Your first workflow

This walks through building a working integration from an empty system. It takes about twenty minutes and produces a workflow that reads records from one API and writes them to another, on a schedule.

You will need a running OpenCelium (Quick install) and credentials for two APIs. If you only want to try the mechanics, any two HTTP endpoints will do — you can use Simple HTTP request steps and skip the invoker work entirely.

Note

Signed in for the first time? The default account is admin@opencelium.io / 1234. Change it.

1. Make the systems known

Each system needs an invoker (how to talk to that kind of API) and a connector (one instance of it, with credentials).

Check Configurations → Invokers first: if an invoker for your product already exists — shipped, uploaded, or synchronised from the Service Portal — you are done with this part. If not, create one, or take the shortcut in step 3.

Then create a connector per system under Connectors → Create:

  1. General Data — a title, the invoker, and optionally a timeout and icon.

  2. Credentials — the fields the invoker’s authentication type requires.

Use Test connection before submitting. If it fails, OpenCelium asks you to confirm explicitly rather than silently storing a broken connector — fix the credentials instead.

Do this twice, once per system.

2. Create the workflow

Press Alt+W, or use Create Workflow in the top bar. The editor opens on an empty canvas with a single Start node.

Give it a title straight away: click the title in the header and type. Titles must be unique, and the check runs as you type.

3. Add the first step

Hover the Start node and click the + handle. The step drawer opens with Choose your next step:

  • Use Connector — pick your source connector, then the operation that reads the records. This is the normal path.

  • Add HTTP Request — if you skipped the invoker: pick the HTTP method and enter the URL yourself.

The step appears on the canvas showing the connector icon and the method name. Note its status dot: green means the connector’s test passed.

Double-click the step to inspect its request. For a connector method the HTTP method is fixed by the invoker; the URL, headers and body are editable.

4. Loop over the results

The source call returns a list, and you want one write per element.

Hover the first step, click +, choose Add Operator, then Loop.

Double-click the Loop node to open its condition. Choose the For operator and point its argument at the array in the first step’s response — use the reference generator: connector → method → field, and pick the array with [*].

The Loop node exposes a loop variable (the iterator). The info panel next to the condition shows it, with examples.

5. Add the write step inside the loop

Use the + handle on the Loop node’s nested path — not the one continuing past it — and add a step on your target connector: the operation that creates or updates a record.

6. Map the data

This is the actual integration. Open the write step’s Body.

The body shows as a JSON tree. Select the field you want to fill, then insert a reference to the source step’s corresponding field. Because you are inside a loop, append the iterator so each iteration takes its own element — for iterator i and field result, that is result[i].

Repeat per field. A field holding only a reference is a direct reference: a straight copy, and the fastest option.

When a value needs transforming — uppercase, a date reformat, concatenating two fields — use Create enhancement on the reference. You get a script seeded with RESULT_VAR = VAR_0, where VAR_0 is your reference and RESULT_VAR is the field’s value:

RESULT_VAR = VAR_0.toUpperCase();

Note

Prefer direct references. Only add a script where you genuinely transform the value — see Data mapping.

7. Test it

Save with Ctrl+S, then start a test run from the header. It executes the current editor state and streams logs into the panel at the bottom.

Expand the tree to see each call with its URL, status, duration and payloads. If a call fails, the viewer takes you to it: red trace dots, loops paged to the failing iteration, the error shown on the step.

Iterate here until it does what you want. Test runs are cheap and only one runs at a time per workflow.

8. Schedule it

Once it works, give it a schedule. Either use the schedules pill in the editor header — Add schedule, scoped to this workflow — or go to Schedules → Create.

A schedule needs a title and a workflow; the cron expression is optional, and the generator builds one for you with a preview of the next runs. Leave logging off unless you need it, and switch it on from the list when you do.

Note

Only one schedule per workflow runs at a time.

9. Get told when it breaks

Create a notification template under Configurations → Notification Templates, then attach a notification to the schedule with event type alert so a failure reaches you by e-mail or webhook.

See Schedule and notify.

Where to go next