my attempt at integrating generative ai into a no-code block programming platform via text-to-visual interactive content simulates highgear environment
dev log:
TODOs before v0.6: semi-stable display algorithm for othogonal graph layout Building Algorithm from Scratch
Branch Ordering (boolean) for decision nodes Automated clustering of subgraphs (subtasks to actions, decision branch grouping) Top down and Left Right should be TB first, then LR
more variety of examples double bend problem snap to integer coords
maybe: try using LLM to determine some layout requirements
v0.5 prototype:
- New UML/pseudo JS hybrid solution in testing
- New parser for the hybrid solution
- adding more support for various formula related features
- expanded the main window display boundary
- better subtasking understanding
v0.4 prototype:
- includes session saving, renaming, iterative modifications on workflows for each session,
- disabled the original image saving and save/load workflow functions,
- working on more elegant exporting system
v0.3 prototype:
- chatgpt-styled multi-session interface
- added more features in the wf script, including subtasks and pseudo contacts lookup
v0.2:
- save feature for workflow json files
v0.1:
- parser integration - esprima base functionality
- few shot prompting and instructions