- Major bug fixes
- Protect against writing a new project from existing directories (crashes)
- Project file saves after the user clicks [SAVE PROJECT]
- A new [New Project] dialog
- Updated the stat bar
- Fully functioning basic Neural Network nodes
- Reorganized
README.md
- Draggable (panning) workspace
- Project will not save the reference to a null workspace file
- Fixed a bug: the constant model
Selector
stores the correct internal data, but does not get updated onto the UI on deserialization - Repurposed the inspection menu into hopefully a helpful little wiki
- Replaced the list nodes with a tree view as the list starts to congest
- In parallel to the congestion, created a new way of placing new nodes continuously
- The text sizing issue each time you open again should be fized
- Actual LICENSE file changed
- Build optimization
- LICENSE CHANGED
- Changed from Apache v2.0 to GPL v3.0 because of PyQt5 and PyInstaller.
- And since its an end-user product, I believe this is a suitable license to support open-source software
- This will be permanent
- (Did attempted to use fbs and PyInstaller and cx_Freeze)
- Binaries is here!
- Using PyInstaller (developer version because of some hidden imports shenanigans)
- Access distribution via https://andrews236.github.io/
- Serialization now is able to serialize constant field datas
- New Optional Type (The last type we will implement for now)
- Added another node, TrainNN, for training Neural Networks
- Internal code clean-up
- Fixed:
- The FileDialog in the nodes does not close when clicked [CANCEL]
- Loading project, then creating a new Workspace with new name causes an E001
- Graphs Testing:
- Linear Regression
- Added few more node types:
- Input Layer NN (Non-functional)
- Hidden Layer NN (Non-functional)
- Output Layer NN (Non-functional)
- CNN (Non-functional)
- Pooling (Non-functional)
- Added variation towards each connector based on type
- .. which means basic type checking in executor
- Serialization update
- updated with typing mechanism
- refactored the executor data saver (it was buggy)
- Fixes:
- Serializing nodes with the same class yields same instances
- File output from the .proj.dat was resulting a file serialization "memory" leak
- (While I was debugging, saving model a lot of time, the .proj.dat >1000 lines)
- Graphic Testing (no change)
- Fixed the
executor
problem - Further improvement on serialization though broken still
- Models are removable from the project reference
- Users has to manually delete the actual file data
- Changed
pip
topip3
in .travis.yml
- removed os from .travis.yml to default
- .travis.yaml -> .travis.yml
- Cleaned up the repository and files
- Added unit/integration tests
- Fixed the connector's initiation position
- Models are now serializable! (Or at least somewhat)
- Allows for deployment
- Added multiple model workspace
- Saves the visual progress (e.g. lines, node position)
- And many more features yet to come stemming from this
- NOTE: load_proj() currently returns partially restored object state
- Refactored some components (Project File Interface & Model Manager)
- Re-organized some GUI components
- Added Standard Decision Tree Classifier
- Added SVM (Support Vector Machine)
- Added a colored version for numerical only input line
- WHITE: Generic (names, files, ...)
- GREEN: Integer Only (numerical, operand, ...)
- Code Cleanup
- Added KMean
- Added an example
- Revamp on executor
- More models
- More major changes
- The model input and output now uses tag names instead of file location (ease of use)
- Changed the project input into a file dialogue
- Added ML models
- Added the project directory on the left-panel (WIP)
- Reorganized the project workspace
- Changed some other front-end feature (meaning some new things have no back-end code to execute it)
- Fixed the problems with the output data of the model
- Changed README.md
- Full Code Refactoring
- PyQt Library states I can use Apache License so long I only distribute only my part of the source code
- Working makeshift AI models
- Other minor changes
- Added (defunct) AI models
- Replaced version 3.6 to 3.8 in Travis
- Added dummy type system to the executor (later to be actual type system)
- Refactored the components
- Added a search bar instead of using keyboard to select models individually
- Working models
- New project listed in the config.yaml automatically gets initialized
- Polished the execution process
- Added input backflow
- Other small changes...
- Execution process working
- Added Models
- Working on execution process
- Added input text field
- The Great Refactor #1 ends
- cleaning-up refactor artifacts
- refactor polishing
- remove, now, useless files and directory
- bug: input field can connect to the output field of the same box
- sphinx was overkill
- finish redirecting code
- code cleanup
- 50 lines of mess is cleaned up in
main.py
- removed old codebase
- 50 lines of mess is cleaned up in
- documenting documentation for users
- redirected the main code to the new codebase
- smooth clicking on nodes
- The Great Refactor #1 starts
- added the code to setup for change of codebase
- original code still running
- Minor change to README
- Testing Travis #1-4
- Added the basic framework for the backend (AI part)
- Minor Changes to the frontend (GUI)
- Organized Code
- Added Travis-CI to the project
- Learning how CI works!
- Added CHANGELOG for tracking updates
- Added more stuff into README
- TODO added for anyone suggesting stuff
- DOC added for now as a software reference and tutorial
- Published project to the GitHub repository
- Project code-named "Hexacone" (don't ask why)