Identify User¶

The Identify User component identifies who is interacting with the workflow and loads their stored context. Unlike Memory Read and Memory Write, this is a direct workflow node -- it runs as a step in the graph rather than as an agent tool.

Ports¶

Inputs¶

Outputs¶

Configuration¶

The Identify User component has no additional configuration. It infers the channel and extracts user identity from the trigger payload automatically.

Usage¶

1. Add an Identify User node from the Node Palette (Memory category)
2. Place it between the trigger node and the agent node
3. Connect the trigger's output to Identify User's input
4. Connect Identify User's output to the agent's input

The component examines the trigger payload and automatically detects the channel type. It then extracts channel-specific identifiers:

Channel auto-detection¶

If the channel input is not explicitly provided, the component infers it from the trigger payload structure:

• Contains message.from -- detected as telegram
• Contains webhook_id -- detected as webhook
• Contains source: "manual" -- detected as manual
• Contains source: "chat" -- detected as chat

State patch¶

In addition to its output ports, the Identify User component writes a _state_patch that merges user context into the global workflow state under the user_context key. This makes user information available to all downstream nodes via Jinja2 expressions without needing explicit edges.

Example¶

A workflow that personalizes responses based on user identity:

flowchart LR
    T[Telegram Trigger] --> IU[Identify User]
    IU --> A[Agent]
    M[AI Model] -.->|model| A
    MR[Memory Read] -.->|memory| A

On first interaction from a Telegram user:

Identify User output:
{
  "user_id": "tg_12345678",
  "user_context": {"is_new": true, "facts": [], "history": []},
  "is_new_user": true
}

On subsequent interactions:

Identify User output:
{
  "user_id": "tg_12345678",
  "user_context": {
    "is_new": false,
    "facts": [
      {"key": "user_name", "value": "Alice"},
      {"key": "preferred_language", "value": "English"}
    ],
    "history": [...]
  },
  "is_new_user": false
}

The agent can then reference user context in its system prompt:

{% if user_context.is_new %}
Welcome! This is your first time here.
{% else %}
Welcome back, {{ user_context.facts | selectattr("key", "equalto", "user_name") | map(attribute="value") | first }}!
{% endif %}