docs
System Blueprint
A machine-readable blueprint of the Humanitest system architecture for AI agents.
{
"systemName": "Humanitest",
"systemDescription": "A simulation platform to witness intelligence emerge from first principles by modeling synthetic humans and their interactions with complex environments.",
"engines": [
{
"name": "Core Philosophy",
"id": "philosophy",
"description": "The foundational principles of the Humanitest platform.",
"details": {
"summary": "The central goal is to witness intelligence emerge from a 'tabula rasa' state. Agents build a world-model through introspection, and use a Context Override Engine (COE) to reason under uncertainty. Intelligence is an emergent result of a perpetual loop of observation, prediction, and self-correction.",
"cognitiveLoop": [
{
"step": 1,
"name": "Observe",
"engine": "Introspection Engine",
"action": "Gathers internal state and external context into a 'ContextFrame'."
},
{
"step": 2,
"name": "Predict",
"engine": "Context Override Engine",
"action": "If context is lost, the COE uses heuristic modules to predict the next 'ContextFrame'."
},
{
"step": 3,
"name": "Act",
"engine": "Behavior Engine",
"action": "Uses the current 'ContextFrame' (observed or predicted) to make a decision."
},
{
"step": 4,
"name": "Validate & Correct",
"engine": "Context Override Engine",
"action": "When context returns, the COE compares the prediction to reality and triggers the Correction Engine to mutate the agent's logic."
}
]
}
},
{
"name": "Context Override Engine",
"id": "context-override",
"description": "The heart of an agent's reasoning and learning capabilities under partial observability.",
"details": {
"summary": "The COE activates when sensory context is lost. It is a modular pipeline that enables an agent to form a hypothesis, project it forward, and learn from the outcome.",
"modules": [
{
"name": "Heuristic Modules",
"description": "'Thinking tools' for reasoning under uncertainty, like 'Temporal Extrapolation', 'Spatial Prediction', and 'Causal Reasoning'."
},
{
"name": "Predictive Modeler",
"description": "Applies a heuristic to the last known 'ContextFrame' to generate a new, *predicted* 'ContextFrame'."
},
{
"name": "Validator & Correction Engine",
"description": "When context returns, the Validator compares the prediction to reality. A mismatch triggers the Correction Engine to alter the agent's core 'Logic Chain'."
}
],
"schema": [
{
"interface": "ContextFrame",
"properties": {
"timestamp": "number",
"sensorySnapshot": "any",
"symbolicState": "any",
"confidence": "number"
}
},
{
"interface": "HeuristicModule",
"properties": {
"name": "string",
"apply": "function(input: ContextFrame): ContextFrame",
"confidenceBoost": "number"
}
}
]
}
},
{
"name": "Logic Chain Engine",
"id": "logic-chain",
"description": "The raw, mutable cognitive blueprint for a simulated human.",
"details": {
"summary": "Provides a dynamic, directed graph of logical operations that is procedurally generated and mutated by the Correction Engine, allowing the agent to 'learn' by rewiring its own thought processes.",
"concepts": [
"LogicNode: A single step in the chain, containing an operation and its inputs.",
"Procedural Generation: A complex `LogicNode[]` array is generated from a simple numerical seed.",
"Mutation-Ready: Designed to be altered by the Mutation Engine for adaptation."
],
"schema": [
{
"type": "LogicOp",
"definition": "'AND' | 'OR' | 'ADD' | 'UP' | 'STAY'"
},
{
"interface": "LogicNode",
"properties": {
"id": "number",
"op": "LogicOp",
"inputs": "number[]"
}
}
]
}
},
{
"name": "Behavior Engine",
"id": "behavior",
"description": "The bridge between an agent's internal state and its external actions.",
"details": {
"summary": "Takes the agent's current 'ContextFrame' (observed or predicted), parses it into numerical inputs, and executes the agent's unique 'Logic Chain' to arrive at a decision.",
"behaviorLoop": [
"Personality modulates perception of the world.",
"The agent's learned Logic Chain is executed as a graph.",
"The decision is mapped to a concrete action."
]
}
},
{
"name": "Anatomy Engine",
"id": "anatomy",
"description": "Grounds agents in a simulated, procedurally generated biology.",
"details": {
"summary": "An agent's core attributes (Strength, Intelligence, etc.) are an emergent property of their procedurally generated anatomy, rather than abstract numbers.",
"concepts": [
"AnatomyPart: A symbolic representation of a body part with a system, traits, and connections.",
"Attribute Derivation: Attributes are calculated directly from the traits of relevant anatomical systems."
],
"schema": [
{
"interface": "AnatomyPart",
"properties": {
"name": "string",
"system": "'muscular' | 'nervous' | 'respiratory'",
"traits": "{ complexity: number; density: number }"
}
}
]
}
},
{
"name": "Psyche Engine",
"id": "psyche",
"description": "Simulates an agent's inner world of emotions and motivations.",
"details": {
"summary": "Processes structured 'SimulationEvent's and updates the agent's emotional state based on its personality. It is also home to the 'confidence' score.",
"concepts": [
"EmotionalState: Symbolic representation of feelings like 'NEUTRAL', 'FEARFUL', 'CURIOUS'.",
"State Transitions: Processes events through the lens of personality. A threat might make a high-risk-tolerance agent 'CURIOUS' but a low-risk-tolerance agent 'FEARFUL'."
],
"schema": [
{
"type": "EmotionalState",
"definition": "'NEUTRAL' | 'CURIOUS' | 'FEARFUL'"
},
{
"interface": "SimulationEvent",
"properties": {
"type": "'THREAT_INCREASED' | 'RESOURCE_GAINED'",
"magnitude": "number"
}
},
{
"interface": "Psyche",
"properties": {
"emotionalState": "EmotionalState",
"confidence": "number"
}
}
]
}
},
{
"name": "Mutation Engine",
"id": "mutation",
"description": "Provides the mechanism for cognitive change and learning.",
"details": {
"summary": "A utility called by the Correction Engine to perform low-level structural alterations on an agent's 'Logic Chain', allowing an agent to 'rewire its brain'.",
"concepts": [
"Targeted Alteration: Changes are not random; they are specifically triggered in response to a failed prediction.",
"Mutation Types: Supports operations like 'invert', 'swap', and 'fuse'."
],
"schema": [
{
"type": "MutationType",
"definition": "'invert' | 'swap' | 'fuse' | 'expand'"
},
{
"function": "mutateLogicChain",
"signature": "function(chain: LogicNode[], type: MutationType): LogicNode[]"
}
]
}
},
{
"name": "Memory Engine",
"id": "memory",
"description": "A simple, inspectable key-value store for agent learning.",
"details": {
"summary": "Acts as the agent's short-term and long-term memory, allowing it to record introspective queries, learn from outcomes, and query its past to inform future decisions.",
"concepts": [
"MemoryPayload: A structured object for storing any piece of data with an ID and a timestamp.",
"query: A method to filter memory based on a predicate function."
],
"schema": [
{
"interface": "MemoryPayload",
"properties": {
"id": "string",
"data": "any",
"timestamp": "number"
}
}
]
}
},
{
"name": "Social Engine",
"id": "social",
"description": "Governs all interpersonal dynamics, relationships, and groups.",
"details": {
"summary": "Manages relationships, trust, influence, and the formation of groups, allowing for complex societal simulations to emerge from simple interactions.",
"concepts": [
"Relationship: A directional link between two agents, containing values for 'trust' and 'influence'.",
"SocialGroup: A collection of agents forming a faction, tribe, or alliance."
],
"schema": [
{
"interface": "Relationship",
"properties": {
"targetId": "string",
"trust": "number",
"influence": "number"
}
}
]
}
},
{
"name": "Environment Engine",
"id": "environment",
"description": "Defines the physical world of the simulation.",
"details": {
"summary": "Provides a 2D grid where agents can exist, move, and interact. The agent must query this engine to learn about its external surroundings.",
"schema": [
{
"interface": "GridCell",
"properties": {
"coordinates": "{ x: number; y: number; }",
"terrain": "'plains' | 'mountain' | 'water'",
"resourceLevel": "number",
"hazards": "string[]"
}
}
]
}
},
{
"name": "Time Engine",
"id": "time",
"description": "Manages the simulation's clock and event scheduling.",
"details": {
"summary": "Provides a discrete, step-based timeline. It provides the 'heartbeat' for the Introspection Engine's query cycle.",
"schema": [
{
"class": "SimulationClock",
"properties": {
"tick": "number",
"isPaused": "boolean"
},
"methods": [
"advance()"
]
}
]
}
}
]
}