Documentation Index

Fetch the complete documentation index at: https://docs.aifabrix.ai/llms.txt

Use this file to discover all available pages before exploring further.

Data sync and freshness

Prev Next

Enterprise Knowledge stays useful only when connected data reflects current business reality. Sync defines how records and documents enter the platform; freshness defines how stale data is detected, communicated, and governed.

Why it matters

AI recommendations built on week-old pipeline data or unsigned contract versions create operational and compliance risk. Architects and operators need a shared vocabulary for sync modes, schedules, and certification impact — without reading pipeline implementation details.

How it works

External system change
  ↓
Sync job (bulk or incremental)
  ↓
Platform copy + business metadata
  ↓
Capabilities + search/MCP surfaces
  ↓
Role Assistant tasks (with as-of context)
Mode Business meaning
Initial bulk load Baseline corpus before AI-assisted use
Incremental sync Ongoing updates from vendor change feeds or schedules
On-demand refresh Operator-triggered reload before sensitive tasks
Document ingestion File/metadata pipelines for document and vector stores

Freshness is not a single timestamp — it combines last successful sync, vendor latency, and certification scope. Operators treat failing sync as a trust incident: capabilities may remain defined but outcomes should not be presented as current.

For push-based updates, see Enterprise webhooks. For the full governed sync story, see Enterprise Synchronization Fabrix. To enable cross-datasource propagation after entities share a resource type, see Enterprise sync bridge — technical guide.

Freshness and certification

Operational certification proves behavior against systems that were reachable at test time. Agent metadata trust assumes field mappings describe live vendor fields. When vendors change APIs or schemas, integrators re-validate and re-run trust pillars — sync configuration alone does not guarantee continued freshness.

Architects document:

  • Maximum acceptable lag per datasource (e.g. CRM hourly, documents daily)
  • Who is notified on sync failure
  • Whether Role Assistants must show “as of” time in evidence

Example

A contract repository syncs nightly. A Legal Assistant summarizing amendments uses document capabilities backed by last-night’s index. If sync fails, operators see alerts; Role Assistant tasks defer or warn rather than silently using expired content.

Common mistakes

Mistake Impact
Certifying once, never monitoring sync Drift undetected
Mixing dev sandbox sync with prod workers Wrong business context
Ignoring document vs record sync differences Missing files or metadata

Limits

Sync capabilities depend on vendor APIs and entity type — not every system supports real-time incremental feeds. Confirm supported modes on the configure track for each datasource type.