> ## Documentation Index
> Fetch the complete documentation index at: https://docs.fapost.in/llms.txt
> Use this file to discover all available pages before exploring further.

# The flow builder

> Building a conversation from blocks.

The builder is where a flow is actually written. It opens from **Builder** on any
row of the [flows list](/using/flows).

Blocks are stacked vertically in the order they run. There is no canvas to
arrange and no lines to draw: a flow reads top to bottom, the way the
conversation happens.

## The screen

**Left — Structure.** A tree of every block in the flow, including those nested
inside loops and subflows. On a long flow this is how you navigate; the main
column is for editing, the tree is for finding.

**Right — Main flow.** The blocks themselves. Each shows its type, its number,
and its configured values inline, so the flow can be read without opening
anything.

**Top — the toolbar.** The flow's name, its published version, and whether the
current work is saved.

## The trigger

The first card in every flow is its trigger — what causes this flow to start.

A new flow shows **No trigger configured**. That is not an error and the flow can
still be built, but nothing will start it: it can only run as another flow's
subflow, or as the assistant's default flow.

## Working with blocks

**+ Add block** appears between every pair of blocks and at the ends. Where you
click is where the block lands — this is how order is chosen, rather than adding
and then moving.

The palette that opens has a search box and seven categories:

| Category        | What it does                                           |
| --------------- | ------------------------------------------------------ |
| **Core**        | Send a message, ask for a reply                        |
| **Logic**       | Branch, wait, loop, end, run a subflow, raise an event |
| **Data**        | Store a value, tag a contact                           |
| **Integration** | Call another system over HTTP                          |
| **AI**          | Query the knowledge base                               |
| **Contact**     | Ask for authentication, notify staff                   |
| **Annotation**  | A comment for whoever reads the flow next              |

The full list with each block's exact type and configuration is in
[Flow node reference](/reference/flow-nodes).

<Note>
  Comments are worth using. A flow is read far more often than it is written, and
  a branch whose condition made obvious sense in the moment rarely does six
  months later.
</Note>

Each block carries its own controls: move up, move down, swap, and delete.

## Two tabs: Builder and Content

**Builder** is the structure — which blocks, in what order, with what settings.

**Content** is the words. The language selector beside it switches between the
languages the tenant has enabled.

The separation is deliberate. Someone translating a flow should not be able to
rearrange it by accident, and someone rearranging it should not have to scroll
past paragraphs of text to see the shape.

## Undo, redo, validate

**Undo** and **Redo** cover editing within the session.

**Validate** checks the flow for problems without publishing it — blocks missing
required configuration, outcomes that lead nowhere, structures that cannot
complete. Run it before publishing rather than after.

## Saving is not publishing

The toolbar shows the save state — **Saved** when there is nothing pending.

**Save draft** stores your work. **Publish** puts it in front of contacts. They
are separate buttons because they are separate decisions, and saving is safe at
any moment: contacts keep receiving the published version until someone
explicitly publishes a new one.

See [Publishing a flow](/using/publishing).