Local AI with meloki: principles and practice
Get to know meloki: local AI models, your own company knowledge and automation on a centrally managed appliance.
meloki combines local AI models, company knowledge and automation on your own appliance. Learn how the platform is structured, which tasks it supports and what you need to use it in your business.
Why a local AI platform?
Many companies want to use AI productively. At the same time, they need to decide which information may be processed, who can access it and which external services are involved. This is an important architectural decision, especially for confidential documents and internal knowledge.
meloki brings AI processing to a dedicated appliance in your infrastructure. Choose the models, provide sources and decide which systems to connect. No public AI cloud is required for core local functions. If you configure external APIs or connectors, assess their data flows separately.
How does meloki work?
The architecture is modular. The publicly described services handle different tasks:
- Ollama provides local language models. Which models can run effectively depends on factors such as model size, GPU memory and the task.
- OpenWebUI provides an interface for chats, models and knowledge sources, along with its own user and permission settings.
- Whisper enables local conversion of speech to text.
- n8n connects AI processing to configured workflows, interfaces and follow-up actions.
The AdminPACK adds central appliance administration to these services. This lets you operate AI applications and infrastructure in one matched system.
What can you do with it?
Make company knowledge accessible
Manuals, policies and other approved documents can serve as knowledge sources. A RAG configuration finds relevant passages and provides them to the language model as context. This is knowledge search followed by answer generation, not automatic training of the model.
How useful the answers are depends on source quality and freshness, preparation and settings. For important decisions, check both the answer and its sources.
Work with documents and enquiries
You can summarise texts, structure content or generate draft answers. This can support internal knowledge work, support teams or HR, for example. Expert, legal and organisational decisions remain with the people responsible.
Automate workflows
With n8n, configured systems and APIs can be integrated into a workflow. For example, an enquiry can first be analysed and then turned into a ticket draft. You define within each process whether a follow-up action runs automatically or requires approval.
What does this mean for getting started?
meloki is offered on matched hardware appliances. The right configuration depends on your tasks, desired models and expected concurrent usage. A blanket user count or fixed introduction period would not be an appropriate measure.
Local operation supports control over data processing. Privacy-compliant use also requires appropriate permissions, technical settings, processes and an assessment of your specific applications.
Start with a suitable task
Start with a manageable task and selected, approved data. Review the results and then expand your use with purpose. In a personal demo we'll discuss your requirements, show available features and talk about the right hardware.
