How do workflows and connections work?
A workflow is a set of components joined by typed connections that pass data between them. A component may also have configuration fields, such as text, selections, or credentials. A normal edge needs at least one matching port type; it does not verify settings or runtime success. See AI Flowi Workflow.
By Ai Flowi · Technical review: Janice Lee, Technical Support, Ai Flowi.
Published · Updated
What is workflow with an example?
A workflow joins components through typed connections. For example, a Playground chat workflow sends a Message from Chat Input to a compatible AI model, then sends its Message response to Chat Output. Configuration fields hold settings such as credentials; connection inputs and outputs determine which data types can pass between components.
| Term | What to inspect |
|---|---|
| Component | Its purpose, required configuration, connection inputs, and outputs. |
| Configuration input | A value you enter or select, such as a match phrase or credential. |
| Connection input | The types it accepts from another component's output. |
| Output | The types it produces for downstream components. |
Common declared types include Message, JSON, Table, Data, LanguageModel, and Tool. For a normal typed edge, at least one source output type must match a type accepted by the destination input. The component catalog lists the documented interfaces.
Check a connection
- Find the source component's output name and declared type.
- Find the destination component's connection input and accepted types.
- Connect the ports when the types overlap. Configure required fields and credentials separately.
- Run a small test and inspect the first component that does not produce the expected data.
For example, the first Playground workflow passes a Message from Chat Input to a compatible model, then a Message response to Chat Output. The webhook tutorial shows why a JSON webhook output is converted to Message before Structured Output.
If a connection fails
Check the exact output and input types first. If the ports connect but execution fails, inspect required settings and credentials, then test the preceding stage again. Common errors lists further checks. After importing a workflow or updating a component, use the workflow lifecycle guide to inspect affected connections and retest them.
Shortcuts and canvas editing
In AI Flowi Workflow, open Settings → Shortcuts to inspect the current keyboard bindings. Double-click a shortcut to record a replacement; a combination already assigned to another action is rejected. Use Restore to return the shortcut list to its defaults. Custom bindings are stored in this browser.
Use the configured Group, Copy, Paste, Duplicate and Delete shortcuts to organise or reuse selected canvas components. Group checks the selected connections and reports an invalid selection instead of creating a group. Undo and Redo move between recorded canvas states when history is available. Undo, Redo, Group, Paste, Duplicate and Delete are blocked while the workflow is locked, while an agent is working on it, or during a version preview.
Playground and attachments
The Playground provides a chat surface for testing a workflow. Its standalone page requires a public workflow with recognised inputs or outputs. Use the attachment control to add files to a message; attached files appear as previews and their uploaded paths are sent with the message. Test whether your configured workflow can process the file you attach.
Stored messages and session boundaries
Settings → Messages opens the stored-message table. A session-specific view filters messages by session and, when supplied, by workflow. Outside the standalone Playground page, you can update message text and delete selected message rows. The standalone Playground page disables message editing and deletion in this table. Deletion targets the selected message IDs; it is separate from deleting the workflow.
Related
What is AI Flowi Workflow? · Your first workflow · Components