Step

A Step represents a single phase within a Context. Each step has its own prompt text, completion criteria, available functions, and navigation rules. The AI advances through steps automatically when criteria are met.

You create steps by calling add_step() on a Context object. All setter methods return self for fluent method chaining.

Properties

name

strRequired

Step name. Must be unique within the parent context.

Methods

add_bullets

Add a POM section with bullet points to the step.

add_gather_question

Add a question to this step’s gather_info configuration.

add_section

Add a POM section to the step.

clear_sections

Remove all POM sections and direct text from this step.

set_end

Set whether the conversation should end after this step completes.

set_functions

Set which SWAIG functions are available during this step.

set_gather_info

Enable structured info gathering for this step.

set_history

Control what the model still sees from earlier steps when this step is entered.

set_reset_consolidate

Set whether to consolidate conversation history on context switch.

set_reset_full_reset

Set whether to completely replace the system prompt on context switch.

set_reset_system_prompt

Set a new system prompt for context switching from this step.

set_reset_user_prompt

Set a user message to inject when this step triggers a context switch.

set_skip_to_next_step

Automatically advance to the next step without evaluating criteria.

set_skip_user_turn

Skip waiting for user input after this step completes.

set_step_criteria

Define when this step is considered complete.

set_text

Set the step’s prompt text directly.

set_valid_contexts

Set which contexts the agent can navigate to from this step.

set_valid_steps

Set which steps the agent can navigate to from this step.

Example

from signalwire import AgentBase
agent = AgentBase(name="my-agent", route="/agent")
contexts = agent.define_contexts()
ctx = contexts.add_context("default")
step = ctx.add_step("collect_info")
# Structured prompt with POM sections
step.add_section("Task", "Collect the caller's account information.")
step.add_bullets("Required Information", [
    "Full legal name",
    "Account number (10 digits)",
    "Reason for calling"
])
# Completion criteria
step.set_step_criteria(
    "Complete when all three pieces of information have been collected "
    "and confirmed with the caller."
)
# Only allow relevant functions
step.set_functions(["lookup_account", "verify_identity"])
# Navigation
step.set_valid_steps(["process_request", "escalate"])
agent.serve()