere on your path, e.g.

ln -sf $(git rev-parse --show-toplevel)/output/bin/Release/linux-x64/publish/Microsoft.Python.LanguageServer ~/.emacs.d/mspyls/
ln -sf $(git rev-parse --show-toplevel)/output/bin/Release/osx-x64/publish/Microsoft.Python.LanguageServer ~/.emacs.d/mspyls/

Credit to lsp-python-ms

Because Microsoft’s Python language server does not ship with a linter, need to install flake8

pip install flake8

;; Configure extra search path
(setq lsp-python-ms-extra-paths
  '("path1"
    "path2"))

;; If you need to work on some python2 projects, make sure use a right flake8
(setq flycheck-python-flake8-executable "python2")

Clangd

clangd needs compile_commands.json to know the compiler flags that are used to build the project, when some custom paths are used.

For cmake project, passing -DCMAKE_EXPORT_COMPILE_COMMANDS=ON to cmake can generate compile_commands.json. If it is not a cmake project, bear is a right tool to generate it.

Pitfalls for compile_commands.json. By default CMAKE_CXX_IMPLICIT_INCLUDE_DIRECTORIES are not exported into compile_commands.json. With compile_commands.json clangd might still not able to figure out the include paths. In that case you can try adding following lines to the beginning of CMakeLists.txt. See cmake issue

if(CMAKE_EXPORT_COMPILE_COMMANDS)
    set(CMAKE_CXX_STANDARD_INCLUDE_DIRECTORIES ${CMAKE_CXX_IMPLICIT_INCLUDE_DIRECTORIES})
endif()

But again, if the project is using gcc instead of clang, you might see some errors like this from clangd. That's because the previous lines add -isystem /usr/lib/gcc/x86_64-linux-gnu/14/include and it is not compatible with clang. You can manually remove it and fine tune the CMake.

E[12:12:01.769] [builtin_definition] Line 18: in included file: definition of builtin function '_mm_getcsr'

To manually verify if clangd works. You can run

clangd --compile-commands-dir="."  --query-driver="clang-tidy" --check="path/to/some.cpp"

Vterm

To make vterm handy, you need some shell-side configuration, add following to .zshrc or .bashrc

if [[ "$INSIDE_EMACS" = 'vterm' ]] \
    && [[ -n ${EMACS_VTERM_PATH} ]] \
    && [[ -f ${EMACS_VTERM_PATH}/etc/emacs-vterm-zsh.sh ]]; then
	source ${EMACS_VTERM_PATH}/etc/emacs-vterm-zsh.sh
fi

GPT

You can choose to use OpenAI, if you want to keep the prompt to yourself. Hosting a local LLM is not difficult either.

Install CUDA

follow guide in nvidia site.

TL;DR

wget https://developer.download.nvidia.com/compute/cuda/repos/debian12/x86_64//cuda-keyring_1.1-1_all.deb
sudo dpkg -i cuda-keyring_1.1-1_all.deb
sudo apt-get install -y nvidia-kernel-open-dkms
sudo apt-get install -y cuda-drivers
sudo apt install nvidia-cuda-toolkit

llama.cpp

Build

git clone git@github.com:ggerganov/llama.cpp.git

mkdir build
cd build
cmake .. -DLLAMA_CUBLAS=ON
cmake --build . --config Release

Run

Run following command in build folder, api server will spin up at localhost:8080

./bin/server --model /path/to/models/dolphin-2.6-mistral-7b-dpo-laser.Q5_K_M.gguf --n-gpu-layers 400

Ollama

Or use Ollama as frontend

curl -fsSL https://ollama.com/install.sh | sh
ollama run dolphin-mixtral

braineo / fate-emacs