AIDP Knowledge Bases currently use an ingestion-based model where source content is ingested, chunked and indexed for use by AI Agents. This provides an effective point-in-time knowledge source, but the Knowledge Base must be refreshed as the authoritative source changes.
It would be useful for AIDP to support two Knowledge Base models, similar conceptually to standard and external catalogues:
- Snapshot / Ingested Knowledge Base: the current model, where content is ingested and indexed within AIDP.
- Live / External Knowledge Base: content remains in the authoritative external repository and relevant knowledge is retrieved at query time.
For example, a SharePoint document library could either be periodically ingested into AIDP or accessed as a live knowledge source, allowing an Agent to use the latest available information without maintaining a separate snapshot.
Providing both options would allow organisations to choose the appropriate approach based on freshness, performance, security and governance requirements, while potentially supporting SharePoint and other enterprise knowledge repositories.
