Skip to content

workflows.mdx

Latest commit

 

History

History
90 lines (68 loc) · 2.24 KB

workflows.mdx

File metadata and controls

90 lines (68 loc) · 2.24 KB
title Workflows
description Create named session-backed flows with optional launcher UI.

Workflows are named app surfaces for guided work.

Use a workflow when a user should start a specific flow from the sidebar, optionally fill out a launcher form, then continue in a session-backed chat.

Create a workflow

wf = app.workflow(
    "Research Company",
    icon="search",
    description="Collect context and draft a brief.",
    scope="user",
)

Workflows appear in the hosted app sidebar.

Start without a custom UI

If the workflow has no @wf.ui() launcher, Capsule opens a simple workflow chat composer.

@wf.start()
async def start(session: cpsl.Session, input: cpsl.WorkflowInput):
    await session.reply("What company should I research?")


@wf.message()
async def continue_research(session: cpsl.Session, msg: cpsl.Message):
    await session.reply(f"Researching {msg.text}...")

Add a launcher UI

@wf.ui()
def research_form():
    return cpsl.ui.WorkflowShell(
        "Research Company",
        children=[
            cpsl.ui.FormSection(
                "Input",
                children=[
                    cpsl.ui.TextInput("Company", name="company", required=True),
                    cpsl.ui.TextInput("Notes", name="notes", multiline=True),
                ],
            ),
            cpsl.ui.ActionBar(
                children=[cpsl.ui.SubmitButton("Start research", primary=True)]
            ),
        ],
    )

The submitted fields are delivered as WorkflowInput.

@wf.start()
async def start(session: cpsl.Session, input: cpsl.WorkflowInput):
    await session.reply(f"Starting research for {input.company}.")

Add actions

@wf.action("approve")
async def approve(session: cpsl.Session, input: cpsl.WorkflowInput):
    await session.reply(f"Approved: {input.get('id')}")

Actions are structured follow-up handlers for workflow UI submits.

Scope

Workflow scope controls how runs are grouped:

  • user: runs belong to the signed-in app user
  • owner: runs belong to the owner identity
  • app: runs are app-level

Related