Getting started
Hi, welcome to Huma. This guide will quickly take you through setting up the development environment for Huma Evaluation Agent System, including individual evaluation agents and signal adapters that power the Decentralized Signal Portfolio(DSP).
See our whitepaper to learn more about the Huma Protocol and the underwriting system's overall design.
To support the Huma Protocol and the lending pools to make good lending decisions. There're 3 major off-chain components we built:
- Decentralized Signal Portfolio (DSP)
- DSP consists of a group of Signal Adapters that gather data from a variety of on-chain and off-chain sources.
- Evaluation Agent
- Each evaluation agent is a microservice employed by a lending pool to make underwriting decisions.
- EA relies on the DSPs to gather data about the loan request.
- You can think DSP is the features and EA is the model in a Machine Learning system.
- Evaluation Agent Host (EAverse)
- EAverse is the node quarterbacking Huma Protocol, DSP and EAs.
- It's currently a centralized platform due to the complexity of computation. As decentralized execution engines become more powerful, we will transition to a decentralized execution platform as well.
We suggest you use python version >= 3.10. The development environment is tested using venv for python env management.
make run-local
Go to
http://localhost:8000/docs and you can make requests through the UI.Poetry is a tool for dependency management and packaging for Python. It creates a lockfile to guarantee reproducible installs and distribution. Highly recommended to browse through poetry usage if not familiar with it already.
Please install it following the official installer. For macOS users, note brew install may not install poetry correctly.
curl -sSL https://install.python-poetry.org | python3 -`
Poetry automatically creates new venv for packages with
pyproject.toml. (optional) VS Code users might find setting poetry config virtualenvs.in-project true makes it easier to manage the python interpreter.poetry install
To ensure code consistency we have CI to check code style. We use
black, flake8 and isort with some customization. You can run the CI check withmake lint-check
To automatically fix issues, run
make lint
In this project, we require all code to use type hints in order to ensure the reliability and maintainability of our codebase.
Using Type Hints and Pydantic in our project allows us to improve the quality and reliability of our code. Type hints provide a way to specify the expected data types for function arguments and return values, which can help catch errors at runtime and improve code readability. Pydantic adds additional features on top of type hints, such as automatic data validation and data conversion, which can help prevent errors and ensure that our code is working with consistent, correct data.
To run the whole test suite.
make test
To run individual tests:
ENV=test poetry run python3 -m pytest {Path to the test file}
- Please create pull requests early to start the conversation about the changes.
- The pull request title should summarize the contribution. Prefix with [WIP] if the PR is still a work in progress.
- All new codes should have an extensive suite of tests with necessary fixtures included. All CI tests need to pass before merging.
We use GitHub workflows to manage CI and CD pipelines. Once the new PR is merged, a new docker image will be built and pushed to container repository. Currently, the final step in applying Terraform change is managed by Huma DAO.
Last modified 3mo ago