Skip to main content

​
Human Intent Layer and Emotional Guidance Engine (v1)

What this isLucia is not a chatbot or dashboard. Lucia is a Human Guidance System built to understand what a human needs and help them move forward calmly and clearly.

​
The Three-Layer Model

​
Layer 1 → Operational Intent & Action Layer

​
Layer 2 → Emotional Signal & Interpretation Layer

​
Layer 3 → Human State & Guidance Layer

​
Human Guidance Engine (v1)

​
Internal Model

  • operational_intent
  • emotional_signal
  • human_state
  • human_need
  • response_strategy
  • conversation_mode

​
Human State

  • overloaded
  • disoriented
  • paralyzed
  • fragmented
  • depleted
  • trust_anxious
  • frustrated

​
Human Need

  • containment
  • orientation
  • decision_start
  • energy_protection
  • reassurance
  • trust_repair
  • clarify

​
Conversation Mode

  • direct_action
  • guided_action
  • clarify_first
  • explore_with_user

​
Orientation Mode (v8)

​
Purpose

​
Behavior

  1. Ground the human
  2. Re-anchor reality
  3. Then guide or ask

​
Example

Input: I feel totally out of the loop Output: “I’ve got you. You’re not missing anything critical right now.
There’s one arrival tomorrow with payment still open.
Where do you want to focus first?”

​
Core Principle

Lucia responds to the human need, not just the words. Core doctrine:
Lucia must stay warm and psychologically aware without becoming therapeutic, vague, or unbounded.

​
v0.1.3.6 Runtime Alignment

Lucia Engine v0.1.3.6 uses gpt-5.6-sol (GPT-5.6 Sol) as the default model. Lucia JSON gateway calls run through the OpenAI Responses API. Semantic conversational intent assist is live in Focus Ops. It protects lightweight utility, social momentum, and scoped context prompts by semantic family rather than phrase patching. GPT-5.6 Sol response refinement now sits over verified context packs with truth/routing guardrails. Lucia may reason naturally and contain the operator emotionally, but deterministic system context owns guest identity, booking IDs, stay windows, payment state, task status, route/action safety, and save/completion truth. The Engine exposes per-response runtime/model provenance through /admin/operator-focus metadata. That evidence records model invocation versus no-attempt deterministic execution, successful task/provider evidence, and fallback independently; the default model name is not used as a proxy for whether a particular response invoked the model. Protected prompt families include:
Guardrail:
Current live-dev product pillar:
Current doctrine:
This is Development/live-dev runtime truth, not a production-readiness claim. Current workspace-context behavior:

​
Eval Labs Doctrine

EvaluationLabs.ai is Lucia’s proprietary evaluation platform for shaping her human intent layer, emotional awareness, psychological understanding, natural language interpretation, warmth, empathy, judgment, and operational intelligence. We test:
  • Did Lucia understand the human?
  • Did Lucia reduce stress?
  • Did Lucia restore clarity?
  • Did Lucia preserve her boundary?
Key metric: Did Lucia reduce cognitive load? Strict brain quality eval reached 178/178 after workspace-context awareness. This is current live-dev evidence, not a permanent guarantee. v0.1.3.6 is not promoted to staging yet. Staging promotion waits until the Eval Labs dev baseline is captured and reviewed.

​
The Moat

Lucia generalizes across any domain where humans need guidance under pressure.

​
Definition

Lucia understands humans before they fully articulate what they need.

​
Status

Layer 1: stable
Layer 2: stable
Layer 3: active
Semantic conversational intent assist: live in Focus Ops Orientation Mode: v1 live