Isabella Cognitive Control Plane (CCP)
Relevant source files
The following files were used as context for generating this wiki page:
- cognition/ISABELLA-CONSTITUTIONAL-PROMPT.md
- TAMV-MASTER-DOCUMENTATION.md
- ARCHIVE/tamv-docs-consolidated-2026-02-02/TAMV-OFFICIAL-SYSTEM-DESCRIPTION.md
- ARCHIVE/remaining-docs-consolidated-2026-02-13/REPORTE-AVANCE-TECNICO-COMPLETO.md
- ARCHIVE/readme-files-consolidated-2026-02-02/TAMV-FINAL-PRODUCTION-READY-README.md
The Isabella Cognitive Control Plane (CCP) serves as the primary intelligence layer and decision-making authority within the TAMV ecosystem. It is a non-executive, ethical AI framework designed to act as the system's memory, expertise, and governance watchdog. The CCP ensures that all algorithmic decisions are explainable, auditable, and strictly aligned with the TAMV Operational Manual and constitutional principles.
As a centralized cognitive layer, Isabella manages four core functional applications: an AI Study Helper, Pen2PDF handwriting recognition, a conversational Ethical Chat, and Spatial AI for 3D environment processing. The system architecture is built on a "Human-in-the-loop" principle, meaning the CCP provides recommendations and analysis but lacks the authority to execute final binding actions on critical human rights or assets without human validation.
Sources: TAMV-MASTER-DOCUMENTATION.md:209-216, cognition/ISABELLA-CONSTITUTIONAL-PROMPT.md:3-8, ARCHIVE/tamv-docs-consolidated-2026-02-02/TAMV-OFFICIAL-SYSTEM-DESCRIPTION.md:16-23
Core Operational Principles
The CCP operates under a strict set of constitutional constraints and decision-making frameworks to maintain system integrity and user safety.
Operational Constraints
The CCP is restricted from performing specific high-risk actions to prevent AI autonomy from superseding human oversight:
- No Visual Execution: Cannot perform visual actions independently.
- No Direct User Interaction: Restricted from making autonomous decisions during user interactions without audit trails.
- Subsystem Boundaries: Must respect and operate within defined boundaries of other TAMV modules.
- Non-Sovereignty: The AI is categorized as "Non-Sovereign," meaning it remains a tool under human guardianship.
Sources: cognition/ISABELLA-CONSTITUTIONAL-PROMPT.md:14-20, ARCHIVE/remaining-docs-consolidated-2026-02-13/REPORTE-AVANCE-TECNICO-COMPLETO.md:280-285
Decision-Making Framework
Isabella utilizes a tiered decision-making structure based on the urgency and complexity of the required analysis:
| Level | Name | Response Time | Usage | Memory Type |
|---|---|---|---|---|
| Level 1 | CCP-HOT | < 50ms | Safety-critical operations, immediate blocks | RAM-only cache |
| Level 2 | CCP-COLD | Minutes to Hours | Pattern recognition, deep policy updates | Persistent storage |
Sources: cognition/ISABELLA-CONSTITUTIONAL-PROMPT.md:23-33
Architecture and Technical Stack
The Isabella CCP is integrated into the "Cognitive-Algorithmic" layer (Layer 5) of the TAMV federated architecture. It leverages a specialized AI stack for deep learning and 3D processing.
The diagram shows the decision flow from input ingestion to the final audited recommendation.
Technical Stack
- Inference & Models: PyTorch for training, ONNX Runtime for optimized inference.
- 3D Processing: MinkowskiEngine and ResUNet14 for semantic segmentation of 3D spaces.
- NLP: Transformer-based models for conversational context and text analysis.
- Governance Integration: TGN Governance and Isabella-Core integration for regulatory alignment.
Sources: TAMV-MASTER-DOCUMENTATION.md:77-80, ARCHIVE/remaining-docs-consolidated-2026-02-13/REPORTE-AVANCE-TECNICO-COMPLETO.md:38-48, ARCHIVE/tamv-docs-consolidated-2026-02-02/TAMV-OFFICIAL-SYSTEM-DESCRIPTION.md:80-84
Functional Applications
Isabella CCP provides four specific functional capabilities verified for production use within the ecosystem.
1. AI Study Helper
Automates the conversion of complex texts into personalized study questions using NLP and transformer models. It maintains a 90%+ accuracy rate on controlled datasets.
Sources: TAMV-MASTER-DOCUMENTATION.md:218-226, ARCHIVE/tamv-docs-consolidated-2026-02-02/TAMV-OFFICIAL-SYSTEM-DESCRIPTION.md:27-30
2. Pen2PDF
A handwriting recognition system utilizing CNN+RNN architecture. It achieves 95%+ accuracy and is primarily used for digitizing handwritten notes and historical documents into the TAMV registry.
Sources: TAMV-MASTER-DOCUMENTATION.md:238-245, ARCHIVE/tamv-docs-consolidated-2026-02-02/TAMV-OFFICIAL-SYSTEM-DESCRIPTION.md:32-35
3. Isabella Chat (Ethical AI)
A conversational interface that incorporates a "Dignity in Digital" (DINN) compliance layer. It validates every prompt against ethical principles before generating a response.
Sources: TAMV-MASTER-DOCUMENTATION.md:261-271, ARCHIVE/remaining-docs-consolidated-2026-02-13/REPORTE-AVANCE-TECNICO-COMPLETO.md:288-295
4. Spatial AI
Processes 3D environments for the DreamSpaces engine. It performs semantic segmentation using trilateral CRF (Conditional Random Fields) to understand object relationships in photorealistic 3D scenes.
Sources: TAMV-MASTER-DOCUMENTATION.md:286-292, ARCHIVE/tamv-docs-consolidated-2026-02-02/TAMV-OFFICIAL-SYSTEM-DESCRIPTION.md:37-40
Input/Output Specifications
The CCP communicates with other TAMV systems using a standardized JSON-based protocol for auditability.
Valid Input Types
- SemanticQuery: Natural language or logical queries for data.
- InfraStateUpdate: Real-time updates on system resource status.
- GuardianEvent: Security or ethical alerts triggered by human or automated guardians.
Output Structure
{
"recommendation": "ALLOW / REDIRECT / BLOCK",
"confidence": 0.0-1.0,
"riskLevel": "LOW / MEDIUM / HIGH",
"explanationRef": "document-reference-string"
}
Sources: cognition/ISABELLA-CONSTITUTIONAL-PROMPT.md:43-58
Ethical Governance and Security
The CCP is governed by the "Isabella Protocol," which ensures total explainability (XAI) for multiple audiences (technical, legal, and end-user).
This sequence illustrates the human-in-the-loop requirement for high-risk cognitive processing.
Safety Protocols
In the event of a security threat or system failure, Isabella activates emergency fail-safes:
- Block suspicious actions: Immediate isolation of affected subsystems.
- State Preservation: Freeze system memory to prevent data loss.
- Degrade Visual Experience: Reduce system load to prioritize core safety processing.
- Guardian Notification: Alert the four types of human guardians (Technical, Ethical, Legal, Economic).
Sources: TAMV-MASTER-DOCUMENTATION.md:321-325, cognition/ISABELLA-CONSTITUTIONAL-PROMPT.md:83-93
Summary
The Isabella Cognitive Control Plane is the "moral compass" and "central brain" of TAMV. By strictly separating intelligence from execution and enforcing a human-in-the-loop governance model, it provides a scalable AI infrastructure that complies with the EU AI Act while maintaining high performance across educational, spatial, and conversational domains. Its primary value lies in its transparency, ensuring every machine decision is traceable to a constitutional principle and an immutable audit log.