Skip to contentRevaal
Log in
Sign up
Log inSign up
All chapters
0:37

Building a workflow#

From an empty canvas to something that does the work every day. The longest chapter here, because this is where most of the work is.

A workflow up close: the steps on the canvas, the inspector, the issues card and the ways to run it.

  1. A fresh canvas

    Press «New workflow» on the workflows page. The canvas opens with two steps already on it, «Start» and «End». Whatever you put between them is what the workflow does.

    There is no save button and you should not look for one. Every change saves itself and the corner of the screen says where that got to. If the connection drops, your changes are held and go up on their own when it comes back.

  2. What each run works on

    «Inputs», on the island at the foot of the canvas, decides what a run is about: fresh content from news sources, the rows of a file or a pasted text, items fetched from an API, or messages arriving on Telegram. Each item in that queue is one run.

    A workflow does not have to take anything in. With no source chosen there is no queue either: a run simply goes from «Start» to «End» once, and the steps work from their own prompts.

  3. Steps, and the wiring between them

    Drag from the handle on a card's edge to draw an edge, or just click it to drop a new step there. There are two kinds: an «Agent», which puts a model to work on the text, and a «Condition», which sends the path down a «Yes» branch or a «No» one.

    A step with no edge in and none out is never reached, and everything you set up inside it does nothing. «Tidy» lays the canvas out again so you can see what is actually joined to what.

  4. Inside a step

    Click a step to open the inspector. «Prompt» is what you are asking for, «Model» is what does the asking, «Tools» is what it may do outside the text, and «Knowledge» is which of the library's documents it gets to read.

    In «Model», «Activity» decides whether the step writes text, makes an image, makes a video, reads the text aloud in a voice, or turns audio into text. The last two want no prompt at all.

  5. The issues card

    The issues card is always saying what state the workflow is in. A blocking fault means it will not run; a warning means it will, but something is not what you think it is. Every line names the step and the reason.

    The card is not stricter than the engine: those are the same refusals the run itself makes. A workflow that is clean here does not stop somewhere else for these reasons.

  6. Test first, run after

    A test session walks the workflow one step at a time: it runs each step on its own and shows you the input, the model's thinking, the output, the time and the cost. You can change the prompt right there and take that step again.

    A step with an effect outside the app, one that sends or publishes something, does not run itself. You have to approve it in the inspector first, so testing a half-built workflow never actually publishes anything.

  7. Three ways to run it

    «Run» has three modes: by hand, automatically on a clock or every so often, and by an API call from outside. The automatic one can be told to try again if a run fails; that is off unless you turn it on.

    Every run spends the credit of the workspace the workflow lives in, not the credit of whoever pressed the button. The automatic ones too, including the ones nobody is watching.

  8. Publishing a version

    «Publish» records a version. Before it does, «Review changes» tells you line by line what moved since the last one: which step was added, which prompt, which model, which edge.

    Publishing is not a switch that turns the workflow on. A run always executes what you see on the canvas; versions are the history, so that later you can see when something changed and go back to it.

Build your first workflow.

Create an account

Revaal

Manifesto

A person’s attention is the most valuable thing they bring to work. Every one of us has a limited amount of energy, focus, and presence each day, and a day that is spent does not come back. This is true for every single person on a team, not only the ones at the top. So the real question for any company is a simple one: what are we asking people to spend their days on?

Too often, the answer is repetition. We ask people to copy, paste, reformat, check the same list again, and write the same message for the hundredth time. The work gets finished, but something is lost. A person doing repetitive work is a person who is not thinking. We believe this is a serious mistake, not because such work is beneath anyone, but because a human being is the wrong tool for it. Machines are patient with repetition. People are not, and they should not have to be.

We are building for the other side of that trade. Models now have the processing power and the intelligence to carry the repeating part: the reading, the sorting, the drafting, everything that follows a known pattern. Hand that over, and look at what is left for people: creating things, staying curious, wrestling with problems that have no certain answer, sitting with a question while it is still unclear, and making something new and then changing it again. That is what a human day should be spent on. We think it should be the normal day, not the rare one.