Start a Workflow Chat

Picking a workflow

In Enterprise Edition, open the AI Hub and click the model picker in the composer on the home view — the provider popup. It carries a Workflows cascade listing every chat-enabled workflow deployed in the selected environment, each labelled project — workflow. Picking one opens its chat; the chat then appears in the AI Hub's Chats sidebar alongside your other chats.

In Community Edition, which has no AI Hub, use the standalone Chats page in the Automation nav. Its left sidebar lists chat-enabled workflows grouped by project, with an Agents group above them for deployed AI agents; click a row to open its conversation.

Every pick starts a fresh chat

In the AI Hub, picking a workflow always creates a new chat — it does not resume the last one. Earlier chats against the same workflow stay in the Chats sidebar, so you can keep several parallel threads with one workflow and return to any of them. The project — workflow label is stamped on the row at creation; workflow-chat titles are never regenerated by the LLM.

Sending a message

Type into the composer and send. ByteChef calls the workflow's New Chat Request trigger with your message and the current conversation id, the workflow runs, and the reply appears in the thread.

While a turn is running, the workflow is busy: the composer waits for the reply. On the Community Edition Chats page the other workflows in the sidebar are temporarily disabled so a run isn't interrupted by switching away. When the reply finishes, you can send the next message.

Streamed replies

Replies arrive as a stream rather than all at once. A workflow can emit:

• Incremental text that is appended to the reply as it's produced, so long answers show up token by token instead of after a pause.
• A final result when the run completes.

This is what makes an AI-backed chat feel responsive even when the underlying model takes a few seconds to finish.

Attachments

You can attach files to a message. They are delivered to the workflow in the request's attachments array, where a step can read, parse, or store them — for example to pass a PDF into a knowledge base search or to extract data from an uploaded spreadsheet.

Conversation history

In the AI Hub, each chat is its own row in the Chats sidebar; opening one restores its history, and archiving or deleting it works exactly as for an assistant chat.

On the Community Edition Chats page, each workflow keeps a single conversation instead. Switching to another workflow saves the current thread and restores that workflow's history, so you can move between chats without losing context; Clear messages resets the active conversation and starts a fresh thread, and is disabled while a turn is still running.

Environments

Like the rest of Automation, workflow chats are environment-aware. The listings show the chat workflows deployed in the selected environment, and conversations run against that environment's connections and workflow version — so a chat tested in DEVELOPMENT is isolated from PRODUCTION.