DataFab Utilities

Version: 1.0 Last Updated: May 2026


What Is a Utility

A Utility in DataFab is a packaged, domain-specific application built on top of the DataFab core framework. Where the core framework provides general-purpose primitives — Knowledge Fabric, Studio, Graph Operations, Graph RAG, AI/LLM Layer, Exchange — a Utility composes those primitives into an end-to-end solution for a particular operational use case.

Each Utility is a first-class platform tenant: it has its own domain, schemas, data sources, DDAs, dialog playbooks, DAG rule chains, and analyst surfaces. Utilities reuse the same security, audit, and operational-mode controls as the rest of the platform; they do not introduce parallel infrastructure.


Utility Anatomy

Every DataFab Utility is structured around the same set of building blocks. The presence and depth of each block vary by utility, but the surface is consistent.

Block Role in a Utility
Domain & Schemas Defines the entities and relationships under analysis (see Schema Management)
Data Sources MCP connectors, customer system connections, OSINT sources (see Knowledge Fabric)
Knowledge Graph Tenant-scoped graph holding subject entities, counterparties, related parties (see Knowledge Fabric)
DAG Rule Chain Self-organising rule graph that evaluates evidence and produces a decision (see Graph Operations › DAG Rules)
Text-to-Cypher Rules Graph-querying rules used by the chain to gather entity-level evidence (see Graph RAG › Text-to-Cypher Rules)
DDAs Schema-bound agents that surface, summarise, or act on chain outputs (see Studio)
Dialog Playbook The analyst-facing conversational layer with intents, tools, guardrails (see Dialog › Playbook-Based Communication)
System Self-Awareness Introspection tools so analysts can interrogate rules, traces, and configuration (see Dialog › System Self-Awareness)
Decision Outputs Structured packs (e.g., escalation, request-for-information, decline, refer) consumed by downstream operational systems
Audit & Provenance Per-rule, per-turn, per-decision audit trails persisted alongside the analysis record

Core Framework Primitives Reused by Utilities

Utilities are deliberately thin: they configure the platform, they do not extend it. The primitives below are owned by the core framework and consumed by every utility.

Primitive Owned By Consumed As
Knowledge Graph + Entity Resolution 03-Knowledge-Fabric Subject and counterparty entity store
Schema Registry 09-Schema-Management Domain schema for the utility
MCP Connectors 03-Knowledge-Fabric Source data ingestion
DAG Rule Engine 08-Graph-Operations Rule chain execution and trace
Text-to-Cypher 13-Graph-RAG Graph-querying inside rules and DDAs
Graph RAG 13-Graph-RAG Retrieval for analyst-facing summarisation
LLM Gateway 05-AI-LLM All LLM calls with provider isolation and provenance
Studio (DDAs / Widgets / Tools) 04-Studio Analyst-facing assets and tool catalog
Dialog Playbooks 15-Dialog Conversational layer
System Self-Awareness 15-Dialog Run, rule, and configuration introspection
Operational Modes (0–4) 04-Studio Automation level per utility
Human-in-the-Loop Controls 04-Studio Review gates on high-risk decisions
Tool Authorization 04-Studio Per-user, per-tenant tool access
Audit Logging 03 / 04 / 05 / 08 / 13 Per-component audit trails

Utility Catalog

Utility Document Status Summary
Transaction Monitoring Transaction Monitoring Production DAG-driven financial-crime alert triage (50+ rules, scoring, decision routing) with tenant BPMN workflow, multi-source OSINT and watchlist screening, explainability + HITL, SAR filing pipeline, outreach gap detection, straight-through processing, dashboards, and policy versioning with ground-truth testing
Car Finance (AutoFab) Car Finance Production Motor finance remediation under the FCA Motor Finance Scheme (population, cohorting, scheme testing, redress, outreach, payment, closure)
Compliance Compliance Production Compliance and conflict-of-interest utility — graph-based conflict detection, OSINT schema management, BPM-driven processes, risk rules + thresholds, adverse-media + watchlist screening (built-in and tenant-specific)

Utility Lifecycle

Stage Description Approvals
Definition Domain, schemas, data sources, rule taxonomy, decision contract are agreed Domain owner
Configuration DAG chain, playbook, DDAs, thresholds, scoring weights are configured per tenant Compliance / risk owner
Validation Test suites run against historical data; thresholds calibrated; analyst playbook tested Independent review
Pilot Limited rollout under Mode 1 (AI-Assisted Manual) with Human-in-the-Loop on every decision Operations owner
Production Full rollout under the chosen operational mode (1–4) with audit and escalation in place Change control
Continuous Tuning Rule changes, threshold updates, schema extensions go through versioning and approval Rule Engine governance

Utility Security and Compliance Posture

Control Implementation
Tenant Isolation Every utility runs in its tenant scope; cross-tenant access is impossible by construction
Operational Mode Binding Each utility declares the mode(s) it supports; out-of-mode invocations are rejected
Decision Auditability Every utility decision is backed by a chain trace and a tool-call audit (08, 04)
Regulatory Mapping Utilities map their decisions to the regulations they support (see Compliance Capabilities)
Model Provenance LLM-driven steps log model ID, prompt hash, and version (see AI & LLM)
Data Residency Utilities respect the tenant’s data residency configuration (see Architecture)
Human-in-the-Loop High-risk decisions are gated by HITL until at least Mode 3

How to Read the Utility Documents

Each utility document follows a consistent structure:

Section Purpose
Component Overview What the utility does, in one paragraph and one capability table
Architecture Overview Block diagram showing reused core primitives and utility-specific configuration
Domain & Schemas The schemas, entities, and relationships the utility analyses
Data Sources MCP connectors, customer systems, OSINT sources used
DAG Rule Chain Tier structure, rule catalog, scoring config, decision config, chain trace
Decision Outputs Structured outputs and how they integrate with downstream systems
Dialog Playbook Intents, tools, guardrails surfaced to analysts
Self-Awareness What analysts can introspect within this utility
Operational Modes Supported modes and HITL gates
Security Controls Utility-specific controls layered on the core posture
Audit Logging Utility-specific audit events and retention
Regulatory Mapping Regulations the utility supports

Cross-References