Skip to content

Latest commit

 

History

History
 
 

docs

README.md

Eliza - Multi-agent simulation framework

https://github.com/elizaOS/eliza

Visit https://eliza.builders for support

🌍 README Translations

中文说明 | Deutsch | Français | ไทย | Español

dev branch

Eliza Banner

As seen powering @DegenSpartanAI and @MarcAIndreessen

  • Multi-agent simulation framework
  • Add as many unique characters as you want with characterfile
  • Full-featured Discord and Twitter connectors, with Discord voice channel support
  • Full conversational and document RAG memory
  • Can read links and PDFs, transcribe audio and videos, summarize conversations, and more
  • Highly extensible - create your own actions and clients to extend Eliza's capabilities
  • Supports open source and local models (default configured with Nous Hermes Llama 3.1B)
  • Supports OpenAI for cloud inference on a light-weight device
  • "Ask Claude" mode for calling Claude on more complex queries
  • 100% Typescript

Getting Started

Prerequisites (MUST):

Edit the .env file

  • Copy .env.example to .env and fill in the appropriate values
  • Edit the TWITTER environment variables to add your bot's username and password

Edit the character file

  • Check out the file src/core/defaultCharacter.ts - you can modify this
  • You can also load characters with the pnpm start --characters="path/to/your/character.json" and run multiple bots at the same time.

After setting up the .env file and character file, you can start the bot with the following command:

pnpm i
pnpm start

Customising Eliza

Adding custom actions

To avoid git clashes in the core directory, we recommend adding custom actions to a custom_actions directory and then adding them to the elizaConfig.yaml file. See the elizaConfig.example.yaml file for an example.

Running with different models

Run with Llama

You can run Llama 70B or 405B models by setting the environment variable for a provider that supports these models. Llama is also supported locally if no other provider is set.

Run with Grok

You can run Grok models by setting the GROK_API_KEY environment variable to your Grok API key and setting grok as the model provider in your character file.

Run with OpenAI

You can run OpenAI models by setting the OPENAI_API_KEY environment variable to your OpenAI API key and setting openai as the model provider in your character file.

Additional Requirements

You may need to install Sharp. If you see an error when starting up, try installing it with the following command:

pnpm install --include=optional sharp

Environment Setup

You will need to add environment variables to your .env file to connect to various platforms:

# Required environment variables
DISCORD_APPLICATION_ID=
DISCORD_API_TOKEN= # Bot token
OPENAI_API_KEY=sk-* # OpenAI API key, starting with sk-
ELEVENLABS_XI_API_KEY= # API key from elevenlabs

# ELEVENLABS SETTINGS
ELEVENLABS_MODEL_ID=eleven_multilingual_v2
ELEVENLABS_VOICE_ID=21m00Tcm4TlvDq8ikWAM
ELEVENLABS_VOICE_STABILITY=0.5
ELEVENLABS_VOICE_SIMILARITY_BOOST=0.9
ELEVENLABS_VOICE_STYLE=0.66
ELEVENLABS_VOICE_USE_SPEAKER_BOOST=false
ELEVENLABS_OPTIMIZE_STREAMING_LATENCY=4
ELEVENLABS_OUTPUT_FORMAT=pcm_16000

TWITTER_DRY_RUN=false
TWITTER_USERNAME= # Account username
TWITTER_PASSWORD= # Account password
TWITTER_EMAIL= # Account email


# For asking Claude stuff
ANTHROPIC_API_KEY=

WALLET_SECRET_KEY=EXAMPLE_WALLET_SECRET_KEY
WALLET_PUBLIC_KEY=EXAMPLE_WALLET_PUBLIC_KEY

BIRDEYE_API_KEY=

SOL_ADDRESS=So11111111111111111111111111111111111111112
SLIPPAGE=1
SOLANA_RPC_URL=https://api.mainnet-beta.solana.com
HELIUS_API_KEY=


## Telegram
TELEGRAM_BOT_TOKEN=

TOGETHER_API_KEY=

Local Inference Setup

CUDA Setup

If you have an NVIDIA GPU, you can install CUDA to speed up local inference dramatically.

pnpm install
npx --no node-llama-cpp source download --gpu cuda

Make sure that you've installed the CUDA Toolkit, including cuDNN and cuBLAS.

Running locally

By default, the bot will download and use a local model. You can change this by setting the environment variables for the model you want to use.

Clients

Discord Bot

For help with setting up your Discord Bot, check out here: https://discordjs.guide/preparations/setting-up-a-bot-application.html

Development

Testing

To run the test suite:

pnpm test           # Run tests once
pnpm test:watch    # Run tests in watch mode

For database-specific tests:

pnpm test:sqlite   # Run tests with SQLite
pnpm test:sqljs    # Run tests with SQL.js

Tests are written using Jest and can be found in src/**/*.test.ts files. The test environment is configured to:

  • Load environment variables from .env.test
  • Use a 2-minute timeout for long-running tests
  • Support ESM modules
  • Run tests in sequence (--runInBand)

To create new tests, add a .test.ts file adjacent to the code you're testing.

Docs Updates

Please make sure to verify if the documentation provided is correct. In order to do so, please run the docs service.

docker compose -f docker-compose-docs.yaml up --build

The docusaurus server will get started and you can verify it locally at https://localhost:3000/eliza.