gitricko / phrasing-bot

A bot that checks your grammar and phrasing using LLM of choice

Phrasing Bot

A web interface bot to check and improve grammar and phrasing of your input, similar to Quillbot or Grammarly.

Use any backend LLM service of your choice, even locally hosted ones (check "Installation" for more)!

Requirements

• node version 20.* (check .nvmrc)
• yarn

Installation

• Copy env.example to .env
  • Change REQUEST_TYPE to one of these: custom, groq, openrouter, openai
  • For groq, openrouter, openai, make sure you fill in the env variables with corresponding prefix
  • For custom
    • Run your own OpenAI-compatible server (llama.cpp, llamafile, Koboldcpp etc) and make sure it is accessible
    • make sure you fill in the env variables with CUSTOM_ prefix
    • CUSTOM_API_BASE_URL should usually end with /v1
    • CUSTOM_API_KEY or CUSTOM_API_MODEL can usually be left empty
    • For CUSTOM_API_PROMPT_TEMPLATE, check available types in src/prompt-template.ts, so like MISTRAL, LLAMA_3 etc

Build frontend cd frontend && yarn && yarn build

Build backend yarn && yarn build

Running

Start server yarn start

The server should now be running on http://localhost:5000/

Development

Bit yanky, need to run the frontend and backend separately

Run frontend cd frontend && yarn && yarn dev

In separate window, run backend yarn && yarn dev

You should now have frontend running on http://localhost:5173/ and backend on http://localhost:5000/ with live reloading etc.