The goal of this project is to create a simple, easy-to-use format for generating and transmitting character files. You can use these character files out of the box with Eliza or other LLM agents.
- Open Terminal. On Mac, you can press Command + Spacebar and search for "Terminal". If you're using Windows, use WSL2
- Type
npx tweets2characterand run it. If you get an error about npx not existing, you'll need to install Node.js - If you need to install node, you can do that by pasting
curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.40.1/install.sh | bashinto your terminal to install Node Version Manager (nvm) - Once that runs, make a new terminal window (the old one will not have the new software linked) and run
nvm install nodefollowed bynvm use node - Now copy and paste
npx tweets2characterinto your terminal again. - NOTE: You will need to get a Claude or OpenAI API key. Paste that in when prompted
- You will need to get the path of your Twitter archive. If it's in your Downloads folder on a Mac, that's ~/Downloads/.zip
- If everything is correct, you'll see a loading bar as the script processes your tweets and generates a character file. This will be output at character.json in the directory where you run
npx tweets2character. If you run the commandcdin the terminal before or after generating the file, you should see where you are.
The JSON schema for the character file is here. This also matches the expected format for OpenAI function calling.
Typescript types for the character file are here.
Basic example of a character file, with values that are instructional examples/example.character.json
Read the example character file and print the contents examples/example.py
Read the example character file and validate it against the JSON schema examples/validate.py
Read the example character file and print the contents examples/example.mjs
Read the example character file and validate it against the JSON schema examples/validate.mjs
You can use the scripts the generate a character file from your tweets, convert a folder of documents into a knowledge file, and add knowledge to your character file.
Most of these scripts require an OpenAI or Anthropic API key.
Convert your twitter archive into a .character.json
First, download your Twitter archive here: https://help.x.com/en/managing-your-account/how-to-download-your-x-archive
You can run tweets2character directly from your command line with no downloads:
npx tweets2characterNote: you will need node.js installed. The easiest way is with nvm.
Then clone this repo and run these commands:
npm install
node scripts/tweets2character.js twitter-2024-07-22-aed6e84e05e7976f87480bc36686bd0fdfb3c96818c2eff2cebc4820477f4da3.zip # path to your zip archiveNote that the arguments are optional and will be prompted for if not provided.
Convert a folder of images and videos into a .knowledge file which you can use with Eliza. Will convert text, markdown and PDF into normalized text in JSON format.
You can run folder2knowledge directly from your command line with no downloads:
npx folder2knowledge <path/to/folder>npm install
node scripts/folder2knowledge.js path/to/folder # path to your folderNote that the arguments are optional and will be prompted for if not provided.
Add knowledge to your .character file from a generated knowledge.json file.
You can run knowledge2character directly from your command line with no downloads:
npx knowledge2character <path/to/character.character> <path/to/knowledge.knowledge>npm install
node scripts/knowledge2character.js path/to/character.character path/to/knowledge.knowledge # path to your character file and knowledge fileNote that the arguments are optional and will be prompted for if not provided.
The license is the MIT license, with slight modifications so that users are not required to include the full license in their own software. See LICENSE for more details.