↔️Workflow Block
Use workflows in workflows.
Advanced/Experimental Contact us or join Discord to report issues
Workflows can be used as building blocks inside other workflows. This is powerful magick. Our UI for this is pretty rough right now, but it works.
Some other examples of utility workflows you might use:
Random person - outputs simple strings like "20-something female pakistani student"
Background Replacer - removes background from input image, outputs an image
and many more
Known issues & caveats
Known bug: All of the other workflow's inputs show up as text inputs. So image inputs, multipick menus are all just text inputs, and defaults aren't parsed well. So be careful!
No safety checks: If you make changes to the other workflow, you might break any dependent workflows! There is not currently any version-linking or similar to prevent this.
Annoyingly: Finding workflows and pasting in workflow IDs is very annoying. It should be a search control, or a button on another workflow saying "use this workflow in your other workflow". Bonus points if any workflow lists prioritizes your personal workflows or workflows that you've liked or ran in the past.
No error bubbling: Errors inside the other workflow aren't surfaced well.
How to use the workflow block
Add the workflow block by clicking the "+" button in the builder and selecting "Workflow Block" in the Experimental/Advanced section.
Select an existing workflow from the dropdown menu, or press the pencil edit icon and paste in a workflow ID.
The inputs from that workflow will appear. You can hardcode the inputs, or use {variables} to compose text to send to the next workflow:
The Microscopic Vision workflow (below) uses a workflow block to call our staff-created StableDiffusion 3 (with negative prompt) workflow. The SD3 workflow generates an image using the Stability.ai API.
Note how the glif ID is edited in (clv55...), and then input values from the Microscopic Vision workflow are passed as variables to the SD3 workflow's prompt and negative fields. e.g. {ar}, {generating_imageprompt}
More examples & use cases
Fetch pages from Wikipedia
Search Wikipedia & return the page text from the top result: https://glif.app/@fab1an/glifs/clufscg2600007pg6wpnb13dh
Fetch the "Today in History" page from Wikipedia: https://glif.app/@jamiedubs/glifs/clqx796yp000610ym0ngo1oef
Enhancing image generation prompts
Uses LLM calls to turn brief prompts into detailed, effective image prompts for different image gen models (SD1.5 vs DALLE vs SD3 etc).
Last updated