SAP AI Core LLM icon

SAP AI Core LLM

AI Agent powered by SAP AI Core — the Deployment ID determines which model and pipeline to use

SAP AI Core LLM

AI-generated

Summary

Send a user message to an SAP AI Core deployment and return its response. Optionally attach n8n AI tools for model-directed function calls, and configure model generation and SAP masking, grounding, or filtering modules.

Inputs

  • Deployment ID (required) — SAP AI Core deployment ID, which selects the deployment and its pipeline.
  • Model Name (required) — Model configured in the deployment; defaults to gpt-4o-mini.
  • Resource Group (required) — SAP AI Core resource group sent with each request; defaults to default.
  • User Message (required) — Message sent to the model. Defaults to the incoming item's chatInput, then message, then a greeting.
  • Tools input — Optional connected AI tools. The model can request tool calls; the node executes them and sends their results back to the model, up to Max Tool Iterations.
  • Options — Set a system message (default: “You are a helpful AI assistant.”), temperature, maximum tokens, top-p, frequency and presence penalties, tool iteration limit, and verbose logging. JSON configuration fields optionally supply SAP data masking, grounding, and content filtering settings.

Output shape

One output item per input item, each containing a JSON object.

Successful items include output text, deploymentId, model, resourceGroup, userMessage, toolsUsed, intermediateSteps, and iterations; systemMessage is included when set. toolErrors is included when tool loading reports errors, and debug is included when verbose logging is enabled. If the model does not produce a final answer before the iteration limit, output contains a limit-reached message. With Continue On Fail enabled, errors are returned as error objects; otherwise execution errors are thrown.

Examples

Example 1: Call a model

Provide deployment and model names, resource group, and a user message. Optionally set generation controls or SAP module configurations.

Example 2: Use connected tools

Connect AI tools; the model can call them, and the node returns intermediate step details with its final answer.

Links

Discussion