1. Individual consulting and demo

We present the Znuny AI add-on live and specifically address your requirements. In doing so, we clarify the key questions:

Infrastructure: Which solution suits you best – a highly secure local instance, a flexible private cloud, or integration with public models
Requirements: We assess your existing IT landscape and define the necessary hardware and software resources

2. Full-service implementation

After the decision, you are not left on your own. We take over or support the complete installation and setup:

-Integration of Znuny AI into your ticket system
-Setup of the middleware (AnythingLLM or n8n)
-Configuration of your dedicated AI server

3. Content optimization and coaching

An AI is only as good as the knowledge it is trained on. We provide valuable guidance on how to best prepare your data (tickets, PDFs, manuals) to achieve the most accurate results in the RAG backend.

Configurable AI agents can perform, among others, the following tasks:

• Routing tickets (assigning queues)
• Determining priority/urgency
• Setting the status
• Categorizing tickets
• Detecting sentiment
• Generating recommended actions
• Generating summaries
• Generating automated responses
• Detecting the language
• Translating content
• Automatically closing tickets upon a thank-you message
• Generating a descriptive ticket title
• Providing recommendations to improve ticket handling
• Extracting information such as invoice number, branch number, etc.
• Setting any dynamic field based on the customer request
  and much more...

With our flexible toolkit, you can create, test, and activate AI agents based on your own ideas.

There are two automation strategies:

1. Adoption of fields from similar tickets based on the vector database
2. Rule-based control through an LLM model based on individual requirements

AnythingLLM as Middleware

Znuny-KI uses AnythingLLM as middleware and thus offers:

• Open-source
• Free to use
• All-in-one RAG (Retrieval Augmented Generation)
• Continuously developed by an active community

Advantages:

• Simple and fast installation and setup – perfect for getting startedAccess to numerous LLM models (local or cloud-based)
• Compatibility with various vector databases
• Easy indexing of your own documents (e.g., tickets, PDFs, websites)
• Ability to use the LLM chat for context verification

Compatible AI providers:

OpenAI
OpenAI (Generic)
Azure OpenAI
Anthropic
Google Gemini Pro
Hugging Face (chat models)
Ollama (chat models)
LM Studio (all models)
LocalAI (all models)
Together AI (chat models)
Perplexity (chat models)
OpenRouter (chat models)
Mistral
Groq
Cohere
KoboldCPP
LiteLLM
Text Generation Web UI

As an alternative to AnythingLLM, n8n can also be used as middleware.

n8n is the “Swiss army knife” of automation. It is the right choice when AI should not only answer questions but also actively interact with processes and make complex decisions.

Advantages:

• Workflow designer: Create visual logic chains
• Agentic workflows: The AI can independently use “tools” (e.g., query an external SQL database or call an API.
• Hundreds of integrations: Connect Znuny not only with AI but also simultaneously with Slack, Teams, Jira, or your ERP system
• Decision trees (if/else logic) that determine whether the AI can respond or a human must take over
• Connection to multiple vector databases (Pinecone, Milvus, Weaviate)
• Use of advanced RAG strategies to improve output

Which technical requirements are necessary for the AI add-on?

• Znuny 7.*
• Docker-Host for AnythingLLM oder n8n
• Access to a powerful language model
  • Local KI-Server, z. B. with 32 GB VRAM
  • or access to a private or public cloud via API

• Open source & Docker-compatible
• Flexible integration of your own knowledge sources
• Support for numerous language models such as GPT-4, LLaMA, Mistral, Mixtral, Teuken and many more.
• Scalable and high-performance - even for larger installations

Compatible large language models:

• Any open-source llama.cpp compatible model
• OpenAI
• OpenAI (Generic)
• Azure OpenAI
• Anthropic
• Google Gemini Pro
• Hugging Face (chat models)
• Ollama (chat models)
• LM Studio (all models)
• LocalAi (all models)
• Together AI (chat models)
• Perplexity (chat models)
• OpenRouter (chat models)
• Mistral
• Groq
• Cohere
• KoboldCPP
• LiteLLM
• Text Generation Web UI

1. Installation of the Znuny AI add-on and the AnythingLLM middleware
2. Setting up the connection between Znuny and AnythingLLM
3. Indexing the relevant data sources (e.g. tickets, PDFs)
4. Configuration and testing of the AI agents

Looking for an AI server? We recommend the HP Z2 Mini!

• Easy and secure integration of SECTOR NORD AI solutions
• Company data remains internal, with no transfer to external clouds – full control over sensitive information
• One-time acquisition cost instead of ongoing cloud fees, making it particularly cost-effective for continuous operation and high query volumes
• The server is compact and fits into any server room
• We use Linux Fedora and stay true to our open-source strategy – full transparency and control with our monitoring solution SNAG-View
• With 128 GB RAM, medium-sized language models can be run locally without relying on external providers
• The integrated NPU is specifically designed for fast and parallel computations, reducing the load on CPU and GPU and improving AI application performance

In short: greater data protection, full control, lower ongoing costs, and flexible use of AI models