api.qtj.ai · RAG-serve
A brand of BREPO PL EN
QTJ.ai — Query Trace Judge
Book a pilot
Private enterprise AI: RAG, quality control and a complete audit trail

Private enterprise RAG. Every answer has a source, a trace and a quality assessment.

QTJ.ai is a private layer between the user, the knowledge layer (RAG) and the selected model — local, Claude or ChatGPT. You can change the model. Knowledge, control and audit stay in QTJ.

Book a pilot in Cloud EU See how it works
Cloud EU Private Cloud On-prem (planned) Built with the AI Act in mind

Don't deploy AI on trust. Deploy AI with evidence.

Meet QTJ
What QTJ.ai is

A controlled layer between a question and the AI answer.

QTJ is not a regular chatbot. The client may use a local model, Claude or ChatGPT — but the knowledge layer, quality control, logging, validation and audit report stay on the QTJ side. Three pillars run on every answer:

Query

A secure AI gateway. It takes the question and retrieves the right context from the selected knowledge layer before anything reaches the model.

Trace

A full trace of every answer: answer_id, sources used, document versions, model, cost and time. It feeds the Evidence Pack — an audit-ready report.

Judge

Quality control before a human sees the answer: sources, hallucinations, sensitive data, procedure risk. Result: PASS, WARNING, REFUSE or ESCALATE.

What QTJ means

One acronym, six layers of meaning.

Q·T·J

Quantum Tunnel Junction

Brand meaning — a controlled passage through a barrier. That barrier is the risk of AI in a company: no sources, no trace, no quality control and no audit evidence.

Q·T·J

Query · Trace · Judge

Functional meaning — a question with the right context, a full trace of AI activity and an evaluation of the answer before it is shown to the user.

Q·T·J

Quorum of Tiny Judges

Technical meaning — a set of small validators: sources, hallucinations, personal data, company policies and procedure risk.

Q·T·J

Qualified Technical Judgment

Evidence meaning — not a raw model output, but a qualified technical judgment: with a source, a document version and a validation result.

Q·T·J

Quality-Traced Justification

Audit meaning — every answer justified and with traceable quality. Trace and the Evidence Pack show the basis on which AI answered.

Q·T·J

Quantifiable Trust, Justified

Governance meaning — trust in AI stops being a declaration. Judge gives a measurable assessment, Trace gives a trail and source.

How QTJ works

From query to evidence — one controlled path.

The generative model is swappable. The knowledge layer, Judge and Trace stay in QTJ, so control and audit do not depend on which model was used.

Query api.qtj.ai RAG-serve Industry / Client sources · versions Engine LOCAL · Claude · GPT Judge PASS / REFUSE local validators Trace answer_id · sources · cost Evidence Pack audit-ready report
  1. A query reaches api.qtj.ai and the RAG-serve layer.
  2. QTJ retrieves context from the Industry, Client layer or both.
  3. The client selects the engine: LOCAL, Claude or ChatGPT.
  4. The model produces an answer on the provided context.
  5. Judge checks the answer before release.
  6. Trace records the full log and feeds the Evidence Pack.
  7. The user gets an answer with sources — or a justified refusal.
Two knowledge layers

A ready-made Industry layer or a private RAG on your documents.

QTJ_INDUSTRY_V1

Ready-made Industry layer

Developed by QTJ for manufacturing, automation, maintenance, energy, quality and technical documentation. It understands the language of PLC, HMI, SCADA, drives, protection devices and switchgear, and ecosystems from brands such as Siemens, Eaton and Schneider Electric.

PLCHMISCADADrivesProtectionSwitchgearFault codes

It is not a trained model. It is an industrial RAG profile: curated indexes, terminology, rules, prompts and validations.

QTJ_CLIENT_V1

Client's private RAG

Built by QTJ on the documents of a specific organization: instructions, procedures, catalogs, machine documentation, service history and quality standards. The client does not build RAG on its own.

InstructionsProceduresCatalogsMachine documentationService history

The client provides documentation and feedback. The QTJ team indexes, selects metadata, manages versions and applies corrections.

Deployment modes and IP protection

Choose your level of control — from EU cloud to air-gapped.

QTJ Cloud EU

Pilot

An instance in the EU cloud. The fastest start.

  • KnowledgeIndustry and/or Client
  • EnginesLOCAL · Claude · ChatGPT
  • KeysQTJ or client BYO
  • ProtectionFull control on the QTJ side

QTJ Private Cloud

A dedicated instance in the client's infrastructure.

  • KnowledgeClient only (no Industry)
  • EnginesLOCAL · Claude · ChatGPT
  • KeysClient keys only
  • ProtectionData stays local; Industry IP not installed

QTJ On-prem

Planned

Local / air-gapped variant, no cloud.

  • KnowledgeClient only
  • EnginesLOCAL only
  • KeysNo cloud keys
  • ProtectionFull locality, no cloud dependency
QTJ.ai

QTJ.ai solutions

Choose the area that matches your organisation's needs — from private RAG to control, manufacturing and deployment requirements.

Private enterprise RAG

Private enterprise RAG on company documents with cited sources, quality control and a complete audit trail. EU Cloud or Private Cloud.

RAG for manufacturing

AI for maintenance and technical documentation. Manufacturing RAG with sources, document versions and answer quality control.

Auditable AI

Auditable AI answers with answer ID, sources, document versions, model, cost, time and quality result in an Evidence Pack.

QTJ and the EU AI Act

Technical mechanisms supporting EU AI Act preparation: logging, transparency, human oversight, monitoring and AI documentation.

Private AI deployment

Compare private AI deployment modes: EU Cloud, dedicated Private Cloud and the planned air-gapped on-prem variant.

QTJ and the AI Act

Technical mechanisms and evidence — not automatic legal compliance.

QTJ supports an organization's preparation for Regulation (EU) 2024/1689: answer transparency, event logging, human oversight, monitoring and documentation of AI use.

EU AI Act · Regulation (EU) 2024/1689 — mechanisms and evidence, data in the EU region (Cloud EU)

The scope of obligations depends on the client's role, the system's purpose and whether a given use is a high-risk AI system. A legal review is recommended before any sales material referring to the AI Act. Source: Regulation (EU) 2024/1689 — EUR-Lex.

Related articles

Art. 4 · AI literacy Art. 12 · event logging Art. 13 · transparency Art. 14 · human oversight Art. 15 · accuracy & cybersec. Art. 26 · deployer duties Art. 50 · transparency duties Art. 113 · timetable
QTJ.ai × BREPO
From BREPO engineering DNA

QTJ.ai is an initiative of the BREPO brand.

Behind QTJ stands BREPO — a Polish industrial automation engineering company. The same core: real technical documentation, engineering discipline and accountability for the result. That is why the Industry layer speaks the language of industry, not generalities.

BREPO Sp. z o.o. · industrial automation · Mikołów, PL

Platform access

QTJ application and API.

Two ways to access QTJ: a web interface for users and an API for integration with systems and applications.

APP

app.qtj.ai

The QTJ web application for working with queries, sources, audit trails and answer quality assessments.

API

https://api.qtj.ai/

A programming interface for connecting systems and applications to QTJ services. Access to API resources requires the correct endpoint and authorisation.

Contact

Let's talk about a QTJ pilot.

The fastest path is a pilot in Cloud EU: the Industry layer, a private RAG on your documents or both — with a full trace and quality assessment from the very first answer.

Ask about a pilot