Make it again with new inputs: reusable graphs
Turn a creation into a graph with Make this reusable, run it with new inputs, continue a stopped run without paying twice, and favorite or copy graphs from Discover.
Updated
A graph is something you made, saved so you can make it again with new material. A listing tour, a product spin or a podcast clip: the graph keeps the steps that made it, and its inputs are what changes next time, such as a listing link, a product photo or a headline. There is no separate template to set up. A graph is the template.
- Make
Finish something you like in a conversation.
- Save
Choose Make this reusable.
- Run
In My graphs, fill in new inputs and run it.
- Continue
If a run stops, Continue picks up where it left off.
Turn a creation into a graph
- Finish something in a conversation that you want to make again: an image, a video or audio.
- Under the agent's reply, in Try next, choose Make this reusable. The button appears when the latest turn made an image, video or audio.
- The agent reads how the work was made and saves it as a graph. Its reply names the graph and its inputs, for example Listing URL, Music style and Director's brief.
Saving doesn't run the graph. The agent's turn is charged like any other turn.
The graph keeps what actually worked: the same models, prompts and edit steps. Choices that were made by hand, such as which photos to use and in what order, become steps that decide them again from the new material. So a listing tour adapts to a house with 22 photos as well as one with 44.
You can also ask in your own words, and say which inputs you want:
Make this reusable as a graph. The inputs should be the product page link and the headline.Check that it worked
Open My graphs at /app/graphs. The new graph is listed with a form for its inputs.
Find your graphs
My graphs lists your Favorites first, the graphs you favorited in Discover, then Your graphs, every graph saved in this workspace. Open it from My library: choose Library tools, then My graphs. You can also go straight to /app/graphs.
Each graph is a card with its name, what it makes, and one field per input. Required and short inputs come first. Inputs with a long default, such as a director's brief, are under More options.
Graphs belong to the workspace: everyone who works in it sees the same graphs. Switch workspace to see that workspace's graphs.
Run a graph with new inputs
- In My graphs, fill in the graph's inputs. Leave an optional input empty, or as it is, to use its default.
- Choose Run with these inputs.
- The top of the page says Running. A long run can take up to an hour. You can leave the page; the run carries on.
- When it finishes, the result appears in the status, and it is in your Library. Choose Open the Library to find it.
A graph runs with no agent: it makes the same kind of piece from the new inputs, deciding what it needs from them, and doesn't take new instructions. To change what it makes, change its inputs, or change the graph itself (below).
Running needs permission to create and run work in the workspace. A run spends credits for the work its steps do, as the same work would in a conversation. See Understand credits, reservations and charges.
Continue a run that stopped
A run can stop before it finishes: it reached the one-hour limit, or a step failed. The status then says why and offers Continue.
Choose Continue to pick up where it stopped, with the same inputs. Steps that already finished are reused. They don't run again and aren't charged again; only the rest runs.
Change a graph
Ask in a conversation. For example:
Change my listing tour graph so the music is calmer and the end card shows the price.The agent edits the saved graph, and the next run uses the change. To change only one run, use its inputs instead.
Add a graph someone gave you
Someone can give you a graph as its file, a .json file such as listing_tour.json. To add it to your graphs:
- Start a conversation, attach the
.jsonfile to your message, and send:
``brief Save this as a reusable graph in my graphs.
- The agent saves it in this workspace and replies with its name and inputs.
- Open My graphs at /app/graphs. The graph is listed, ready to run with your own inputs.
The copy is yours: change it as you like. It doesn't change when the person who gave it to you changes theirs.
Graphs from Discover: favorite one or add a copy
Wavemaker publishes ready-made graphs in Discover under Graphs, such as a listing tour for real estate. Open one to see what it makes, its inputs and its examples. Then choose either:
- Favorite (the bookmark at the top of the card). The graph appears under Favorites at the top of My graphs, ready to run as published. It is also in My library › Saved.
- Add to my graphs. This saves your own copy under Your graphs. You can ask the agent to change it, and it doesn't change when its creator publishes a new version.
A favorite keeps the exact version you chose. When its creator publishes a later version, its card in My graphs says so; choose Use version 2 (or whichever is newer) to switch. If the creator withdraws a graph, it leaves your favorites; a copy you added stays. To stop seeing a favorite, choose Remove from favorites on its card. On a favorite's card, Add a copy to my graphs makes your own copy, and Open listing returns to its page in Discover.
Favorites and your graphs belong to the workspace, so everyone who works in it sees the same ones.
Publish a graph
Wavemaker staff publish graphs to Discover from My graphs: Publish on a graph's card opens the same steps as publishing from a conversation, with the category set to Flow graph. Publishing an unchanged graph again keeps its current version; after a change it becomes the next version. See Publish your work and manage listings.
Use a graph from an agent, the CLI or the API
A graph is a file in your workspace, for example wm://memory/graphs/listing_tour.json. An agent connected to Wavemaker can read it and run it; ask it to "run my listing tour graph" with the inputs you want.
For developers: adding a graph is one write to a wm://memory/graphs/ uri, with the JSON as content or a link to it as from_url. In code, wm.graph.run(uri, inputs, key) runs a graph inside run. From the command line, wm graph wm://memory/graphs/listing_tour.json --input listing_url=https://... --key tour-1 runs it, and --check checks the graph and inputs without spending anything. Running again with the same key continues a stopped run.
Troubleshooting
| What you see | What it means | What to do |
|---|---|---|
| No graphs yet | Nothing has been saved as a graph in this workspace. | Make something in a conversation, then choose Make this reusable. |
| Make this reusable isn't offered | The latest turn didn't make an image, video or audio. | Ask for it in your own words, as above. |
| Fill in Listing URL to run this graph. | A required input is empty. The message names it. | Fill it in, then run again. |
| The run stopped | It reached the time limit, or a step failed. | Choose Continue. Finished steps aren't charged again. |
| A favorite is gone from My graphs | Its creator withdrew that version. | Find the graph in Discover again, or use the copy if you added one. |
| This run's inputs could not be read back | Continue couldn't recover the run's inputs. | Run it again from its form. |
If a run fails again at the same step, see Recover interrupted or uncertain work.
Related
- Library: find, organize and reuse your work
- Check graph and method requirements, for Comfy graphs you import
- Connect an external agent with MCP