Others

Utility components for annotation, date/time, conversation history, and structured output parsing.

---

Notes

A non-executable annotation node. Use it to add comments, labels, or documentation to sections of the canvas without affecting flow execution.

Use in a flow

1. Add a Notes node anywhere on the canvas.

2. Click the pencil icon on the node to edit its description.

3. Write your notes using plain text or Markdown.

Notes do not accept inputs, produce outputs, or run during flow execution.

Markdown support

The description field supports Markdown: headings (#), bold (**text**), italic (*text*), bullet lists (- item), and links ([text](url)).

Parameters, inputs, and outputs

Notes has no configurable parameters, inputs, or outputs. All content lives in the editable description on the node.

Limitations

• Non-executable — Notes never run and cannot be wired to other components.

• Canvas-only — Content is visible on the flow canvas; it is not passed to agents or downstream steps.

---

Current Date

Outputs the current date and time in a selected timezone. Wire the result into Prompt components, conditions, or agents that need time-aware context.

Use in a flow

1. Add a Current Date node to the canvas.

2. Select a Timezone (defaults to UTC).

3. Connect Current Date output to a Prompt, If-Else, Agent, or other downstream component.

Current Date (Current Date) ──► Prompt ──► Agent

Parameters

Parameter	Default	Hidden	Description
Timezone	UTC	No	Timezone used when formatting the current date and time.

Outputs

Output	Description
Current Date	A message with the current date and time in the selected timezone.

Limitations

• Runtime value — The timestamp is generated when the node runs, not when the flow is saved.

• Message format — Output is a single text message; use a Parser or downstream logic for different formats.

---

Message History

Retrieves stored conversation messages for the current or specified session. Use it to inject chat history into Prompt templates, provide context to an Agent, or pass prior turns to downstream steps.

Use in a flow

1. Add a Message History node to the canvas.

2. Optionally connect an External Memory component; if left empty, messages are read from the platform’s session storage.

3. Adjust filters (sender type, message count, order) as needed.

4. Connect Data, Message, or DataFrame output to the next step.

Message History (Message) ──► Prompt ──► Agent

Parameters

Parameter	Default	Hidden	Description
External Memory	—	No	Optional memory component to read from. When empty, uses platform session message storage.
Sender Type	Machine and User	Yes	Filter by Machine, User, or both.
Sender Name	—	Yes	Filter messages by a specific sender name.
Number of Messages	20	Yes	Maximum number of messages to retrieve. Set to 0 to return none.
Session ID	—	Yes	Session to read from. When empty, uses the current session.
Order	Descending	Yes	Ascending (oldest first) or Descending (newest first).
Template	{sender_name}: {text}	Yes	Format for the Message output. Supports {text}, {sender}, {sender_name}, and other message fields.

Outputs

Output	Description
Data	Retrieved messages as a list of message objects.
Message	Messages formatted as a single text string using the Template.
DataFrame	Messages as a table for filtering or further processing.

Limitations

• Session scope — Without Session ID, only messages from the active session are returned.

• Storage dependency — History depends on messages being stored during the conversation (e.g., Store Messages enabled on Input).

---

Structured Output

Parses text into a defined JSON schema using a connected language model. Use it to extract fields from agent responses, normalize LLM output, or produce consistent data objects for downstream steps.

Use in a flow

1. Add a Structured Output node to the canvas.

2. Connect a Language Model component to Language Model.

3. Wire the text to parse into Input Message (for example, from an Agent or Input).

4. Define fields in Output Schema (name, type, description per field).

5. Connect Structured Output or DataFrame to the next step.

Agent (response) ──► Structured Output ◄── Language Model
Structured Output (Structured Output) ──► Data Operations / Output

Parameters

Parameter	Default	Hidden	Description
Language Model	—	No	LLM used to extract and format structured data. Must support structured output. Required.
Input Message	—	No	Text to parse into the defined schema. Required.
Output Schema	one field row	No	Table defining output fields: Name, Description, and Type (str, int, float, bool, dict). Required.
Format Instructions	built-in default	Yes	System instructions telling the model how to extract and format values.
Schema Name	—	Yes	Name for the output schema model (defaults to OutputModel when empty).

Output Schema fields

Each row in Output Schema defines one property on the output object:

Column	Description
Name	Key in the resulting JSON object.
Description	What the field represents; guides the model during extraction.
Type	Data type: str, int, float, bool, or dict.

Outputs

Output	Description
Structured Output	Parsed results as Data (JSON objects under results).
DataFrame	Parsed results as a table; multiple objects become multiple rows.

Limitations

• Model support — The connected Language Model must support structured output.

• Schema required — Output Schema cannot be empty.

• Extraction quality — Accuracy depends on the model, Format Instructions, and input text clarity.