Technical stack
The technologies behind OZ-Technology's solutions
Artificial intelligence built into business tools, PHP backend, Linux servers, relational databases, connected energy (HEMS/IoT), open APIs and VoIP telecom : here, in detail, are the technology building blocks used to design solutions that are reliable, secure and built to last.
- AI and RAG connected to the company's real data
- A performant, secure and maintainable PHP backend
- Linux infrastructure, Docker and cloud best practices
- Open APIs, the MCP protocol and third-party integrations
Artificial intelligence built into the CRM and business tools
Rather than a generic chatbot bolted onto the front end, the AI developed by OZ-Technology is plugged directly into the company's real data : product catalogue, technical documentation, customer base, invoices. The goal is to save real time on concrete tasks : quoting, document search, drafting, and querying data.
In practice, this relies on vector search (RAG – Retrieval-Augmented Generation), which lets a language model answer based on the company's own documents and data rather than on its general knowledge alone, with strict control over what the assistant can read or write.
Concrete use cases
- A pre-sales assistant that queries the product catalogue to speed up a quote
- Querying the company's database in natural language, no Excel export needed
- Augmented search over large legal or technical documentation
- Secure connection of external AI tools (such as Claude) to the CRM via API/MCP, with controlled read access
A concrete example : CRM Lumeco comes with two native AI assistants for sales teams and management, and the OZI project illustrates this same approach applied to augmented document search.
PHP backend: robust and quick to implement
PHP remains at the core of the applications developed by OZ-Technology, in particular for business logic, internal APIs and data processing. A pragmatic choice : a mature ecosystem, very broad hosting compatibility, excellent performance for database-driven applications, and controlled operating costs.
Every application is built with constant attention to security (prepared statements, strict input validation, session management) and maintainability : code organised by business domain rather than stacked scripts, with a clear separation between logic, data access and presentation.
What it enables
- REST/JSON APIs to power dashboards, mobile applications and third-party integrations
- Scheduled data ingestion and processing scripts (cron jobs, queues)
- Document generation (quotes, invoices, reports) and business automation
- Integration with modern frameworks (Laravel) depending on project size
Linux servers, Docker and cloud infrastructure
All applications are hosted and run on Linux servers (Debian/Ubuntu), with Nginx or Apache as the front end and PHP-FPM for execution. This open-source foundation guarantees transparency, stability and no dependency on a proprietary vendor.
Services are containerised with Docker where relevant : environment isolation, reproducible deployments between development and production, and easier scaling to orchestrate. Monitoring, automated backups and system hardening (firewall, restricted SSH access, regular updates) are built into every production release.
Best practices applied
- Hardened Linux servers (firewall, fail2ban, key-based SSH access)
- Docker containerisation for isolation and portability
- Automated backups and a disaster recovery plan
- Application and system monitoring to anticipate incidents
Databases: MySQL, MariaDB and time series
MySQL and MariaDB form the main storage foundation : reliable, widely documented, and performant for the relational workloads typical of a CRM or a business application (customers, quotes, invoices, schedules).
For high-frequency measurement data — solar production, electricity consumption, energy prices by quarter-hour — the schema is designed from the outset for time series : tailored indexing, pre-computed aggregations and archiving strategies, to keep performance stable even with several million measurement points accumulated over time.
In the field
- Data modelling tailored to each business (billing, energy, HR, sites)
- Query and index optimisation for responsive dashboards
- Regular backups and tested restore procedures
- Handling very large volumes of measurements for HEMS platforms
HEMS and IoT: connected energy
A HEMS (Home Energy Management System) links together the equipment of an energy installation — solar inverters, home batteries, powermeters, digital meters (P1/DSMR) — to provide a unified view and automated control, where each manufacturer usually offers its own isolated portal.
Field data is collected via IoT protocols such as MQTT and manufacturer APIs, cross-referenced with external sources (solar forecasts, Elia data, dynamic electricity tariffs), to generate battery charge and discharge strategies that genuinely reduce the bill.
Typical components of a HEMS architecture
- Real-time ingestion via MQTT and manufacturer APIs (e.g. Huawei inverters)
- Digital P1/DSMR meters and powermeters for fine-grained measurement
- Solar forecasting and weather weighting to anticipate production
- Battery optimisation engine based on dynamic and capacity tariffs
See the concrete implementation on the Fluxi HEMS project.
API, integrations and the MCP protocol
No solution lives in isolation : applications expose REST/JSON APIs to connect to third-party tools — accounting, banking (CODA import), energy suppliers, telecom operators — and to power dashboards or external applications.
For AI, the open MCP (Model Context Protocol) protocol makes it possible to connect an assistant like Claude to a CRM or a business database in a standardised, secure way, with precisely defined access rights (read only, limited scope), rather than building an ad hoc integration for every AI tool.
Types of integrations delivered
- REST/JSON APIs for web applications, mobile apps and dashboards
- Webhooks and synchronisation with third-party business tools
- Automatic import and reconciliation of bank statements (CODA)
- Secure connection of external AI assistants via API/MCP
Telecom and VoIP: more than 25 years of field expertise
Before AI and the web, there's the network. Design and operation of reliable VoIP and SIP infrastructure : operator interconnections, SBC, routing, real-time monitoring and voice traffic security.
This expertise now translates into IP telephony platforms monitored continuously, with automatic generation of service-quality statistics and indicators (SLA, KPI), and seamless integration with CRMs and business tools to link every call to a customer record.
Areas covered
- SIP/TDM interconnections with Belgian and international operators
- IP telephony servers, SBC and routing management
- Real-time monitoring and SLA/KPI reporting
- VoIP ↔ CRM integration for unified customer tracking
Details and concrete cases on the Telecom & VoIP page.
Security: a cross-cutting requirement, not an option
On every project, security is treated as a design constraint, not as a step added at the end of development.
- HTTPS/TLS encryption across all exposed services
- Prepared statements and strict input validation to prevent injection attacks
- Precisely scoped AI data access (read only, defined perimeter)
- Hardening of Linux servers (firewall, restricted access, tracked updates)
- Attention to GDPR from the design stage (personal data, logging, retention)
Technical stack at a glance
Frequently asked questions about the technologies used
How is AI integrated into a CRM ?
By connecting a language model to the company's real data through vector search (RAG) : catalogue, documents, customer base. The assistant then answers from data that is up to date and specific to the company, with controlled access — including via the MCP protocol to connect tools such as Claude.
Why choose PHP over another language ?
Modern PHP offers an excellent balance between development speed, performance and hosting cost for data-driven applications. With good practices (prepared statements, clear architecture, testing), it makes it possible to deliver robust business tools that are easy to maintain over time.
What is a HEMS and what is it for ?
A HEMS (Home Energy Management System) centralises solar production, consumption and battery data to automatically control energy flows, taking weather forecasts and dynamic tariffs into account, in order to reduce the electricity bill.
What is the MCP (Model Context Protocol) protocol ?
An open protocol that allows an AI assistant to connect in a standardised, secure way to external tools and databases, with precisely defined access rights, without building a custom integration for every assistant.
What infrastructure do the applications run on ?
Linux servers (Debian/Ubuntu), with Nginx or Apache as the front end, PHP-FPM for execution and MySQL/MariaDB for storage. Services are containerised with Docker where relevant, for reproducible, isolated deployments.
Which databases are used for energy and business data ?
MySQL and MariaDB, with a schema designed for time series when it comes to high-frequency measurements (production, consumption, prices), to keep good performance even with millions of data points.
Have a technology in mind, or a specific need ?
AI connected to your data, a business CRM, HEMS, API or VoIP infrastructure : let's talk about your project and the most pragmatic way to bring it to life.