Security, made visible
HTTP and TLS configuration checks, explained through the CipherBeam prototype.
Explore CipherBeam ↗BUILD SYSTEMS.MAKE THEM WORK.
AI systems, security engineering and automation.
Built for the real world. Tested before launch.
Connect the moving parts
Intake, routing and orchestration, connected to the tools a team already uses.
Explore automation ↗
Knowledge into answers
Earlier RAG work, with documents, retrieval and visible sources. The current Studio evolution is in progress.
Explore Ragnar Studio ↗1 of 3
Three disciplines.
One accountable person.
AI systems first, security always, automation where it connects the two.
03 / SELECTED WORK
Work in progress.
Engineering in public.
Three current product lines. Existing implementations and the next engineering step, clearly separated.
01 / AI SYSTEMS
KNOWLEDGE → RETRIEVAL → ANSWER
CONCEPTUAL ARCHITECTURERAG application · Studio evolution planned
Ragnar Studio
A self-hosted assistant application with document ingestion, retrieval and source-aware answers.
FastAPI / SQLite / Ollama / Qdrant
Current: static console and ingestion pipeline. Next: canonical knowledge ownership and governed providers. Production readiness is not established.
Project overview →02 / SECURITY ENGINEERING
SCOPE → EVIDENCE → FINDING
CONCEPTUAL ARCHITECTUREWeb-check prototype · unification planned
CipherBeam
A web security prototype that inspects HTTP and TLS configuration and explains its findings.
Python / Flask / Requests
Current: web configuration checks. Next: authorized repository assessment and captured evidence. The unified platform is not implemented.
Project repository ↗03 / AUTOMATION & INTEGRATIONS
CONNECT → PLAN → REVIEW
CONCEPTUAL ARCHITECTURERead-and-planning app · consolidation planned
Social Connector
A social account application for OAuth connections, partial metrics, drafts and calendar preparation.
Laravel / React / PostgreSQL
Current: read and planning workflows. Next: authentication and workspace isolation. Automated publishing and real AI generation are not available.
Project overview →Built with a method.
See an attack get blocked ↗- 01
Understand.
Map the process, the data and access.
- 02
Build.
Create the smallest useful system.
- 03
Attack.
Test hostile inputs and misuse.
- 04
Harden.
Fix, retest and document the boundaries.
FIELD NOTES / FROM THE STUDIO
The thinking
behind the build.
Why a RAG project starts with preparing knowledge, not choosing a model
RAGnar and the RAGnar Knowledge Processor show why a good AI system starts with tidy preparation of documents, chunks and sources.
07 / ABOUT
A solo studio, on purpose
I'm Dario Mijić. TOKMATIK is a one-person studio, so there is no hand-off between a salesperson, a developer and a tester. I build the system, attack it, document it and stay accountable for it. Work is done in English or Croatian.
- LOCATION
- Based in Croatia
- LANGUAGES
- English, Croatian
- CONTACT
- dario@tokmatik.eu
08 / FAQ
Questions before starting
The first question is rarely "which tool?". It is "which problem first?"
Do we need AI for this?
Not always. Often the biggest gain comes from cleaner intake and clearer routing. AI is added only where it earns its place.
Can you test an agent or RAG system we already built?
Yes. Testing is done only on systems you own or have given written permission to test, with an agreed scope and rules of engagement.
Where does our data go?
Private and self-hosted setups are possible, including local models. Data handling is agreed in writing before any work starts.
What do you need from us?
A short description of the process, access to the relevant tools or documents, and one person who can make decisions.
How long does a first project take?
It depends on scope. The aim is a first useful system quickly, not an endless transformation.
What happens after delivery?
You can stop at the delivered system, or continue with monitoring, improvements and retesting.