This document describes the structural layout, component responsibilities, and processing flow of the Metacortex v0.1 system. The system is designed as an experimental neural-symbolic architecture that interfaces with external LLMs (e.g., Mistral via Ollama) and applies Fourier-based structural decomposition to reason over axiomatic semantic dimensions.
development/
├── metacortex/
│ ├── v0.1/
│ │ ├── core/ # Core processing logic
│ │ ├── api/ # FastAPI endpoints
│ │ ├── modules/ # Optional plugins or subcomponents
│ │ └── README.md # This documentation
├── compass/
│ ├── v0.1/ # Semantic decomposition logic (pre-superposition)
-
Purpose: Acts as the external interface to receive queries.
-
Framework: FastAPI
-
Responsibilities:
- Accept text input via POST.
- Normalize request.
- Forward to
core/metacortex_engine.py.
-
Purpose: Central controller for processing, transformation, and reasoning.
-
Responsibilities:
- Build symbolic and structural network.
- Integrate Fourier-based vector processing.
- Trigger
SuperpositionEngine.
-
Purpose: Analyze and process overlapping vector structures.
-
Responsibilities:
- Accept semantic vectors.
- Apply interference logic (not mean-based).
- Respect axis-defined weighting via axioms.
- Forward to
token_engine.
-
Purpose: Generate response text from interfered vector space.
-
Responsibilities:
- Synthesize meaningful output.
- Log key semantic anchors.
- Return final result.
User Input → FastAPI (api/routes.py)
↓
Metacortex Engine (core/metacortex_engine.py)
↓
Superposition Processing (core/superposition_response.py)
↓
Token Generation (core/token_engine.py)
↓
Response → Console / API Response
- Q-Transform Layer: Quantization and decoherence of vector states.
- Goal Anchoring: Matching against long-term goal vector (COMPASS interface).
- Semantic Mirror: Add module to reflect human conceptual axis back into feedback loop.
- Interactive Tuning: Interface to adjust weightings per Axiom manually.
- Ollama: Container-based Mistral interface for external LLM reasoning.
- COMPASS: Handles semantic decomposition and axiom grounding.
- ArangoDB: Will store historical responses and vector resonance snapshots for learning.
All components are subject to modular replacement or upgrade. This version (v0.1) serves as a minimal working scaffold for symbolic-structural reasoning pipelines.
Author: dwpplumb License: Experimental, CC-BY 4.0 or similar