Every client is deployed within a fully dedicated cloud account. This means no compute, storage, networking, or database resources are shared between clients - ever. Your data exists exclusively within your own infrastructure boundary, completely separate from any other organization using the platform.This model provides the highest level of infrastructure isolation available in cloud deployments stronger than the access-control-based separation used by most platforms.
When documents are ingested - whether through integrations or direct uploads they are processed entirely within your dedicated cloud environment. Documents are parsed into text, broken into chunks, and converted into vector representations using models running within your environment. These representations are stored securely in your cloud-hosted vector database.By default, no document content, metadata, or access permission data ever leaves your environment unless you explicitly configure it to do so.
When a user submits a query, the system determines whether additional context is needed before generating a response. If context is required, it is retrieved exclusively from your ingested knowledge base within your cloud environment.By default, the only external system that query data is sent to is the LLM configured for your deployment. No other external communication occurs unless explicitly enabled by your administrator.
Administrators may configure third-party integrations to extend the platform’s capabilities. All integrations are disabled by default and must be explicitly enabled. The data implications of each are outlined below:Embedding modelDocument content may be sent to an external API for vector embedding. By default, embedding runs entirely within your dedicated cloud environment using an on-premise model.Document processing serviceDocument content may be sent to a third-party service for enhanced text extraction.Web searchSearch queries may be sent to a configured external search provider when enabled. Only the query string is transmitted - no document content or user data is shared.Image generationWhen enabled, prompts may be sent to a third-party image generation service. This feature is off by default and activated only upon administrator request.Custom actionsAdministrators may configure API-based actions available to the assistant. The scope of any data shared externally depends on the specific actions configured for your deployment.All external integrations are disabled by default. Any configuration that results in data leaving your dedicated environment is controlled exclusively by your organization’s designated administrators.
