Skip to main content
This guide will teach you the basics of Runchat by showing you how to generate multiple prompts and images parametrically. This workflow will use images of buildings as an example, but you can easily adapt the prompts to suit your use case.
Before starting this guide, head to https://runchat.app/ and sign in with Google or email and password. No Runchat account? Create a free Starter account at https://runchat.app/signup.

​
Creating a new Runchat

Open the Runchat editor and click New Runchat to create a new workflow. Runchat will save your work whenever you add a node or re-run your workflow from the Editor toolbar at the top of the canvas. You can enter a new file name and manually save your work using the toolbar at the top of the editor.

​
Creating nodes and inputs

Everything in Runchat is built from five basic nodes: Inputs for entering data, Agent for processing data and using tools, Create for working with generative models, Fetch for connecting to external services and Code for writing javascript. There are several ways to display the node creation menu in Runchat:
  • Double click on the canvas
  • Press the space bar
  • Dragging from an input or output parameter and dropping on an empty spot on the canvas
  • (On mobile) tap the + button in the bottom toolbar
Display the node creation menu and create a new Create -> Image node.
If you need to, you can delete a node or an edge by clicking on it then pressing Delete or Backspace.

​
Write your first Prompt

The Create node is used for all generative AI models in Runchat. You can create and edit images, generate videos and meshes, and install your own custom models from Fal if required. If you’re using Runchat for design you will be working with the Create nodes 90% of the time. You give the Create node instructions for what you want it to do using the Prompt input. You can enter your prompt directly into the text input field on the node, or by connecting data from some other node in your workflow. For now, enter “A modernist concrete building with sweeping curved forms”.

​
Running nodes

Nodes won’t do anything until you run them. Now that we’ve set an input to our prompt, press the blue play button in the toolbar below the node to run it and generate an output. You can run all nodes in your workflow by pressing the blue play button in the toolbar at the top of the editor.

​
Inputs and outputs

After the node finishes running, it will display the output. You can flick back to the input form by hovering over the node (or tapping it once on mobile) and clicking the OUTPUT toggle.
Running a node uses 1 credit. The free Starter plan gives you 500 credits to play around with. The Imagen 4 Fast model is great for generating draft images and costs 50 credits. The credit cost for running the model as well as the output type is displayed next to the model name in the selection dropdown.

​
Creating input parameters

Now we will start building up some UI for our workflow that makes it easier to create variations on our prompt. From the node menu, create an Input -> Text node, and move this to the left of our canvas. Enter our initial design concept: “A modernist building with sweeping concrete forms”. We can now connect this text input to any input parameter in our workflow and use it in multiple places if need be.

​
Connecting nodes

Runchat is useful for defining parametric relationships between generative AI models. We can control how data is passed from one node to another by creating edges. Add an Agent node and move this to between our input text node and our Create node. Click and drag from the green output handle on our Input node to the blue Context input handle on our Agent node. Our Agent node will now use the contents of our Input node as “context” for your prompt. Then click and drag the output of our Agent node to the input on our Create node. Now we have a full parametric workflow set up: We can provide some input (e.g. an initial design concept), modify this input with our first prompt (e.g. to create variations), and pass these variations to a final node to illustrate them.
Edge colours indicate what type of data is moving between nodes. For more information in this check the Core Concepts article on Data Types.

​
Working with lists

Now let’s get our Agent node to generate the variations on our initial concept. In the node settings bar, change the output format from Markdown to List. This will force our output to be a list of values instead of paragraphs of text. Enter “Generate 4 variations on this concept” into our prompt input field.

​
Scheduling updates

Whenever we change the inputs to a node, it will be scheduled for an update on the next run. Nodes scheduled for an update are indicated with dashed input edges. When we press the editor Play button in the top menu bar, all scheduled nodes will be run first. If there are no nodes scheduled for update then the entire workflow will be re-run. Press the editor Play button in the top menu bar to update all of the nodes in our workflow.
List outputs are displayed in a table view by default. You can toggle between table and page view by hovering over the output page on a node and clicking on the icons in the bottom right corner. This is also where you can download the output data as a csv file.

​
Next Steps

You’ve now covered the basics of Runchat, but there’s a lot more it can do. Work your way through the next tutorial on creating a tool, take a look at some examples to learn about other use cases, or check out the concepts guides to get a better understanding of how runchat works under the hood.
⌘I