Run, save, and deploy
Running the pipeline
You can run the pipeline right from the builder screen to test pipeline execution. Make sure to meet all the requirements before doing so:
- Pipeline has at least one input node
- Pipeline has at least one output node
- All nodes and handles are connected
- All inputs are non-empty in the left sidebar
- All required variables are set in env variables page.
Simply click “Run” near pipeline execution, and you’ll see the outputs in the bottom drawer once the execution completes.
execution result of a simple input-LLM-output pipeline
Saving your work
You don’t have to explicitly save every change. Workshop version is saved automatically!
Commit a new version
If you want to store the current pipeline to be able to come back later, click “Commit” at the top of the page, and give this new version a name. This will save the current workshop version snapshot as a commit, which you can later choose from the versions dropdown. Commit versions can be pointed to by endpoints and used elsewhere.
Deploying a pipeline version
Once you are happy with your pipeline and ready to use it in external applications, be it testing or production, you can deploy expose your pipeline as an API endpoint.
- Make sure you are happy with all node names, especially inputs and outputs. Remember, node names must be unique. See Naming nodes.
- Commit the pipeline version. This creates an immutable snapshot.
- Select the newly created (latest, top-most) commit from the version drop-down.
- Click “Set as Target” at the top of the page.
After the deployment is successful, you can invoke the pipeline execition from your application. Learn more by clicking “use API” in the page or refer to API reference