Flow editor components

The Flow Builder is made up of a handful of major pieces:

• Toolbar: export, settings, and test controls for the whole flow.
• Settings: flow metadata and advanced options.
• Flow env variables: constants shared across every step in the flow.
• Flow inputs: the parameters the flow itself accepts.
• Actions: the individual steps that run when the flow executes.
• Action editor: where you configure each action’s behavior.

Toolbar features

From the toolbar you can reach:

• Summary: rename the flow.
• Undo/Redo: step back and forward through your edits.
• Path: set who can access the flow.
• Menu options: deployment history, exporting as JSON/YAML, and editing the YAML directly.
• Tutorial access: guided walkthroughs, with the option to reset or skip them.
• Diff viewer: compare the current draft against the last deployed version.
• AI Builder: generate or modify the flow with Orvanta AI.
• Sticky notes: annotate the canvas for yourself or teammates.
• Test controls: run the whole flow, or step through it up to a chosen point.
• Draft/Deploy: save without publishing, or push the change live.

Settings configuration

Under flow metadata you’ll find:

• Summary: an optional, human-readable title.
• Path: the unique identifier combining owner and flow name.
• Description: markdown instructions shown to whoever runs the flow.
• Advanced options: worker group selection, caching, early stop, early return points, and shared-directory access.

Flow environment variables

These are flow-scoped constants, referenced from any step’s input as flow_env.VARIABLE_NAME. They come in three flavors:

• Plain string values, optionally linked to a workspace variable.
• JSON objects, for structured data.
• Connection references, resolved when the step runs.

Flow inputs configuration

There are three ways to add an input:

1. Customize manually: set its name, type (Integer, Number, String, Boolean, Array, Object, Any), description, display title, and placeholder.
2. POST request: send a sample JSON payload to the capture endpoint and let Orvanta infer the fields.
3. Copy first step inputs: reuse whatever the first step already declares.

Depending on the type, advanced settings can add currency formatting, regex validation, file encoding, enum choices, or a specific connection type.

Action steps

A step can be:

• An inline script (Python, TypeScript/Deno, Go, Bash, Nu, one of the supported SQL dialects, a REST call, GraphQL, or PowerShell).
• A reference to a script on the Orvanta Hub.
• A reference to a script already in your workspace.

Action editor components

The action editor itself breaks down into:

• Header: rename the step and jump to retries, concurrency, cache, early stop, suspend, sleep, or mock settings.
• Script editor: context variables, input variables, connections, a reset button, the LSP-backed assistant, formatting, and a read-only script view.
• Step configuration: define each input as a templated string, a dynamic expression, or a static value.
• Test section: run this one step in isolation with a chosen set of inputs and inspect the result.

Inputs accept three shapes — a templated string using ${variable} interpolation, a dynamic JavaScript expression, or a static value — and can be filled in one of two ways: appending a reference into the field, or connecting it directly to another step’s output.