OPTIMADE API implementation for AiiDA

Latest release	Build status	Activity

This is a RESTful API server created with FastAPI that exposes an AiiDA database according to the OPTIMADE specification.

It is mainly used by Materials Cloud to expose access to archived AiiDA databases through the OPTIMADE API. But it may be freely implemented by any to fulfill a similar purpose.

The server is based on the test server "template" used in the optimade-python-tools package. Indeed, the filter grammar and parser and pydantic models from optimade-python-tools are used directly here.

Prerequisites

Environment where AiiDA is installed.
AiiDA database containing StructureData nodes, since these are the only AiiDA nodes that are currently exposed with this API (under the /structures endpoint).

Installation

The package is relased on PyPI, hence you can install it by:

$ pip install aiida-optimade

Otherwise, you can also git clone the repository from GitHub:

$ git clone https://github.com/aiidateam/aiida-optimade /path/to/aiida-optimade/parent/dir
$ pip install -e /path/to/aiida-optimade

Development

For developers, there is a special setuptools extra dev, which can be installed by:

$ pip install aiida-optimade[dev]

or

$ pip install -e /path/to/aiida-optimade[dev]

This package uses Black for formatting. If you wish to contribute, please install the git pre-commit hook:

/path/to/aiida-optimade$ pre-commit install

This will automatically update the formatting when running git commit, as well as check the validity of various repository JSON and YAML files.

Initialization

You should first initialize your AiiDA profile.

This can be done by using the aiida-optimade CLI:

$ aiida-optimade -p <PROFILE> init

Where <PROFILE> is the AiiDA profile.

Note: Currently, the default is optimade_sqla, if the -p / --profile option is now specified. This will be changed in the future to use the default AiiDA profile.

Initialization goes through your profile's StructureData nodes, adding an optimade extra, wherein all OPTIMADE-specific fields that do not have an equivalent AiiDA property are stored.

If in the future, more StructureData nodes are added to your profile's database, these will be automatically updated for the first query, filtering on any of these OPTIMADE-specific fields. However, if you do not wish a significant lag for the user or risking several GET requests coming in at the same time, trying to update your profile's database, you should re-run aiida-optimade init for your profile (in between shutting the server down and restarting it again).

Running the server

Locally

Using the aiida-optimade CLI, you can do the following:

$ aiida-optimade -p <PROFILE> run

Where <PROFILE> is the AiiDA profile you wish to serve.

Note: Currently, the default is optimade_sqla, if the -p / --profile option is now specified. This will be changed in the future to use the default AiiDA profile.

You also have the opportunity to specify the AiiDA profile via the environment variable AIIDA_PROFILE. Note, however, that if a profile name is passed to the CLI, it will overrule and replace the current AIIDA_PROFILE environment variable.

# Specifying AiiDA profile as an environment variable
$ export AIIDA_PROFILE=optimade
$ aiida-optimade run

Navigate to http://localhost:5000/v1/info

Tip: To see the default AiiDA profile, type verdi profile list to find the colored profile name marked with an asterisk (*), or type verdi profile show, which will show you more detailed information about the default profile.

Note: The aiida-optimade run command has more options to configure your server, run

$ aiida-optimade run --help

for more information.

With Docker

Adapt profiles/test_django.json and profiles/docker-compose.yml appropriately.

$ docker-compose -f profiles/docker-compose.yml up --build

Navigate to http://localhost:3253/v1/info

Stop by using

$ docker-compose -f profiles/docker-compose.yml down

Jinja templates

If you are familiar with Jinja, there are two templates to create the JSON and YAML files: profiles/config.j2 and profiles/docker-compose.j2, respectively.

Configure the server

You can configure the server with the aiida_optimade/config.json file or set certain environment variables.

To learn more about this, see the optimade-python-tools repository.

Design choices

Q: Why create an individual config.json file instead of just mounting an existing .aiida directory and using that directly?
A: This, currently, wouldn't work because the REPOSITORY_URI needs to point to the right path inside the container, not on the host. Furthermore, storing all configurations in the same file can be fragile.