Operating models
Alcilla can start small and be expanded later. Personal working data, shared knowledge, and compute-intensive AI services can be operated separately in a deliberate way.
The models are not mutually exclusive. Local or centralized operation can additionally use Alcilla.Station or Nexus as an intelligent KI loadbalancer for the company’s own infrastructure.
Local single-user workstation
In local operation, Alcilla runs on a designated computer. Personal content such as emails, notes, chat histories, recordings, and local Kanban boards remains there.
This model is suitable for:
- individuals and small pilot groups,
- particularly confidential personal work areas,
- a quick start without a central platform,
- companies that first want to gain experience with local AI models.
If the AI models and additional services are also operated locally, the workstation can function largely independently of central AI infrastructure.
Alcilla.Hub
In the centralized model, each user continues to have a local Alcilla instance. A central service additionally provides approved knowledge and shared content.
A search can therefore consider two areas:
- the user’s personal local knowledge,
- the centrally approved company knowledge.
The results are processed together without automatically transferring personal content to the central area.
Which content remains local?
In the intended central operating model, the following personal content in particular is not stored centrally:
- emails,
- chat histories,
- personal notes,
- Recorder content,
- local working data.
Shared Kanban boards are a deliberate exception. They can be stored centrally and selectively shared with people or groups.
Sign-in and permissions
Centralized operation is secured through an enterprise sign-in based on OIDC. This allows Alcilla to connect to an existing identity provider, such as Keycloak or another compatible enterprise solution.
Permissions can be based on groups and approvals. Areas that have not been approved are not displayed in search results.
We can assist with connecting existing identity providers when required. Additional interfaces, such as connections to ADFS or other authentication systems, can also be implemented on request.
Alcilla.Station
Alcilla.Station provides selected compute-intensive functions centrally on infrastructure supplied by Alaska.Labs. The Alcilla application and personal working data can continue to be operated and stored on the workstation.
Alcilla.Station can provide services including:
- AI language models for chat, summarization, and extraction,
- embedding models for semantic search,
- document conversion through Docling,
- transcription of recordings,
- web search and research through SearXNG.
The specific scope of functions provided through Alcilla.Station is defined in the relevant agreement.
Which data remains local?
Alcilla.Station does not replace the local Alcilla environment as a whole. The following data in particular can continue to be stored locally:
- emails and attachments,
- the local database and knowledge index,
- notes and chat histories,
- local Kanban boards,
- configuration and personal working data.
For processing performed centrally, however, the content required for the relevant task is transferred to the configured service. This may include text, documents, images, or audio data.
Who is Alcilla.Station suitable for?
Alcilla.Station is particularly suitable for:
- companies without their own GPU or AI infrastructure,
- workstations that cannot run compute-intensive models locally,
- organizations with limited resources for model operation and maintenance,
- a quick start using centrally provided additional services,
- environments where personal data may be stored locally while selected processing steps are performed centrally.
If content is not permitted to leave the workstation even for processing, all required models and additional services must be operated entirely locally.
Alcilla.Nexus
Unlike Alcilla.Station, Alcilla.Nexus does not involve us operating AI servers for you. Instead, we provide an intelligent KI loadbalancer that standardizes your own AI servers within your infrastructure, distributes requests adaptively, and presents local workstations with a consistent interface.
If your company already operates multiple AI servers—whether at one location or distributed across several locations—Nexus handles the standardization: It offers an intelligent loadbalancer with health monitoring, load distribution, and failover for all consolidated servers.
What does Alcilla.Nexus provide?
With Nexus, you can:
- operate one or more AI servers in your infrastructure,
- standardize and distribute requests across your AI servers through the intelligent Nexus loadbalancer,
- connect local workstations to them,
- use health monitoring and automated failover,
- retain complete control over hardware, models, and locations.
Nexus does not perform substantive processing; it is the intelligent mediation layer with load distribution, health monitoring, and failover. All content remains within your infrastructure.
Who is Alcilla.Nexus suitable for?
Alcilla.Nexus is particularly suitable for:
- companies with their own AI infrastructure that want to consolidate multiple servers in one place,
- organizations with heterogeneous or distributed AI server landscapes,
- companies that require full control over their AI infrastructure for compliance reasons,
- organizations that want to make AI services centrally accessible across several locations.
Which option is right for your business?
| Requirement | Suitable model |
|---|---|
| A single workstation should get started | Local single-user workstation |
| Personal data must be stored and processed exclusively on the workstation | Local single-user workstation |
| Several people should be able to search shared knowledge | Centralized operation |
| Shared Kanban boards should be available centrally | Centralized operation |
| Central sign-in and group permissions are required | Centralized operation |
| AI models should run centrally while personal working data remains stored locally | Local workstation with Alcilla.Station or centralized operation with Nexus |
| Document conversion or transcription should be used without internal GPU infrastructure | Alcilla.Station |
| The company’s own AI servers should be consolidated and made accessible | Alcilla.Nexus |
| Shared knowledge and centrally provided AI services are required | Centralized operation with Alcilla.Station or Nexus |
Differences at a glance
Alcilla.Station is not a replacement for a local or central Alcilla installation. The service supplements these operating models with AI models and additional services provided by Alaska.Labs.
Alcilla.Nexus is likewise not a replacement. It provides intelligent mediation with KI loadbalancing to standardize your own AI servers and make them failover-capable.
This results in five typical operating variants:
| Operating variant | Where does Alcilla run? | Where is personal working data stored? | Shared knowledge and boards | Where do AI and additional services run? | Suitable for |
|---|---|---|---|---|---|
| Fully local single-user workstation | On the workstation | On the workstation | Available locally only | Entirely on the workstation | Maximum local control and operation without an external AI service |
| Local workstation with Alcilla.Station | On the workstation | On the workstation, with transfer and data processing through Alcilla.Station | Available locally only | Selected models, Docling, transcription, or research through Alcilla.Station | Quick introduction without the company’s own GPU and AI infrastructure |
| Centralized operation with the company’s own infrastructure | On the workstations and as a central company service | Personal data on each respective workstation | Approved knowledge and shared boards on the central Alcilla instance | On the workstations or the central Alcilla instance | Multiple users, shared knowledge, and fully self-managed operation |
| Centralized operation with Alcilla.Station | On the workstations and as a central company service | Personal data on each respective workstation, with transfer and data processing through Alcilla.Station | Approved knowledge and shared boards on the central Alcilla instance | Selected models and additional services through Alcilla.Station | Multiple users with shared knowledge but without a complete in-house AI infrastructure |
| Alcilla.Nexus—self-managed operation with KI loadbalancing | On the workstations and as a central company service | Personal data on each respective workstation | Approved knowledge and shared boards on the central Alcilla instance | The company’s own AI servers, served by Nexus loadbalancer | Multiple users with shared knowledge, their own AI infrastructure, and a loadbalancing requirement |
Hybrid configurations are possible
Not every function has to be operated in the same location. For example:
- emails can continue to be processed and stored locally on the workstation,
- shared knowledge can be provided through a central Alcilla instance,
- language models can be used through Alcilla.Station,
- document conversion can be provided centrally,
- individual highly confidential workflows can remain entirely local,
- research can be enabled only for selected users.
No existing server environment? A local pilot project is still possible. For subsequent multi-user operation, we can assist with architecture, sizing, and deployment of the required services.