Understand the Private Cloud capability boundary¶
Private Cloud is the implemented but disabled HiveAI deployment path for an approved private chat service, private embedding service and private search database. An embedding is a numerical representation used to find relevant information; it is derived from source content and needs the same careful governance as the content it represents.
This path is intended to keep those AI processing components within a specifically reviewed deployment boundary. The boundary identifies the approved services, information handling, region, suppliers and accountable organizations. It does not remove Zerohive's permission, classification, source-access, oversight or spending requirements.
Current availability: Private Cloud, private providers, private embeddings and private search storage remain disabled in the inspected configuration. This guide is preparation and boundary documentation. It does not announce availability, promise a delivery date or provide a self-service activation procedure. Non-private HiveAI retains its separate default availability subject to current controls.
Three meanings that must stay distinct¶
| Term or screen | What it concerns | What it does not establish |
|---|---|---|
| HiveAI Private Cloud capability | A governed private AI processing and retrieval boundary | An enabled or qualified customer deployment |
| A commercial Private Cloud plan or contract | Commercial terms and hosting context | AI model availability, source access or permission to change infrastructure |
| Legacy Private Cloud Review | Staff review of existing commercial arrangements after the plan split | An AI activation switch or automatic infrastructure migration |
The latest inspected commercial designs show a review queue and an empty state. Their presence does not change HiveAI's disabled status. Use the commercial catalog and contracts guide for that separate responsibility.
What the private path contains¶
The source implementation includes an approved deployment profile, a versioned policy projection for each workspace, private model and embedding adapters, and a PostgreSQL-based search provider. Search information is separated into ordinary workspace content, Secret workspace content and platform support content.
A model name or connection setting is only one part of this arrangement. The complete profile also needs compatible processing and search dimensions, current supplier evidence, a defined information ceiling, approved regions and jurisdictions, clear responsibilities and an unsuspended policy. A valid profile still does not prove that actual services are reachable or that recovery has been rehearsed.
Private service failure does not automatically move content to another provider, public service or region. Changing the processing boundary is a reviewed deployment decision.
Business responsibilities¶
The organization considering this path needs named owners for deployment, model and embedding supply, inference operations, database operations, evidence, security, data protection, legal review, incident reporting, monitoring, corrective action, backup/recovery and retention/deletion. Conditional supply-chain roles also require an explicit decision.
These are accountability requirements in the implementation and operating process, not a claim of legal certification. Decision-makers need the actual deployment assessment and supporting evidence. Current approval can expire or be suspended; ownership must continue after initial preparation.
Infrastructure, supplier operation, support and recovery create ongoing responsibilities. No universal price, savings or service commitment is established by this documentation.
Data and recovery¶
Private search data is derived from authoritative records and documents. Preparing it, verifying it and making it available are separate steps. Replacing an active search generation retires the earlier generation; it is not a way to restore deleted information or bypass withdrawn access.
Removing information from search is also different from satisfying a complete deletion request across source systems, retained evidence, exports and backups. Retention and legal-hold decisions belong to the responsible business and operational processes. A database method alone cannot certify that those obligations were met.
Where implementation still needs qualification¶
The database checks inspect selected schema, privileges, tenant isolation and active support-search information. They do not prove that every workspace's data is complete. The alternative accelerated search mode has an evidence contract, but its live readiness integration is incomplete. A healthy result for a disabled component means only that it is disabled.
The private worker also has gaps in the current support-publication and workspace-manual publication composition. The presence of general search-storage methods does not establish a complete user workflow. These limitations remain open before any future activation.
Use the Private Cloud preparation and support guide for responsibilities and safe next steps. Provider governance, knowledge lifecycle, governance and incidents and credits explain the shared controls.