The visual code editor for AI agents can be come difficult to work with as the agent complexity grows (orchestrators, executors, multiple tools, etc). Some ideas to improve the user/builder:
- A snap to grid or similar auto format of the flows and connectors.
- Clicking the individual nodes could be improved by adding a 'right-click' menu that allows one to select the child-node to link to
- AI-assistance, (i.e. connect executor A to tools 1,3) can aide in building complex relationships
- Show a visual on the 'build flow' page that highlights which node is in process while using the playground. Essentially tie the Lang chain trace to the box that is being 'run'
