repository·main·Indexed 26 days ago
https://github.com/the-swarm-corporation/autohedgeAn autonomous agent hedge fund designed for trading in the Solana ecosystem. AutoHedge uses a multi-agent architecture consisting of Director, Quant, Risk Management, and Execution agents to perform market analysis, risk assessment, and order execution. It provides a programmatic interface via the AutoHedge class and an interactive CLI REPL for executing trading tasks through natural language prompts.
AutoHedge operates using a specialized multi-agent pipeline where each agent handles a specific stage of the trading lifecycle:
AutoHedge requires several environment variables for API access and trading. Set these in your environment or a .env file.
Required for token prices and search tools. Obtain a key at https://portal.jup.ag.
JUPITER_API_KEYRequired for experimental agent functionality (OpenAI/Anthropic).
OPENAI_API_KEYANTHROPIC_API_KEYWORKSPACE_DIR: The directory used for agent workspaces (e.g., "agent_workspace").Required for executing trades on Solana.
WALLET_PRIVATE_KEY: Your Solana wallet private key.# Jupiter API (token price & search tools)
JUPITER_API_KEY=
# OpenAI (experimental agents)
OPENAI_API_KEY=
ANTHROPIC_API_KEY=
WORKSPACE_DIR="agent_workspace"
# Trading
WALLET_PRIVATE_KEY=""autohedge command in your terminal.autohedge Install the latest version of the AutoHedge package using pip.
pip install -U autohedgeThe AutoHedge CLI provides an interactive REPL (Read-Eval-Print Loop) for executing research and hedging tasks via natural language prompts.
To start the interactive session, run the autohedge command from your terminal. Once inside the REPL, you can enter task prompts directly.
REPL Commands:
<task>: Enter a natural language instruction (e.g., Analyze NVDA for 50k allocation) to execute a task.help, ?, or h: Display usage tips.quit, exit, or q: Exit the REPL session.When initializing AutoHedge, you can specify the output_type to determine the format of the returned trading logs. Supported types are:
list: Returns messages as a list (default).dict: Returns messages as a dictionary.str: Returns the conversation history as a single string.AutoHedge class is the main entry point for the trading system. It coordinates agents and manages the trading cycle. You can configure the system's name, description, and how output data is stored and formatted.require_openai_key() to check if the OPENAI_API_KEY environment variable is set. This is a requirement for using Swarm's gpt-4.x agents. The function returns True if the key is present and False otherwise.If you are building on top of AutoHedge rather than using the CLI, you can import the AutoHedge class to execute tasks programmatically.
Use the .run(task=...) method to pass a string task prompt. The method returns the result of the executed task.
load_env() to automatically locate and load a .env file. The function searches for a .env file by walking up the directory tree from the current working directory until it finds one (the project root approach). If a .env file is found in the project root, it is loaded without overriding existing environment variables. If no project-root .env is found, it falls back to loading from the current working directory. This ensures API keys are available even when running scripts from subdirectories.find_project_env() function returns the Path to the .env file found by walking up from the current working directory. If no .env file is found in any parent directory, it returns None. This is useful if you need to explicitly reference the path of the configuration file.AutoHedge class is the primary entrypoint for the library. Upon importing the package, load_env() is automatically called to load necessary environment variables. You can instantiate AutoHedge to begin managing automated hedging operations.