OctoBot AI Hedge Fund

Multi Agent Crypto AI Hedge Fund

OctoBot AI Hedge Fund is a practical & real-world approach to AI multi-agent driven crypto investment strategies.

⚠️ Work in Progress: This repository is currently under development and not yet ready for production use.

It’s a ready-made framework to:

1. Code or re-use specialized analysis, strategy or trading agents to build your AI hedge fund team.
2. Test and optimize your hedge fund using historical data and backtesting.
3. Launch your hedge fund for continuous live analysis and real or paper trading using OctoBot.

This AI multi-agent system designed to replicate traditional hedge funds thinking by leveraging a team of expert AI agents.

1. Analyst agents use dedicated input data, and outputs their analysis based on their knowledge
2. Strategy agents receive analysts’ outputs and apply their own investment strategy to conclude on ideal states
3. Trading agents receive ideal state updates and decide whether or not to take action based on their configuration

The whole AI Hedge Fund is a self-improving system where agents learn from their past experience and are trained with historical and live data.

Available agents

Here are the currently ready-to-use agents built into the framework. Feel free to create your own or improve the existing ones. New agents are added on a regular basis.

Data Analysis Agents

• Social Networks Agent
• Technical Analysis Agent

Strategy Agents

• Strategy Manager Agent

Trading Agents

• Signal Agent
• Bullish Agent
• Bearish Agent
• Judge Agent
• Risk Agent
• Distribution Agent

Using OctoBot AI Hedge Fund

Python installation

pip install octobot-ai

Docker installation

docker build -t octobot-ai .

Quick start

This part is in WIP and will soon be updated.

A state of the art & research driven agentic portfolio management

OctoBot AI Hedge Fund is created as a practical and research driven approach to offer the best ways to leverage AI to make data-driven investment decisions.

This framework is designed to go beyond typical market data analysis and global market sentiment investment to:

1. Consider signals from any source.
   Here, think market price, newspapers, social network feeds, internet browsing or anything either maths or LLMs (or a combination of both) can handle

2. Gather signals and applying them to a portfolio according to well known investment frameworks and target risk levels. Agents can be configured to trade following famous investors strategies such as Warren Buffett’s value investing, Michael Burry’s Contrarian Investment or Cathie Wood’s Growth investing.

3. Take decisions using its analysis and its past experience in a self-learning virtuous cycle as highlighted in this Orchestration Framework for Financial Agents research paper.

Avoiding common multi-agent LLM pitfalls

There are many reasons why multi-agent LLM systems fail in the context of finance and investment and OctoBot AI Hedge Fund is built to avoid them. Here are the most common ones and how the framework handles it.

1. Poor Specification & System Design: ambiguous roles and tasks = Agents disobey tasks or ignore instructions.
   Using langchain and a custom framework to accurately define the role of each agent solves this problem. Other solutions have been explored in this research: AlphaAgents: Large Language Model based Multi-Agents for Equity Portfolio Constructions.

2. Communication Breakdowns: Natural-language = unstructured and probabilistic = Ignoring input & withholding info.
   Enforcing formatted and standardized communication between agents largely suppresses this risk. Solutions are explored in this research: When Machines Meet Each Other: Network Effects and the Strategic Role of History in Multi-Agent AI.

3. The Trust-Vulnerability Paradox: more agents means = better performance but also a massive increase cascading failure risk.
   Using agent memory to constantly learn from failures and successes paired with a dynamic trust score makes the framework resilient to the Trust-Vulnerability Paradox. This paradox is explored in detail in The Trust Paradox in LLM-Based Multi-Agent Systems: When Collaboration Becomes a Security Vulnerability.

License

GNU General Public License v3.0 or later.

See GPL-3.0 LICENSE to see the full text.

Give a boost to OctoBot AI Hedge Fund

Do you like what we are building with OctoBot AI Hedge Fund? Consider giving us a star ⭐ to boost the project's visibility!