Khaos Research · Use CaseFederated Ecosystem · DCAT-AP Metadata Pipeline · Hybrid AI Search Broker

Federated Data Space Ecosystem (EDAAN)

An enterprise-grade data space architecture combining automated DCAT-AP metadata extraction, hybrid AI search with conversational agents, and sovereign data exchange over high-availability Kubernetes infrastructure.

Next.js 16FastAPIPythonQdrant (Vectors)Elasticsearch 8Pydantic AIDCAT-AP v2JSON-LDKubernetesDockerApache TikaMongoDB

GitHub Visit project

Data Space ArchitectureSovereign P2P Mesh

Sovereignty Mesh

01. INGEST

Raw Datasets+14 Familias · Tablas, Ráster, Audio, Código

→

02. EXTRACT

MetaDataXtractExtracción Determinista + LLM & DCAT-AP

↓

03. GOVERN

ÁGORA / EDAANPolíticas ODRL 2.2 · Conector EDC

←

04. DISCOVER

MetadataSearchHybrid RRF & AI Discovery

↑

Zero data leakage · Bilateral contracts enforce sovereign transfer

3 ReposFederated EcosystemIngestion · Sovereignty · AI Discovery

100% SoberanoData SovereigntyData never leaves producer until bilateral contract is signed

+14 FamiliasDeterministic IngestionDeep technical extraction & semantics without exposing raw data

Búsqueda HíbridaSemantic DiscoveryNatural language search overcoming synonyms & jargon with AI

Inter-Repository Architecture

The Data Space Triad: Ingestion, Search & Sovereign Governance

A data space cannot function with missing metadata, nor can published assets be discovered without intelligent search. These 3 independent repositories work in concert to power the regional data ecosystem.

Repo 01

MetaDataXtract

Extraction & Semantics · Paper in Prep

Inspects +14 format families, extracts deterministic structural schemas, and generates strict DCAT-AP JSON-LD.

Exports validated JSON-LD →

Repo 02

ÁGORA Data Sphere

Sovereign Portal & ODRL Contracts

Official gateway for EDAAN: transactional dataset offers, bilateral access contracts, and Kubernetes cluster.

Synchronizes catalog assets →

Repo 03

metadataSearchTool

Hybrid RRF Search & Pydantic AI

Bridges vocabulary barriers uniting Elasticsearch 8 lexical BM25 and Qdrant dense vectors with conversational agents.

Guides consumers to data assets

MetaDataXtractIndependent Research Line · Paper in Prep · v1.5.0

Security-first automated pipeline extracting structural metadata from +14 format families into DCAT-AP v2.0 JSON-LD without disclosing raw files.

GitHub Repo Live Platform

Explore Capability:

Interactive Metadata Knowledge Graph Across +14 Format Families

Select any format family from the official reference. The graph dynamically branches both common base metadata (always present) and format-specific technical properties extracted deterministically:

section: tabular

Common Base Metadata (Always Present)

MetaDataXtract

Specific Properties: Tabular & Spreadsheets

Metadato Base Común

LLM Deep Description & Utility

AI Semantic Narrative

Extracted Value / Schema:Semantic narrative generated when SUGGESTIONS_LLM is configured

Generated when SUGGESTIONS_LLM_BASE_URL is active. Synthesizes purpose and data utility; safely degrades if LLM is disabled.