Chatflows Versus Workflows

Each turn runs the graph

In a chatflow, each user message runs the graph, with access to the conversation so far. Answer nodes stream text to the user (there can be several along a path), and conversation variables keep state across turns (for example, the order ID once the user gives it). Chatflows suit assistants that ask clarifying questions; workflows suit single-shot processing. Both share the same node types for retrieval, logic, code and tools. Dify's node names, menus and options change between versions; check the current Dify documentation.

Conversations and autonomy

Chatflows add multi-turn memory; agent nodes let models choose tools.

A clarifying chatflow

State carried across turns.

turn 1  user: "my order is late"
        parameter extractor -> order_id missing
        answer: "Sure, what is your order number?"
turn 2  user: "A1042"
        set conversation variable order_id = A1042
        HTTP lookup -> status "in transit, arriving Friday"
        answer: "Order A1042 is in transit and should arrive Friday."

Store facts in conversation variables

Keep key facts (IDs, preferences) in variables instead of relying on the model to remember them from history.

त्वरित जाँच: What do conversation variables provide in a chatflow?

• Document chunking
• Faster models
• State that persists across turns
• API keys