# For developers

This chapter is mainly intended to developers who want to contribute to Ersilia's infrastructure. If you intend to contribute AI/ML models, please refer to the [Model contribution](https://ersilia.gitbook.io/ersilia-book/ersilia-model-hub/model-contribution) chapter.

## The Ersilia CLI

The main codebase of Ersilia is the [Ersilia CLI](https://github.com/ersilia-os/ersilia). For a full reference of all commands available, please visit [this page](https://ersilia.gitbook.io/ersilia-book/ersilia-model-hub/developer-docs/command-line-interface). A detailed reference of the Ersilia API can be accessed [here](https://ersilia-os.github.io/ersilia). 

To start contributing to Ersilia, please [fork our master branch](https://github.com/ersilia-os/ersilia/fork) and work on it locally. We recommend installing Ersilia in editable mode inside a Conda environment:

```bash
conda create -n ersilia python=3.12
conda activate ersilia
git clone https://github.com/your_username/ersilia
cd ersilia
pip install -e . 
```

### Working with models as Docker containers

All models incorporated in Ersilia are [dockerized](https://hub.docker.com/u/ersiliaos) for easy deployment. While the dockerization step happens as part of our [CI/CD workflows](https://ersilia.gitbook.io/ersilia-book/ersilia-model-hub/developer-docs/ci-cd-workflows), it is recommended to install Docker for model testing purposes.

{% hint style="success" %}
Running models as Docker container is the recommended way since it maximizes interoperability across systems and persistency.
{% endhint %}

### Working with models from source

Ersilia models can also be packaged from source. The source code and parameters for a given model are available in its corresponding GitHub repository. Some model checkpoints are too large (>100MB) for GitHub storage. If you want to work with models from source, please make sure that [git-lfs](https://git-lfs.com/) is installed and active in your system to push large files to the model repository.

{% hint style="info" %}
Working with models from source is often recommendable to quickly explore potential issues with the model. Typically, a dedicated Conda environment will be created by Ersilia automatically at fetch time.
{% endhint %}

## CI/CD workflows and testing

Ersilia relies heavily on GitHub Actions workflows for automation and testing. Visit the corresponding sections to learn more about:

* [Basic concepts of CI/CD at Ersilia](#ci-cd-workflows-and-testing)
* [The model testing command](https://ersilia.gitbook.io/ersilia-book/ersilia-model-hub/developer-docs/model-testing)
* [A fully-featured testing playground for the Ersilia CLI](https://ersilia.gitbook.io/ersilia-book/ersilia-model-hub/developer-docs/cli-testing)
* [Styling and guidelines to ensure code quality in an automater manner](https://ersilia.gitbook.io/ersilia-book/ersilia-model-hub/developer-docs/developer-guide-for-codebase-quality-and-consistency)

## Model packaging

An important part of the Ersilia infrastructure is the packaging of models with [Ersilia Pack](https://github.com/ersilia-os/ersilia-pack). Please see the [Ersilia Pack documentation ](#model-packaging)for more information.

{% hint style="warning" %}
The legacy method for packaging models was strongly based on [BentoML](https://github.com/bentoml/). This is progressively being deprecated in favor our Ersilia Pack, which is built on top of [FastAPI](https://fastapi.tiangolo.com/).
{% endhint %}