Data Retention vs. Processing
What elvex Retains
For File-Based and Website Datasources (Notion, Confluence, Wikis):
elvex stores the following in its secure database:
Raw content: Text of every page, document, and file
Original file metadata: Creation dates, modification timestamps, file names, URLs
Processed text chunks: Content split into semantic segments (typically 2-4 sentences each)
Vector embeddings: Numerical representations of each chunk for semantic search
Relationship metadata: Links between pages, hierarchical structures, cross-references
What elvex Processes Temporarily
elvex does not retain database content:
Query results only: Only the specific data returned by generated SQL queries is temporarily processed
No pre-ingestion: Database content is not crawled, indexed, or stored in advance
Session-based processing: Query results are processed only for the duration of the user request
No persistent storage: Database content does not become part of elvex's permanent knowledge base
Specific Behavior for Common Knowledge Management Systems
Notion Integration
When connecting Notion to elvex:
What gets retained:
All page content including nested pages and databases
Page properties, tags, and metadata
Access mechanism:
Uses OAuth and matches the connecting users permissions
Requires explicit permission grants for each Notion workspace
Can access pages and subpages provided credentials can reach
Sync behavior:
elvex subscribes to files so is notified when there is a change rather than periodic syncs
Confluence Integration
When connecting Confluence to elvex:
What gets retained:
All accessible pages and blog posts
Page metadata, labels, and space information
Macro-generated content and dynamic elements (rendered as static text)
Access mechanism:
Uses OAuth and matches the connecting users permissions
Can traverse space hierarchies and cross-space links
Sync behavior:
elvex subscribes to files so is notified when there is a change rather than periodic syncs
Website/Wiki Crawling
For public or authenticated websites:
What gets retained:
All crawled page content within configured depth limits
Page metadata and URL structures
Navigation structures and internal linking
Crawling scope:
Configurable maximum depth (1-7 levels from starting URL)
Configurable page limits (1-100 pages per website)
Include/exclude rules using regular expressions
Custom authentication headers for protected content