Ingest

Ingest is how documents enter your VaultCrux corpus. Upload files through the UI, point to URLs for automatic fetching, or push documents via the API. Supported formats include Markdown, PDF, HTML, and plain text. Each document is chunked, embedded, and encrypted before storage.

The ingest panel shows processing status in real time: parsing, chunking, embedding, and indexing stages with progress indicators. Failed documents surface clear error messages explaining what went wrong, whether it is an unsupported format, an oversized file, or an embedding generation failure.

Documents are organised by domain. You assign a domain when ingesting, and domains become the unit of health monitoring. The Domain Health dashboard shows coverage statistics, freshness scores, and gap analysis per domain. Domains with high query volume but low document coverage are flagged so you know where to invest next.

Relation detection runs automatically during ingest. When a new document references or relates to existing documents, the system creates relationship links that power expanded retrieval. Parent-child document families, cross-references, and version chains are all detected and indexed.