The GUI project is found under the gui/ subfolder. It does not currently have any linting / unit tests set up. This repo uses components from shadcn.
This issue aims to introduce a new view, which can be found in the sidebar under Playground > Workflow View. The vision for this page / view is that whenever an LLM Agent executes a tool, the graph will show the executed tools. This is because an agent will learn to build workflows with tools. For example, a task it might have to execute is "Reply to my latest email", in which the agent will use the read_last_email tool, then write_email to reply back to the user. This means that the agent will be able to reuse this workflow, so that it can just simply call this workflow as a tool with the required input parameters. The goal of this workflow view is to see workflows built up over time in a graph.
I want you to use the react flow library for making the graph. The nodes should be able to be moved and interacted with, and clicked on to view more details. If this library is not suitable (for free) for constructing these graph, choose another one.
This change might require a slight restructuring of the API fetching system. The API that this interacts with is found in woodwork/gui/gui.py. If this is necessary, use tanstack-query to create a generic fetching hook that can be reused. The frontend should listen to events exposed by the agent, and then provide functionality to add a node to the graph, and remove nodes from a graph.
Show workflows in a graph, this graph should automatically updated whenever new workflows are added, either by an agent or manually. This will have to require some push-based system.
The GUI project is found under the gui/ subfolder. It does not currently have any linting / unit tests set up. This repo uses components from shadcn.
This issue aims to introduce a new view, which can be found in the sidebar under Playground > Workflow View. The vision for this page / view is that whenever an LLM Agent executes a tool, the graph will show the executed tools. This is because an agent will learn to build workflows with tools. For example, a task it might have to execute is "Reply to my latest email", in which the agent will use the read_last_email tool, then write_email to reply back to the user. This means that the agent will be able to reuse this workflow, so that it can just simply call this workflow as a tool with the required input parameters. The goal of this workflow view is to see workflows built up over time in a graph.
I want you to use the react flow library for making the graph. The nodes should be able to be moved and interacted with, and clicked on to view more details. If this library is not suitable (for free) for constructing these graph, choose another one.
This change might require a slight restructuring of the API fetching system. The API that this interacts with is found in woodwork/gui/gui.py. If this is necessary, use tanstack-query to create a generic fetching hook that can be reused. The frontend should listen to events exposed by the agent, and then provide functionality to add a node to the graph, and remove nodes from a graph.
Show workflows in a graph, this graph should automatically updated whenever new workflows are added, either by an agent or manually. This will have to require some push-based system.