Chat Trigger¶

Trigger

The Chat Trigger receives messages from external chat clients via the message gateway. When a user sends a message through a connected chat interface, the gateway forwards it to Pipelit, and the trigger fires to start a new workflow execution.

Component type: trigger_chat

Ports¶

Outputs¶

Inputs¶

This component has no input ports. It is an entry point.

Configuration¶

The Chat Trigger requires no configuration. Add it to the canvas and connect it to downstream nodes.

Usage¶

1. Drag a Chat Trigger node onto the canvas from the Node Palette (under Triggers).
2. Connect its output to an Agent, Code, or any other downstream node.
3. Connect an AI Model sub-component to your agent.
4. Open the ChatPanel in the workflow editor (bottom panel).
5. Type a message and press Enter.

Each message sent through the ChatPanel creates a new workflow execution scoped to the Chat Trigger. The user's message text is available on the text output port, and the full payload (including metadata) on the payload port.

Accessing Chat Input Downstream¶

In your agent's system prompt or any node's configuration, reference the chat input using Jinja2 expressions:

{{ trigger.text }}

Or reference the specific trigger node by its ID:

{{ trigger_chat_abc123.text }}

Conversation Memory¶

When the Chat Trigger connects to an Agent with Conversation Memory enabled, the agent maintains conversation history across multiple chat messages. This gives the agent continuity -- it remembers what was said in previous messages within the same workflow.

The thread ID for conversation memory is constructed from the user's profile ID and the workflow ID, so each user gets their own conversation thread per workflow.

Example¶

A minimal chat-based workflow:

graph LR
    CT[Chat Trigger] --> AG[Agent]
    AM[AI Model] -.-> AG

    style CT fill:#f97316,color:white
    style AM fill:#3b82f6,color:white

1. Chat Trigger receives the user's message.
2. Agent processes the message using the connected AI Model, reasons through the response, and produces output.
3. The response is displayed back in the ChatPanel.

Multi-trigger workflow¶

A Chat Trigger and a Manual Trigger can both feed the same agent:

graph LR
    CT[Chat Trigger] --> AG[Agent]
    TM[Manual Trigger] --> AG
    AM[AI Model] -.-> AG

    style CT fill:#f97316,color:white
    style TM fill:#f97316,color:white
    style AM fill:#3b82f6,color:white

In this layout, the same agent handles messages from both chat and manual triggers. The {{ trigger.text }} expression works regardless of which trigger fired.

API Endpoint¶

Chat messages are sent via:

POST /api/v1/workflows/{slug}/chat/

Request body:

{
  "text": "Hello, can you help me with something?",
  "trigger_node_id": "trigger_chat_abc123"
}

The trigger_node_id field is optional. If omitted, Pipelit uses the first trigger_chat node found in the workflow.