AFL-automation 1.2.0

Hardware interfacing code for the Autonomous Formulation Lab

NIST Autonomous Formulation Laboratory - Automation Software

This package contains the core laboratory automation software used in the NIST AFL platform.

Its core is the 'DeviceServer' API, a simple way of exposing functionality in simple Python classes to the outside world via HTTP servers. It includes robust item queueing support, output rendering, and hooks to allow for 'smart' generation of user interfaces automatically.

Specific deviceserver instances are provided for a variety of hardware used in the AFL platform: syringe pumps, valves, multiposition flow selectors, UV-Vis spectrometers, x-ray and neutron scattering instruments/beamlines. There are further deviceserver classes that integrate these base devices to perform higher-level functions, e.g. "loading". These classes aim to specify instructions for running a particular protocol in a hardware-agnostic way.

Virtual servers for offline testing

Several stub modules are provided so you can launch the AFL stack without any physical hardware. Running one of these modules the first time creates ~/.afl/config.json containing the Tiled server URL, API key and port bindings. Update this file as needed.

Example commands to start the dummy services:

# start a dummy OT2 robot
python -m AFL.automation.prepare.Dummy_OT2_Driver

# start a virtual sample loader
python server_scripts/virtual_instrument/DummyLoader.py

# optional virtual detectors
python server_scripts/virtual_instrument/VirtualSANS_data.py
python server_scripts/virtual_instrument/VirtualSpec_Data.py

For a full offline demo you can start everything at once:

python server_scripts/virtual_instrument/AllDummy.py

These stub servers expose the same HTTP API as the real hardware allowing you to test an AFL workflow entirely in software.

Production deployment

By default the APIServer will use the waitress WSGI server if it is installed. To fall back to Flask's built-in server pass --no-waitress to AFL.automation.shared.launcher.

Running tests

This repository uses pytest for unit tests. A GitHub Actions workflow runs the tests automatically on every push and pull request using .github/workflows/test.yaml. The workflow installs the package along with the dependencies listed in pyproject.toml.

To execute the tests locally, install the package in editable mode and run pytest. If the installation fails because the package version cannot be determined from Git tags, set SETUPTOOLS_SCM_PRETEND_VERSION=0.0.0 as in the CI workflow:

SETUPTOOLS_SCM_PRETEND_VERSION=0.0.0 pip install -e .
pytest