pedrobcst / xerus-streamlit Goto Github PK

Streamlit interface for XRay Estimation Using Refinement and Similarity (XERUS)

License: MIT License

Python 100.00%

xerus-streamlit's Introduction

NOTE:

Main development of this interface has moved to the MAIN XERUS repository, and it has been merged there. All the development of the interface will be done there.

Xerus-streamlit

This is the repository for the Streamlit interface of XRay Estimation Using Refinement and Similarity (XERUS) Currently, we are in beta phase. Most of functionalities of Xerus is already implemented and can be use via the interface without the need of any coding or use of jupyter notebooks.

INSTALLATION:

Please first install the latest version of Xerus following the instructions in the main repository

After, you can clone this repository and install the Streamlit dependencies doing:

git clone https://www.github.com/pedrobcst/Xerus-streamlit
pip install -r requirements.txt

USING

To use please be aware this is just an interface around Xerus, and thus all the requirements to use Xerus are still necessary. That is:

MongoDB server is running (and/or the configurations for a remote server is set in Xerus package)
The api-key settings (for materials project) are correctly set in Xerus

If this is set, just run:

streamlit run app.py

And the interface should load in the web (defaulting to port 8501).
We are currently working in writing a documentation describing how to use step by step, and a Streamlit recording of the usage should be available soon.

xerus-streamlit's People

Contributors

Watchers

xerus-streamlit's Issues

Remove "View data" button

By default, the data should be automatically plotted after it is uploaded. The 'view data' button is unnecessary.

Add an option to export the results of data

Ideally we should be able to export all the 'results' folder to zip after analysis is finished.

Filter the number of columns from dataframe

Filter the number of columns of the default dataframe to:

name
rwp
wt
spacegroup
crystal system

Update README

Write updated README on how to install, set up and start the interface working.

Fix columns dropping when only one phase is run

When one phase is found, the dataframe does not have the colums "npos" and "n_runs". However, we are always dropping these columns, leading to error (if only one phase is found).

We should change this to just keep the columns we want instead of dropping the ones we dont want.

Multipage Support?

It might be nice (after updating the main Xerus) if we could have mutlipage for:

Simulating + Correlations only
Launching BBO from whatever specified CIFs (Just refinement optimization interface, without the need to launch whole analysis process)
Others?

Add more information about the steps

Add:

Ability to pre-simulate the patterns and calculate correlations without need to run Analyze
Show at analyze at which step we are (Downloading structures, Simulating, Refinement)

Change the search paramaters from inputs to form.

Change these inputs to form so the app is only reload when run analysis is pressed.

Export optimization results (structure)

Export the structural results coming frm optimization

Add SessionState to buttons

Add SessionState to buttons Analyze and Initialize.

Enhace the UI (Beautify)

Work on enhacements of the UI. Beautify it.
Currently is all over the place

Create requirements.txt or enviroment.yml

Create a requriements.txt file for ease of installation or provide a premade conda enviroment file that can easily be used.

Currently used custom packages outside of the ones used by Xerus:

streamlit-aggrid
streamlit

editable correlation plot list

In correlation plot, is it possible to change the order of appealing item (phase)?

Add option to show highest correlated patterns.

Currently the visualization of results are quite lacking. We only show the results of a given refinement process. Sometimes, being able to see all top K highest correlated patterns with the data can be quite useful. Do:

Accept K as st.number_input, to see how much to be chosen
Add the option to plot these K highest correlated patterns.
BONUS: Would be nice if it is possible to do that before running the full analysis (Needs to change Xerus main code to allow that, or just run correlations twice?)
BONUS 2: Show the full plot of each run? (Ie, the process of peak removal and phases found. This can also be quite useful / enlightening for parameter tuning if necessary.)

Current TODO:

Finish the plotting function to show Data, Background, and remove background depending on the options provided
More to come.