Conscious Agent Architecture (Syntrometrie/RIH Implementation)

graph TB %% Global Styling classDef agent fill:#E8DAEF,stroke:#A569BD,stroke-width:2px,color:black,font-size:14px classDef module fill:#D5F5E3,stroke:#58D68D,stroke-width:2px,color:black,font-size:14px classDef metric fill:#FCF3CF,stroke:#F4D03F,stroke-width:2px,color:black,font-size:14px classDef state fill:#D6EAF8,stroke:#5DADE2,stroke-width:1px,color:black,font-size:14px classDef io fill:#FADBD8,stroke:#EC7063,stroke-width:2px,color:black,font-size:14px classDef data fill:#FEF9E7,stroke:#F8C471,stroke-width:1px,color:black,font-size:12px %% Inputs subgraph "Inputs" In_State["State (s_t, 12D)"]:::state In_Reward["Reward (r)"]:::data In_History["State History (H_t)"]:::state end %% Agent Core Processing (Forward Pass) subgraph "Agent Core (Forward Pass)" Agent["ConsciousAgent.forward"]:::agent In_State --> Encoder["Encoder (Linear)"]:::module Encoder --> GNN_Input["Encoded State"]:::state GNN_Input --> BuildGraph["Build Graph"]:::module BuildGraph --> GraphData["Graph Data (Nodes, Edges)"]:::data GraphData --> RunGNN["Run GNN (Layers)"]:::module GNN_Input --> RunGNN RunGNN --> BeliefEmb["Belief Embedding"]:::state BeliefEmb --> SelfReflect["Self Reflect Layer"]:::module SelfReflect --> TranscendentEmb["Transcendent Embedding"]:::state %% Metric Calculation Branch TranscendentEmb --> ComputeMetrics["Compute Metrics"]:::metric In_History --> ComputeMetrics ComputeMetrics --> GeoMetrics["g_ik, Gamma, R"]:::data ComputeMetrics --> RIHMetrics["I(S), rho, tau_t"]:::data ComputeMetrics --> OtherMetrics["Stability, Zeta, t, dA/dt, Consistency"]:::data GeoMetrics & RIHMetrics & OtherMetrics --> AllMetrics["Forward Metrics"]:::data %% Value Prediction Branch BeliefEmb --> ValueHead["Online Value Head"]:::module ValueHead --> ValuePredS["Value V(s)"]:::data %% Emotion Branch In_State --> EmoState["Emotion State (R1-6)"]:::state EmoState & In_Reward --> EmoModule["Emotional Module"]:::module EmoModule --> Emotions["Updated Emotions (R1-6)"]:::state %% Qualia Branch TranscendentEmb --> QualiaHead["Qualia Output Head"]:::module QualiaHead --> QualiaFeatures["Qualia Features (R7-12)"]:::state %% Assembly & Feedback Emotions & QualiaFeatures --> AssembleState["Assemble Full State"]:::module AssembleState --> FullStateOut["New Full State (12D)"]:::state TranscendentEmb --> FeedbackNet["Feedback Network"]:::module FeedbackNet --> FeedbackSignal["Feedback Signal"]:::data end %% Outputs & Interactions subgraph "Outputs & Interactions" ForwardResult["Forward Result (Tuple)"]:::data Emotions --> ForwardResult BeliefEmb --> ForwardResult RIHMetrics --> ForwardResult OtherMetrics --> ForwardResult GeoMetrics --> ForwardResult FeedbackSignal --> ForwardResult FullStateOut --> ForwardResult ValuePredS --> ForwardResult ForwardResult --"Emotions"--> Avatar["Avatar Expressions"]:::io ForwardResult --"Attention Score"--> GenerateResponse["agent.generate_response"]:::agent GenerateResponse --"Context"--> GPT["TransformerGPT"]:::module GPT --> DialogueOut["Dialogue Response"]:::io ForwardResult --"Qualia (R7-12)"--> EnvFeedback["env.update_qualia_feedback"]:::io end %% Learning Loop (Conceptual) subgraph "Learning Loop (Async)" Memory["Replay Memory (PER)"]:::data LearnTask["_run_learn_task"]:::agent --"Samples Batch"--> Memory BatchData["Batch (s, b, r, s', d)"]:::data --> Learn["Agent.learn"]:::agent Indices["Indices"]:::data --> Learn Weights["PER Weights"]:::data --> Learn Learn --"Uses Online Nets (s)"--> AgentCore["Agent Core"] Learn --"Uses Target Value Net (s')"--> TargetNet["Target Value Head"]:::module Learn --"Calculates"--> Loss["Combined Loss (Value + RIH)"]:::metric Learn --"Updates"--> AgentCore Learn --"Updates"--> TargetNet Learn --"Updates"--> Memory[("PER Priorities")] Learn --"Adds Processed Batch"--> Memory end %% Connect Inputs to Agent In_State --> Agent In_Reward --> Agent In_History --> Agent