TREN
Request demo
SQL Server · RAG + LLM

Find the cause
behind performance issues.

MiniRAG is a RAG + LLM powered performance analysis solution that brings SQL Server performance signals together with relevant technical knowledge.

Book a demo →How does it work?

Screens and metrics are design samples, not live product output.

Performance WorkbenchDemo view
Query StoreReadySource concept
Plan Analysis3 plansSample data
RAG Sources12 chunksSample context
Query latency · sample chart
MiniRAG Assistant
Query plan + wait statistics
Relevant runbooks and DBA notes
Source-backed findings and draft recommendation
Capabilities

Not just answers.
Evidence and context.

Handles different signals in one review flow instead of reducing performance to a single metric.

Performance context

A shared review space for Query Store, execution plans, wait stats and performance telemetry.

Technical knowledge via RAG

Retrieves context from runbooks, documentation, past incidents and DBA notes.

LLM-assisted analysis

Explains likely causes clearly and separates assumptions from real signals.

Cross-signal correlation

Interprets CPU, I/O, waits, plan changes and trends over time together.

Expert control

Shows recommendations with sources and leaves the final decision to the DBA.

Repeatable analysis

Standardises review steps and shares operational knowledge across the team.

Reference architecture

The path from signal to recommendation.

This flow is a reference architecture for product design; it does not guarantee current integration scope.

01 · INPUT
SQL Server

Query Store, plans, wait stats and performance telemetry.

02 · KNOWLEDGE
Knowledge base

Runbooks, technical docs, case history and DBA know-how.

03 · RETRIEVAL
RAG layer

Retrieve query-relevant sources and keep provenance.

04 · REASONING
LLM analysis

Likely causes, rationale and evidence-linked explanation.

05 · REVIEW
DBA decision

Validation, impact assessment and a controlled action plan.

Trust approach

Not automatic answers.
Controlled decision support.

Designed for enterprise use; security, traceability and change control are core design criteria.

Read-only start

A low-risk, read-only observation model is recommended first.

Source citation

Recommendations should link to their signals and knowledge sources.

Data governance

Identity, RBAC, data minimisation and sensitive-data policies must be designed in.

DBA approval

Human review and change management are kept for production changes.

These are proposed design principles, not declared as implemented product features.

NEREUS MiniRAG

Turn your SQL Server expertise
into knowledge.

Request a product demo and architecture review.