Skip to content

R&D playground to play with agents and OpenBB

Notifications You must be signed in to change notification settings

mmistroni/openbb-agents

 
 

Repository files navigation

OpenBB LLM Agents

Work-in-progress.

This is a project that leverages LLMs and OpenBB Platform to create financial analyst agents that can autonomously perform financial research, and answer questions using up-to-date data. This is possible as a result of agents utilizing function calling to interact with the OpenBB Platform.

Installation

Currently, we support Python 3.10+. We will be adding support for more version of Python relatively soon.

openbb-agents is available as a PyPI package:

pip install openbb-agents --upgrade

Setup

OpenAI API keys

To use OpenBB LLM Agents, you need an OpenAI API key. Follow these steps:

  1. Get API Key: Sign up on OpenAI and get your API key.
  2. Set Environment Variable: Add this to your shell profile (.bashrc, .zshrc, etc.):
    export OPENAI_API_KEY="your_openai_api_key"

OpenBB Platform data provider credentials

To use the OpenBB Platform functions, you need to configure the necessary data provider API credentials. This can be done in one of two ways:

  1. Local Configuration: Specify your credentials in a ~/.openbb_platform/user_settings.json file. Follow the local environment setup guide for detailed instructions.
  2. OpenBB Hub: Create a personal access token (PAT) via your OpenBB Hub account. This PAT can then be passed to the agent as an argument.

Getting started

It is highly recommended to take a look at the Getting Started Notebook, which runs you through the features of openbb-agents at a high level.

Usage

>>> from openbb_agents.agent import openbb_agent
>>> result = openbb_agent("What is the current market cap of TSLA?")  # Will print some logs to show you progress
>>> print(result)
- The current market cap of TSLA (Tesla, Inc.) is approximately $695,833,798,800.00.
- This figure is based on the most recent data available, which is from January 15, 2024.
- The market cap is calculated by multiplying the current stock price ($218.89) by the number of outstanding shares (3,178,920,000).

To use your data provider credentials stored in OpenBB Hub, you can pass in your OpenBB Hub PAT directly to the agent:

>>> openbb_agent("What is the stock price of AAPL?", openbb_pat="<openbb-hub-pat>")

Note: The agent dynamically configures itself based on the available data provider credentials. Consequently, certain data sources and functions may be inaccessible without the appropriate API key. By default, yfinance is included as a data provider and does not require an API key. For a comprehensive list of functions and their supported data providers, refer to the OpenBB Platform documentation.

Queries can be relatively complex:

>>> openbb_agent("Perform a fundamentals financial analysis of AMZN using the most recently available data. What do you find that's interesting?")

Queries can also have temporal dependencies (i.e the answers of previous subquestions are required to answer a later subquestion):

>>> openbb_agent("Who are TSLA's peers? What is their respective market cap? Return the results in _descending_ order of market cap.")

An async variant of the agent is also available:

>>> from openbb_agents.agent import aopenbb_agent
>>> await aopenbb_agent("What is the current market cap of TSLA?")

Development

  • Create a new virtual environment, with poetry
  • poetry install

Linting and Formatting

We're currently experimenting with ruff as a drop-in replacement for black, isort and pylint.

You can run linting checks as follows:

ruff check

Or fix linting errors:

ruff check --fix

Or format the code:

ruff format

We've also included these in the pre-commit, if you'd prefer to have these checks run automatically before commiting code. You can install the pre-commit hooks as follows:

pre-commit install

Testing

We use pytest as our test runner:

pytest -n 8 tests/

About

R&D playground to play with agents and OpenBB

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 66.6%
  • Jupyter Notebook 33.4%