Isabella AI Capabilities

Relevant source files

The following files were used as context for generating this wiki page:

• 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
• ARCHIVE/remaining-docs-consolidated-2026-02-13/TAMV-COMPLETE-IMPLEMENTATION-PLAN.md

Isabella AI is the core ethical artificial intelligence framework of the Territorio Autónomo de Memoria Viva (TAMV) ecosystem. It is designed as a non-executive system, meaning it functions as an explanation and recommendation engine that requires explicit human validation for critical actions. Isabella ensures that the technical infrastructure remains aligned with human dignity, transparency, and the Ethical Framework.

The framework serves multiple roles across the TAMV ecosystem, ranging from personalized education through the UTAMV platform to spatial understanding within XR DreamSpaces. By utilizing explainable AI (XAI) and multi-layer supervision, Isabella provides a sovereign alternative to extractive digital platforms.
Sources: TAMV-MASTER-DOCUMENTATION.md:213-219, TAMV-OFFICIAL-SYSTEM-DESCRIPTION.md:21-25

---

Core Functional Applications

Isabella AI is structured into four primary functional modules, each targeting a specific pillar of the TAMV socio-technical infrastructure.

1. AI Study Helper

The AI Study Helper is integrated with UTAMV (TAMV University). It uses Natural Language Processing (NLP) with transformer models to convert raw text into personalized assessments and study materials.

• Target Accuracy: 90%+ on controlled datasets.
• Key Logic: Preprocesses text to extract key concepts and generates difficulty-scaled questions.
  Sources: TAMV-MASTER-DOCUMENTATION.md:223-234, TAMV-OFFICIAL-SYSTEM-DESCRIPTION.md:32-35

2. Pen2PDF

This module utilizes a CNN+RNN (Convolutional Neural Network + Recurrent Neural Network) architecture to recognize handwriting and convert it into digital text.

• Target Accuracy: 95%+ accuracy.
• Datasets: Validated against ScanNet and SYNTHIA (3,099+ scenes).
• Use Cases: Digitization of university notes and historical document conversion.
  Sources: TAMV-MASTER-DOCUMENTATION.md:243-255, TAMV-OFFICIAL-SYSTEM-DESCRIPTION.md:37-40

3. Isabella Chat

The conversational interface of the framework. It operates under a strict ethical supervision pipeline where messages are validated for compliance before a response is generated.

• Architecture: Explainable Language Models (LLMs) with an added "Explainability Layer" to provide context on why a specific response was given.
• Use Cases: Technical support, personalized tutoring, and content moderation.
  Sources: TAMV-MASTER-DOCUMENTATION.md:264-279

4. Spatial AI

Designed to understand 3D environments for the DreamSpaces engine. It employs ResUNet14, MinkowskiEngine, and trilateral CRF for semantic segmentation of 3D spaces.

• Functionality: Processes photorealistic scenes for intelligent navigation and object recognition.
• Status: Functional with documented limitations.
  Sources: TAMV-MASTER-DOCUMENTATION.md:290-298, TAMV-OFFICIAL-SYSTEM-DESCRIPTION.md:42-45

---

Technical Architecture

Isabella's architecture is federated and designed for high-risk AI compliance, following the requirements of the EU AI Act.

System Data Flow

The following diagram illustrates how Isabella processes a user request, emphasizing the mandatory ethical validation phase.

The data flow ensures that every action is audited for ethical compliance before execution and recorded in the immutable BookPI registry.
Sources: REPORTE-AVANCE-TECNICO-COMPLETO.md:144-165, TAMV-OFFICIAL-SYSTEM-DESCRIPTION.md:27-30

Key Components Table

Component	Technology	Role
Isabella Core	Python 3.9+, FastAPI	Main orchestration and API server
Inference Engine	ONNX Runtime	High-performance model execution
Vector Memory	pgvector (PostgreSQL)	Stores embeddings for RAG and context
3D Processing	MinkowskiEngine	Sparse 3D semantic segmentation
Audit Layer	TGN Governance	Human-in-the-loop validation
Sources: TAMV-MASTER-DOCUMENTATION.md:95-103, REPORTE-AVANCE-TECNICO-COMPLETO.md:21-35

---

Ethical Governance and Supervision

Isabella is strictly governed by the DINN (Dignidad Digital) principle. No AI system within TAMV executes final, binding actions on critical rights or assets without human oversight.

Supervision Layers

Isabella's decisions are overseen by the TENOCHTITLAN security framework:

1. Twin MOS: A cross-validation system where "Agent A" validates an action while "Agent B" questions it.
2. Human Guardianship: Four types of human guardians (Technical, Ethical, Legal, Economic) hold final authority.
3. Explainability (XAI): Isabella must provide multiple versions of "why" a decision was made, tailored to different audiences.
   Sources: TAMV-MASTER-DOCUMENTATION.md:308-316, TAMV-FINAL-PRODUCTION-READY-README.md:91-96

API Implementation (Sovereign Interaction)

Isabella provides a standard API for interaction, often integrated through the TAMVAI server which provides OpenAI-compatible endpoints.

class IsabellaChat {
    async processMessage(message, context, userProfile) {
        // 1. Validate ethical compliance
        const ethicalCheck = await this.validateEthicalCompliance(message);
        if (!ethicalCheck.approved) {
            return this.generateEthicalResponse(ethicalCheck.reason);
        }
        
        // 2. Generate response and add XAI layer
        const response = await this.generateContextualResponse(message, context, userProfile);
        return this.addExplanabilityLayer(response);
    }
}

Sources: TAMV-MASTER-DOCUMENTATION.md:267-278, REPORTE-AVANCE-TECNICO-COMPLETO.md:35-50

---

Implementation Status and Metrics

Isabella is currently integrated into the tamvai-api server, supporting text completions and embeddings.

Metric	Target	Status
API Availability	99.99%	Verified
Response Latency	< 100ms (p95)	Verified
Ethics Score	> 90%	Verified
Detection Time	< 2 seconds	Measured in tests
Sources: TAMV-OFFICIAL-SYSTEM-DESCRIPTION.md:104-115, TAMV-FINAL-PRODUCTION-READY-README.md:102-108

---

Conclusion

Isabella AI Capabilities represent a shift from traditional executive AI to an advisory, ethical framework. By combining high-performance technical components like MinkowskiEngine and FastAPI with a rigorous human-supervised audit layer, Isabella ensures that the TAMV ecosystem remains a sovereign and dignified space for its citizens. Its primary significance lies in its compliance with the EU AI Act and its role as the cognitive orchestrator of the MSR blockchain and XR DreamSpaces.
Sources: TAMV-OFFICIAL-SYSTEM-DESCRIPTION.md:188-192, REPORTE-AVANCE-TECNICO-COMPLETO.md:120-130