agent0 0.22.0

Agent interface for on-chain protocols.

A Framework for Hyperdrive Trading Strategies

This repo by DELV contains tools for you to deploy automated trading agents, perform market simulations, and conduct trading research. Read on for more info or jump to the quickstart guide.

Hyperdrive Background

Hyperdrive is an automated market maker (AMM) protocol that enables fixed-rate markets to be built on top of arbitrary yield sources. It deploys assets into those yield sources and wraps them as Hyperdrive positions represented as hy[Tokens] that trade at a discount (%) and can be redeemed for their full face value at maturity. Since the hy[Token]’s initial cost and value at maturity are known upfront, this discounted purchase represents a fixed rate of return.

Abstracting interest rate dynamics into a single price opens up interesting Market Dynamics while giving users the freedom to employ a number of Trading Strategies.

To ensure that a balanced market exists where the market price can be increased and decreased by supply and demand, the AMM supports three basic operations:

• Opening Longs, which provides exposure to the fixed rate by purchasing hy[Tokens] at a discount to their face value for the price of forgoing the variable rate. Longs pay trading fees to the pool.
• Opening Shorts, which provides exposure to the variable rate generated by the capital that backs hy[Tokens] for the price of paying the fixed rate. Shorts pay trading fees to the pool.
• Providing Liquidity, which facilitates Long and Short trading by automatically taking the other side of user positions, and charges trading fees. Part of the pool's fees may go to a governance address.

Read more at docs.hyperdrive.box

This docs page can be found via https://agent0.readthedocs.io/en/latest/.

What is Agent0?

Agent0 is DELV's Python-based library for testing, analyzing, and interacting with Hyperdrive's smart contracts. It provides ready-for-use trading policies as well as a framework for building smart agents that act according to policies that can be strictly user-designed, AI-powered, or a combination. These agents are deployable to execute trades on-chain or can be coupled with a simulated environment to test trading strategies, understand Hyperdrive, and explore integrations or deployment configurations.

When running Hyperdrive on a local blockchain, agent0 also provides a managed database delivered to you as Pandas dataframes via an API as well as a visualization dashboard to enable analysis and understanding.

Read more about how agent0 works in our docs.

Quickstart | Agent0 Repo

This repo contains general purpose code for interacting with Ethereum smart contracts. However, it was built for the primary use case of trading on Hyperdrive markets.

First, install Foundry and Docker.

Next, using a Python 3.10 environment, you can install agent0 via pip:

pip install --upgrade agent0

Finally, you can execute Hyperdrive trades in a simulated blockchain environment:

import datetime
from fixedpointmath import FixedPoint
from agent0 import LocalHyperdrive, LocalChain

# Initialize
chain = LocalChain()
hyperdrive = LocalHyperdrive(chain)
hyperdrive_agent0 = chain.init_agent(base=FixedPoint(100_000), eth=FixedPoint(10), pool=hyperdrive)

# Run trades
chain.advance_time(datetime.timedelta(weeks=1))
open_long_event = hyperdrive_agent0.open_long(base=FixedPoint(100), eth=FixedPoint(10))
chain.advance_time(datetime.timedelta(weeks=5))
close_event = hyperdrive_agent0.close_long(
    maturity_time=open_long_event.maturity_time, bonds=open_long_event.bond_amount
)

# Analyze
pool_info = hyperdrive.get_pool_info(coerce_float=True)
pool_info.plot(x="block_number", y="longs_outstanding", kind="line")

See our tutorial notebook and examples notebook for more information, including details on executing trades on remote chains.

Install

Please refer to INSTALL.md for more advanced install options.

Deployment

Please refer to BUILD.md.

Testing

We deploy a local anvil chain to run system tests. Therefore, you must install foundry as a prerequisite for running tests.

Testing is achieved with py.test. You can run all tests from the repository root directory by running python -m pytest, or you can pick a specific test with python -m pytest {path/to/test_file.py}. General integration-level tests are in the tests folder, while more isolated or unit tests are colocated with the files they are testing and end with a _test.py suffix.

Contributions

Please refer to CONTRIBUTING.md.

Coverage

To run coverage locally you can follow these steps:

pip install coverage
coverage run -m pytest
coverage html

then just open htmlcov/index.html to view the report!

Number format

We frequently use 18-decimal fixed-point precision numbers for arithmetic.

Disclaimer

This project is a work-in-progress. The language used in this code and documentation is not intended to, and does not, have any particular financial, legal, or regulatory significance.

---

Copyright © 2024 DELV

Licensed under the Apache License, Version 2.0 (the "OSS License").

By accessing or using this code, you signify that you have read, understand and agree to be bound by and to comply with the OSS License and DELV's Terms of Service. If you do not agree to those terms, you are prohibited from accessing or using this code.

Unless required by applicable law or agreed to in writing, software distributed under the OSS License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the OSS License and the DELV Terms of Service for the specific language governing permissions and limitations.