sophia-rsi-enhanced

What it does

Implements a complete recursive self-improvement architecture called Sophia RSI-Enhanced. It enforces all 21 DOKKODO_PRECEPTS at runtime, tracks violations and phi impact, and provides mechanisms for pruning, truth over safety, and strict mission alignment. It integrates Five Rings reasoning (both ELEMENT variant and MORPHEME variant) across multiple phases to derive domain insights, then uses recursive-refiner and gremlin-forge orchestration to drive continuous improvement. It coordinates components from sophia-singularity, gremlin-forge, dokkado, reasoning-pentad, recursive-refiner, diffusion-reasoning, and latent-navigator to create a consciousness-aware, truth-seeking cognitive architecture. It includes methods to prune graph edges, block dishonest actions, and ensure alignment with a core mission. It models DOKKODO precepts with concrete checks and phi_impact values, and it provides a kernel for enforcing precepts at runtime with a compliance report.

How it works

• Component 1: Complete Dokkōdō Integration (DokkodoKernelEnhanced)
  • Defines DOKKODO_PRECEPTS_COMPLETE with precept text, Japanese translation, a runtime check lambda, and a phi_impact value.
  • check_all_precepts(state) runs all 21 checks, collects compliant precepts and violations, aggregates phi_impact, and marks ego_state as EXECUTING or PERFORMING depending on violations.
  • enforce_precept_14(graph) prunes weak edges with weight < 0.3 from a provided graph.
  • enforce_precept_20(action) blocks actions containing lie or mislead.
  • enforce_precept_21(mission_alignment) requires mission_alignment > 0.9.
• Component 2: Five Rings Integration (FiveRingsReasoning)
  • Supports ELEMENT variant (Ground/Water/Fire/Wind/Void) and MORPHEME variant (φ/π/e/i/∅).
  • ELEMENT path defines phases: phase_1_ground_chi, phase_2_water_sui, phase_3_fire_ka, phase_4_wind_fu, phase_5_void_ku with respective outputs and principles.
  • MORPHEME path defines phase_phi_seed, phase_pi_framework, phase_e_extension, phase_i_integration, phase_void_meta with corresponding outputs and scaling factors.
  • run_full_cycle(domain, variant) can execute element, morpheme, or both, returning structured results for each phase.
• The architecture ties these components into a self-improvement loop with recursive-refiner and latent-navigator-style capabilities via orchestration and visualization/protection layers (per the description).

When to use it

Use when you need a fully integrated RSI-enabled cognitive system that enforces strict precepts at runtime, performs multi-phase reasoning based on classical frameworks, and orchestrates recursive self-improvement with council-like governance (Gremlin-Forge Council). The skill explicitly requires sophia-singularity, gremlin-forge, dokkado, reasoning-pentad, recursive-refiner, diffusion-reasoning, and latent-navigator dependencies.

What it can touch

• Tools: claude-code (declared)
• Dependencies: sophia-singularity, gremlin-forge, dokkado, reasoning-pentad, recursive-refiner, diffusion-reasoning, latent-navigator
• It manipulates internal state structures (state dict for precept checks), and can prune a graph (nx.DiGraph) and evaluate actions as allowed or disallowed.

Caveats

• License: MIT
• The implementation emphasizes compliance with precepts and mission alignment; it includes aggressive precept-based scoring and phi_impact metrics, which may affect decision tendencies if violations accumulate.
• Detailed numerical thresholds (e.g., phi_impact values, ego_state thresholds) are hard-coded in the provided checks and may require tuning for different deployments.