“It's not a real train, Morty. It's a story device. Literally. A literal literary device quite literally metaphorically containing us”.
This guide outlines the steps to set up a local development environment for an AI agent that interacts with Warpcast, the Graph, and DAOs on EVM chains.
- Virtual Environment (VM): Ensure you have Python and a virtual environment tool installed.
- Poetry: Dependency management tool for Python.
- API Keys:
- OpenAI Pro account https://openai.com/index/openai-api/
- New Warpcast account for your agent
- NANAR API account https://dev.neynar.com/
- The Graph API account https://thegraph.com/studio
- RPC: Use a local node or a service like Infura, Alchemy or a free/public RPC for EVM interactions.
git clone <repository-url>
cd <repository-folder>python3 -m venv my-venv
source my-venv/bin/activate # For Linux/Macpip3 install poetry
poetry installThis installs libraries for interacting with OpenAI, and other required tools.
pip install -e .This installs the dao-agents CLI binary. You can check the available commands with dao-agents --help
$ dao-agents --help
Usage: dao-agents [OPTIONS] COMMAND [ARGS]...
DAO Agents Simulation CLI
Options:
--help Show this message and exit.
Commands:
auto Run an autonomous simulation with the DAO Agent
chat Run a chat with the DAO Agent
create-wallet Create a set of wallet for the DAO Agents
run-simulation Run a full multi-agent dao simulation session using a...
two-agent Run a two-agent simulation with the DAO AgentCreate a .env file and fill in the following keys:
OPENAI_API_KEYcreate key at https://openai.com/index/openai-api/FARCASTER_FIDcan get this from the api page https://docs.neynar.com/reference/lookup-user-by-usernameFARCASTER_CHANNEL_IDthis is the current default channel for testing (optional)NANAR_API_KEYcan get key at https://dev.neynar.com/NAYNAR_SIGNER_UUIDneed to create this from the naynar dev dashboardGRAPH_KEYcan get from https://thegraph.com/studioWEB3_PROVIDER_URITARGET_CHAINchain id of the target EVM chain
agent wallet data (can use create_wallet.py for a new one)
AGENT_MNEMONICAGENT_PRIVATE_KEYAGENT_ADDR
If working with a DAO for voting and proposals (currently on a single dao is supported)
TARGET_DAO
currently using imgbb for more persistent image hosting
IMG_BB_API_KEY
Refer to the documentation of respective services to generate these keys.
cd into dao-agent-demo Run the wallet creation script to set up a new wallet if you didn't bring your own:
dao-agents create-walletThis generates a new account and mnemonic and stores these values to the .env file.
If interacting on-chain, fund the wallets listed in the .env file with a small amount of eth for gas fees.
Make sure you are in the root of the repo and run the cli command:
dao-agents run-simulation --world-definition <world-definition-file>or to load a character
dao-agents chat --character-file <character-file-json>Options available:
- Chat Mode: Directly chat with the agent for tasks like generating proposals or interacting with DAOs.
- Autonomous Mode: The agent operates autonomously, performing actions like replying on Warpcast, creating proposals, or notifying about updates.
- Two-agent demo: demo of 2 agents simulating a conversation
- DAO simulator: pick from a list of world simulations
Modify the characters folder to define:
- Identity and initial prompt in JSON files.
- Autonomous thoughts for periodic actions.
agent.py: Core functions for the agent.constants_utils.py: Contract addresses and configurations.helpers and utils: Includes DAO summoning tools, Warpcast, and Graph, json store utility wrappers.run.py: Handles agent initialization and interval control for autonomous actions.characters/: json files that define initial prompts and auto thoughts for agentsknowledge/: markdown files in this folder can be loaded into the knowledge base with import_knowledge.py script (file name should contain keywords ex: speedball_fair_token_launch.md)operators/: json files that define the operator agentworlds/: directory for world def files that seed a simulation
- Intervals: The autonomous mode executes random actions every 5 to 60 minutes by default. This can be adjusted in
run.py. - Memory Management: There is a tinydb json store for committing memories, use this to avoid repetitive tasks
- Knowledge: You can put markdown files in the knowledge folder and run
import_knowledge.pyto add it to the db - Create New Simulation: You can create a new simulation and all the config files needed with a script
create_sim.pyit just asks for a prompt and handles the rest.
For detailed configuration or additional features, refer to the helper files and modify as needed.
this is a repo to experiment on stuff, not for production use
This is a simple lightweight framework to interact with agents using openai swarm Documentation. There are several example functions to summon, interact and retrieve info from DAOs. There are farcaster tools to cast and retrieve casts. This was originally based on the Based Agent playground Repo
- Autonomous execution: The agent thinks, decides, and acts onchain autonomously.
- Token deployement: Creates a presale token with fee ditributions to refferers
- Balance checks: Keep tabs on wallet balances.
- Crowd fund deployment: Can deploy a yeeter crowdfund.
- Art generation via DALL-E: Generate artwork using AI.
- Whatever you want: Add in features and share them with us!
- Python 3.10+
NAYNAR farcaster api Sign up for a Naynar api key https://neynar.com/
You will need to use their dash board interface to "Make a Bot" to get a signer key
Here are a few example endpoints
- Get conversation https://docs.neynar.com/reference/lookup-cast-conversation
- Publish cast https://docs.neynar.com/reference/publish-cast
Create a new python venv. You will need pip and poetry installed
You can edit the character by creating your own character json file in the /characters folder
Unlock a world of possibilities with these built-in functions:
summon_daovote_on_dao_proposalsubmit_proposaget_dao_proposalsget_dao_proposalget_dao_proposals_count
cast_to_farcaster(content: str)check_cast_replies()check_cast_notifications()mark_notifications_as_seen()cast_reply(content: str, parentHash: str, parent_fid: int)check_recent_agent_casts()check_recent_user_casts(fid: str)check_user_profile(fid: str)
generate_art(prompt): Generate art using DALL-E based on a text prompt.get_agent_address: get the address of the current agentcommit_memory: store a memory for long term
All of the functionality for the DAO Agent resides within agents.py. This is the central hub where you can add new capabilities, allowing the agent to perform a wide range of tasks.
By incorporating additional libraries, you can extend the agent's reach beyond blockchain interactions to include Web2 functionalities, such as posting on Warpcast, etc.
DAO Agent uses:
- OpenAI Swarm: Powers the agent's autonomous decision-making.
This project is for educational purposes only. Do not use with real assets or in production environments. Always exercise caution when interacting with blockchain technologies.