Releases: debiai/DebiAI
v0.30.10
DebiAI-gui now supports new parameters #212
# Select data folder path
debiai-gui start -d ./my_data
# No browser
debiai-gui start --no-browser
# Add Algo-providers
debiai-gui start -ap http://localhost:4000 http://localhost:8000
# Add Data-Providers
debiai-gui start -dp http://localhost:4000 http://localhost:8000
# Display help
debiai-gui start --help
usage: debiai-gui start [-h] [-d DATA_FOLDER] [-p PORT] [-nb]
[-dp DATA_PROVIDER [DATA_PROVIDER ...]]
[-ap ALGO_PROVIDER [ALGO_PROVIDER ...]]
options:
-h, --help show this help message and exit
-d DATA_FOLDER, --data-folder DATA_FOLDER
Path to the data folder
-p PORT, --port PORT Port to run the server on
-nb, --no-browser Do not open the browser on startup
-dp DATA_PROVIDER [DATA_PROVIDER ...], --data-provider DATA_PROVIDER [DATA_PROVIDER ...]
One or more data-provider URLs
-ap ALGO_PROVIDER [ALGO_PROVIDER ...], --algo-provider ALGO_PROVIDER [ALGO_PROVIDER ...]
Easy Data provider Python module
A debiai_data_provider python module has been created to make the creation of Data-providers easier.
Easy Algo provider Python module
An algo_provider python module has been created to make the creation of Algo-providers easier.
2D point plot fixed range option #248
We can now fix the ranges for the 2D point plot, keeping aspect ratio when zooming or resizing the widget.
Tactile update #226
Gridstack is now up-to-date, tactile now works as expected on phone or big touch screens :
Algo-providers table results
List of dictionaries are supported as an Algo-Provider output, they will be displayed as an array:
Bug fixes
Layout save #244: Increased the Gridstack npm package caused some issues when saving layouts, it has been fixed.
Text column filter issue #224
The Algo-Provider's API implementation is now more robust and less prompt to errors #221
Minor improvements
With the [Enter]
key support and an automatic name generation, it's now easier to add Data-providers from the Dashboard.
Loading and creating selections is now possible from the DebiAI data insertion Python module.
DebiAI size has been divided by 3 with the removal of heavy dependencies.
v0.29.9 Null values support and new Array widget
Null values support #191
It is now possible to import missing (null, Nan, none) values into DebiAI.
The columns now display the percentage of null values:
The Statistics
widget also displays the percentage of missing values for each color:
You can also filter null
values:
We put in a lot of effort to make sure every widget works well with this change. If you find something that isn't right, please write us an issue.
Array widget remake #213
The previous Array widget was slow and had some miss-alignment between rows and columns. We completely remade it to support a large amount of data and to support dynamic size rendering:
As you can see from the image, missing values are supported by this widget.
v0.28.0 Easy start and Arrays & Dictionaries support
Easy start update
DebiAI is now available as a standalone python module using pip.
As of now DebiAI can be installed with Docker or git for production. But now using pip you can switch between production and development more quickly by installing once DebiAI and running your script.
New command to install DebiAI as python package:
pip install debiai-gui
New command to run DebiAI:
debiai-gui start
New commands to launch the backend in development:
cd debiai/debiaiServer
pip install -r debiaiServer/requirements.txt
python3 run_debiai_server_dev.py
Arrays & Dictionaries support
Arrays and Dictionaries are now supported by DebiAI and can import by the python module or by the data providers.
Arrays and Dictionaries can't be used as columns for most widgets (apart from the Sample array
widget) but they can be unfolded from the dashboard. This allows you to explore the project's data whatever their dimension. For example, you can analyze the Woodscape images and obstacles data without having to create a second project.
Unfolded columns:
Columns can be unfolded recursively
Unfolding an array column will add more data to the project:
Mode details on the Array & Dictionaries unfolding documentation
New features
Improvements
v0.27.2 Metadata for selections
New Feature
- Display selection metadata #203
v0.27.0 Menu on right-click, models metadata & other improvements
New Features
- Widget menu available on right click #192
- Web Data-providers and python module now supports Models Metadata #194
- Widgets are now in full width upon opening #164
- Ability to select secondary Y axis on range slider plot #179
Bug fix
- Display of warnings during axis selections for the point plot #193 , #152
- Fixed DebiAI-PY env url if "/" or "/#/" are added to it. #166
- Ability to save correlation widget configuration #167
Other changes
v0.26.0 Integrated algo provider & other improvements
Features update
The Regression metrics provide new outputs #169 and added a Classification metric algorithm #138
Improved project page look and feel #153 #142 #141
The integrated data-provider can now be read-only #134
Bug fix
- Project name is now displayed instead of the project id #155
- Fixed issues with algo-hub id_list #148
- Prevented crash when Python DP model doesn't exist #137
Other changes
v0.25.0 Complete UI restyle, new widgets and improvements
Complete UI restyle #124
We spend the last weeks completely restyling DebiAI. The theme is now clearer, the widget buttons are placed in a more logical way and the analysis dashboard menu and widget catalogue have been completely redone:
New Widgets: 2D density plot and Night stars plot #20 #5
2D density plot:
Night stars plot:
More information on the widget documentation page.
Model data-id loading improvements #130
To reduce the number of requests made to the Web data-providers, we have improved the model results data ID loading procedure. We have also added a short memory cache system to the ID loading. The time to reach the analysis page should appear shorter.
Bug fix:
Pressing the [Enter]
key when creation a selection now creates the selections instead of closing the modal #128
v0.24.0 Cache, column groups and analysis export
Cache improvements #13
When DebiAI will load a project data, it will store the loaded data in the browser store, as long as a project_update_date
is provided.
If an update date isn't provided, DebiAI won't be able to know if the loaded data are up to date.
A new panel in the project page allows you to unable/disable/clear the cache:
Columns group #117
You can now add a "group" : "my-group"
in a column description when asked information about your project structure. This will group the columns in the dashboard, allowing a clearer analysis:
Analysis export #18
You can now export the plot as an image thanks to a new button on top of the widget:
You can also add comments to a widget.
You can also export a full page as a zip file, a markdown file will list all the generated images, the widget comments, the widget configuration and the widget title will be included.
Learn more on our web documentation
v0.23.0 AlgoProviders & dashboard layout management
Custom algorithms
DebiAI now comes with a way to add custom algorithms to the analysis dashboard, they can be used to gerenate metrics that can be analyzed like any other columns.
You can add custom algorithms by either creating an Algo-provider or by editing the DebiAI's integrated Algo-provider.
Learn more on our Custom algorithms documentation
Dashboard layout management
DebiAI now comes with a way to save and load dashboard layouts, this allows you to get back to a previous layout faster. The layouts also save the widget configuration.
Learn more on our Dashboard layout management documentation
v0.22.0 Analysis loading & data-provider API improvements
New features
Range slider backgroud color
It is now possible to set a column as a background color:
#103
Cancel analysis
It is now possible to cancel an analysis with a new "cancel" button
#99
Time remaining
A estimated time remaining is displayed near the progress bar
#59
Data-providers
Project and model deletions
The data-providers API now support project and model deletion
#101
#62
More information on analysis start & end
New information are added when requesting id list & dada:
- Unique analysis ID (string)
- Analysis start (boolean)
- Analysis end (boolean)
#96