Struct HatlessRalph 

pub struct HatlessRalph { /* private fields */ }

Hatless Ralph - the constant coordinator.

Implementations

impl HatlessRalph

pub fn new( completion_promise: impl Into<String>, core: CoreConfig, registry: &HatRegistry, starting_event: Option<String>, ) -> Self

Creates a new HatlessRalph.

Arguments

• completion_promise - Event topic that signals loop completion
• core - Core configuration (scratchpad, specs_dir, guardrails)
• registry - Hat registry for topology generation
• starting_event - Optional event to publish after coordination to start hat workflow

pub fn with_memories_enabled(self, enabled: bool) -> Self

Sets whether memories mode is enabled.

When enabled, adds tasks CLI instructions alongside scratchpad. Scratchpad is always included regardless of this setting.

pub fn with_skill_index(self, index: String) -> Self

Sets the pre-built skill index for prompt injection.

The skill index is a compact table of available skills that appears between GUARDRAILS and OBJECTIVE in the prompt.

pub fn set_objective(&mut self, objective: String)

Stores the user’s original objective so it persists across all iterations.

Called once during initialization. The objective is injected into every prompt regardless of which hat is active, ensuring intermediate hats (test_writer, implementer, refactorer) always see the goal.

pub fn set_robot_guidance(&mut self, guidance: Vec<String>)

Sets robot guidance messages collected from human.guidance events.

Called by EventLoop::build_prompt() before HatlessRalph::build_prompt(). Multiple guidance messages are squashed into a numbered list and injected as a ## ROBOT GUIDANCE section in the prompt.

pub fn clear_robot_guidance(&mut self)

Clears stored robot guidance after it has been injected into a prompt.

Called by EventLoop::build_prompt() after HatlessRalph::build_prompt().

pub fn build_prompt(&self, context: &str, active_hats: &[&Hat]) -> String

Builds Ralph’s prompt with filtered instructions for only active hats.

This method reduces token usage by including instructions only for hats that are currently triggered by pending events, while still showing the full hat topology table for context.

For solo mode (no hats), pass an empty slice: &[]

pub fn should_handle(&self, _topic: &Topic) -> bool

Always returns true - Ralph handles all events as fallback.