Does this skill support non-English datasets?

The skill is specifically configured to produce all diagnostic reports, taxonomies, and schemas in Portuguese (PT-BR).

What inputs are required for the audit?

To get an accurate report, you should provide the total document count, the domain/industry, the embedding model used, and a few examples of your document content.

How does this skill help with hallucinations?

By identifying the 'precision precipice' and providing a structured hierarchical and graph-based schema, it ensures the LLM retrieves more relevant, contextually grounded data.

What is semantic collapse in RAG?

Semantic collapse refers to the loss of retrieval precision in vector stores that occurs when document volume or embedding dimensionality causes distinct data clusters to overlap, leading to irrelevant results.

RAG Semantic Auditor

Name: RAG Semantic Auditor
Author: prof-ramos

byprof-ramos

•

Ciencia de Datos y ML

Diagnoses semantic collapse and optimizes RAG architectures using hierarchical taxonomies and Graph-RAG schemas.

Acerca de

RAG Semantic Auditor is a specialized diagnostic tool designed to prevent performance degradation in Retrieval-Augmented Generation systems. It identifies 'semantic collapse'—the point where vector search precision fails due to high dimensionality or volume—and provides a three-step remediation path. The skill generates a brutalist risk assessment report, a structured 3-level hierarchical indexing strategy to reduce search space, and a deterministic Graph-RAG schema to transition from simple vector similarity to sophisticated relationship-based retrieval, specifically optimized for PT-BR deployments.

Características Principales

Automated 3-level hierarchical taxonomy generation (Root-Branch-Leaf)
1 GitHub stars
Semantic collapse risk assessment with precision drop projections
Deterministic Graph-RAG schema design with node and edge definitions
Search space reduction calculations to optimize indexing efficiency
Traversal simulations to validate knowledge graph query paths

Casos de Uso

Designing hierarchical routing logic for multi-domain RAG systems
Transitioning from basic vector stores to advanced Knowledge Graph architectures
Diagnosing hallucinations and accuracy loss in large-scale vector databases

Acerca de

Características Principales

Automated 3-level hierarchical taxonomy generation (Root-Branch-Leaf)
1 GitHub stars
Semantic collapse risk assessment with precision drop projections
Deterministic Graph-RAG schema design with node and edge definitions
Search space reduction calculations to optimize indexing efficiency
Traversal simulations to validate knowledge graph query paths

Casos de Uso

Designing hierarchical routing logic for multi-domain RAG systems
Transitioning from basic vector stores to advanced Knowledge Graph architectures
Diagnosing hallucinations and accuracy loss in large-scale vector databases