Agentic Ingestion — Unique AI Documentation
Agentic Ingestion
Agentic Ingestion brings AI-powered enrichment capabilities into the Knowledge Base ingestion pipeline. These capabilities help administrators improve how documents are processed, enriched, and made searchable for end users.
This page groups the administrator-facing Agentic Ingestion features available in Knowledge Base.
Who This Is For
This section is intended for Knowledge Base administrators and space administrators who configure document ingestion behavior for folders, scopes, or chat uploads.
For infrastructure deployment, service configuration, Redis, feature flags, and runtime troubleshooting, see Agentic Ingestion Infrastructure.
Available Capabilities
| Capability | What It Does | Admin Page | Status |
|---|---|---|---|
| Agentic Metadata Extraction | Extracts structured metadata from ingested document content using a configurable schema and language model. | Agentic Metadata-Extraction | BETA |
| Agentic Image Content Extraction | Extracts text and meaning from charts, figures, diagrams, and other visual elements inside PDFs processed through Microsoft Document Intelligence. | Agentic Image Content Extraction | BETA |
Recommended Reading Order
- Start with Agentic Metadata-Extraction if you want to enrich documents with searchable metadata fields.
- Read Agentic Image Content Extraction if you want to improve extraction from charts, figures, diagrams, or image-heavy PDFs.
Configuration Scope
Agentic Ingestion settings can apply in different places depending on the upload path:
| Upload Path | Where Admins Configure It |
|---|---|
| Knowledge Base uploads | Folder or scope ingestion configuration |
| Chat uploads | Space / assistant advanced settings through ingestionConfig |
The exact configuration steps depend on the capability. See the individual subpages for setup instructions, examples, limitations, and troubleshooting guidance.