Content filtering
Checking what goes into a model and what comes out of it against content rules, and stopping or flagging a request that fails them. Distinct from data maskingData maskingReplacing the personal or confidential values in a prompt with placeholders before it leaves for a model, and putting them back in the answer. Anonymisation, where the values are gone for good, and pseudonymisation, where they return, are the two forms; which fields count as sensitive is the application's question, not the module's.SAP only, which hides values: a filter refuses content.
Overview
A module of the orchestration service in SAP AI CoreSAP AI CoreSAP's AI service on BTP: the runtime for AI assets, and through the Generative AI Hub inside it the access to foundation models, the orchestration service with its templating, masking, filtering and grounding modules, the Prompt Registry and Document Grounding.sap official, btp, run on the input and on the output separately; SAP names hateful, violent, sexual and vulgar content as the categories.
Not covered: At those categories and those two points. Prompt injection, a tool result that lies and whether an answer is right are not content filtering; the first of them sits with tool governanceTool governanceControlling what an agent may call, with which authorisations, and leaving an audit trail of what it did. The authorisation question that MCP exposure raises but does not answer.SAP and community, unassessed.
Every provider behind this door is SAP's, so core-only is a assessment of the corpus rather than a quote from the core.
No community artifact is documented here.
@sap-ai-sdk/orchestration@sap-ai-sdk/orchestrationThe Orchestration Service client of SAP's AI SDK, which applies templating, grounding, filtering and masking around a model call rather than in the application.sap official, cap exposes the same input and output filters to a Node.js or CAP service, as part of each orchestration configuration.
Not covered: At the SDK call, which is the same boundary as the service's.
No community artifact is documented here.
Artifacts
Content
Seen working
Every CDS service in both samples is marked `@requires: 'any'` and left publicly accessible, which the README says simplifies the testing and deployment process, and it asks for proper authentication mechanisms to be applied in their place.
Details
Coverage
For your agent
- Ref
capability:content-filtering- JSON
- /capabilities/content-filtering.json
- MCP
- Connecting an agent