SAP AI Core Generative AI Hub

https://help.sap.com/docs/sap-ai-core/generative-ai/
Admin
The generative AI hub is a capability of SAP AI Core that enables access to and orchestration of...
Tokens:16,979
Snippets:119
Trust Score:9.8
Update:2 months ago
Show doc for...
Tokens:
Raw
### Create Optimization Configuration via cURL

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/create-configuration-for-optimization?version=CLOUD

Example of a POST request to the SAP AI Core API to create an optimization configuration with required bindings.

```bash
curl --request POST "$AI_API_URL/v2/lm/configurations" \
--header "Authorization: Bearer $TOKEN" \
--header "AI-Resource-Group: $RESOURCE_GROUP" \
--header "Content-Type: application/json" \
--data '{
    "name": "genai-eval-conf",
    "scenarioId": "genai-optimizations ",
    "executableId": "genai-optimizations ",
    "inputArtifactBindings": [
        {
            "key": "datasetFolder",
            "artifactId": "1f50fbf8-18d6-4e72-ae10-04fdbc087815"
        }
    ],
    "parameterBindings": [
        {
            "key": "basePrompt",
            "value": "<scenario>/<promptName>:<promptVersion>"
        },
        {
            "key": "modelParams",
            "value": "{"<modelName1>:<modelVersion1>":{"temperature":0.7,"max_tokens":1000},"<modelName2>:<modelVersion2>":{"temperature":0.9}}"
        },
        {
            "key": "dataset",
            "value": "data.json"
        {
            "key": "optimizationMetric",
            "value": "JSON_Match”
        },
        {
            "key": "variableMapping",
            "value": "{"<prompt_variable1>":"<data_variable1>"}"
        },
        {
            "key": "promptTemplateScope",
            "value": "resourcegroup"
        }, 
        {

            "key": "targetModels",
            "value": "<modelName>:<modelVersion>”
        },
        {
            "key": "targetPromptMapping",
            "value": “<modelName>:<modelversion>=<promptName>:<promptVersion>”
        },
    ]
}‘
```

--------------------------------

### GET /v2/lm/scenarios/foundation-models/models

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/choose-model?version=CLOUD

Retrieves a list of all available foundation models, including their executable IDs, model names, and version details.

```APIDOC
## GET /v2/lm/scenarios/foundation-models/models

### Description
Retrieves a list of available foundation models that can be deployed within SAP AI Core.

### Method
GET

### Endpoint
$AI_API_URL/v2/lm/scenarios/foundation-models/models

### Request Headers
- **AI-Resource-Group** (string) - Required - The resource group identifier (e.g., 'default').
- **Authorization** (string) - Required - Bearer token for authentication.

### Response
#### Success Response (200)
- **count** (integer) - Number of models available.
- **resources** (array) - List of available models.
  - **executableId** (string) - Executable under which the model is available.
  - **model** (string) - The model name.
  - **description** (string) - Description of the model.
  - **versions** (array) - List of available model versions.
    - **name** (string) - The model version name.
    - **isLatest** (boolean) - Indicates if the version is the latest.
```

--------------------------------

### List available foundation models using cURL

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/choose-model?version=CLOUD

Sends a GET request to the foundation models endpoint to retrieve a list of available models.

```bash
curl --location '$AI_API_URL/v2/lm/scenarios/foundation-models/models' \
--header 'AI-Resource-Group: default' \
--header 'Authorization: Bearer $AUTH_TOKEN'
```

--------------------------------

### Example JSON response for foundation models

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/choose-model?version=CLOUD

A sample JSON structure returned by the foundation models API, showing model metadata and configuration parameters.

```json
{
  "count": 41,
  "resources": [
    {
      "accessType": "Remote",
      "allowedScenarios": [
        {
          "executableId": "azure-openai",
          "scenarioId": "foundation-models"
        },
        {
          "executableId": "orchestration",
          "scenarioId": "orchestration"
        }
      ],
      "description": "OpenAI GPT-4o engineered for speed and efficiency, supporting text, images and audio using Chat Completions API",
      "displayName": "GPT-4o",
      "executableId": "azure-openai",
      "model": "gpt-4o",
      "provider": "OpenAI",
      "versions": [
        {
          "capabilities": [
            "text-generation",
            "image-recognition"
          ],
          "contextLength": 128000,
          "cost": [
            {
              "inputCost": "0.00312"
            },
            {
              "outputCost": "0.0092"
            }
          ],
          "deprecated": false,
          "inputTypes": [
            "text",
            "image",
            "audio"
          ],
          "isLatest": false,
          "metadata": [
            {
              "meanWinRate": "0.938"
            },
            {
              "chatBotArenaScore": "1285"
            },
            {
              "airBenchRefusalRate": "0.528"
            }
          ],
          "name": "2024-05-13",
          "retirementDate": "",
          "streamingSupported": true
        }, ...

    ]
}
```

--------------------------------

### Retrieval Output Structure

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/contextualized-retrieval-using-metadata-and-vector-search-911c1ca88d3243e4ab34509d08ceb962?version=CLOUD

Example of the JSON structure returned by the retrieval process, containing content and associated metadata.

```json
[
    [
        {
            "content": "Joule is the AI copilot that truly understands your business. Joule revolutionizes how you interact with your SAP business systems, making every touchpoint count and every task simpler.",
            "metadata": {
                "source": [
                    "/tmp/document_1.pdf"
                ],
                "webUrl": [
                    "https://<sharepoint_url>/sites/<Documents>/sap_joule.pdf"
                ],
                "document_webUrl": [
                    "https://<sharepoint_url>/sites/<Documents>/sap_joule.pdf"
                ]
            }
        },
        {
            "content": "AI-powered capabilities in Joule with custom-developed enterprise skills tailored to your company's specific needs. By integrating with both SAP and third-party applications, Joule can instantly apply its newfound knowledge within SAP systems.",
            "metadata": {
                "source": [
                    "/tmp/document_1.pdf"
                ],
                "webUrl": [
                    "https://<sharepoint_url>/sites/<Documents>/sap_joule.pdf"
                ],
                "document_webUrl": [
                    "https://<sharepoint_url>/sites/<Documents>/sap_joule.pdf"
                ]
            }
        }
    ]
]
```

--------------------------------

### Chat Inference Response Structure

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/chat?version=CLOUD

Example JSON response showing the structure of chat history and orchestration results after an inference request.

```json
{
  "request_id": "5445f8d8-8b68-43c3-a149-26c1e6a88a22",
  "module_results": {
    "templating": [
      {
        "role": "user",
        "content": "Create 3 paraphrases of I love you."
      },
      {
        "role": "assistant",
        "content": "1. I have deep affection for you.\n2. My feelings for you are strong and caring.\n3. You mean everything to me."
      },
      {
        "role": "user",
        "content": "Can you please make them more romantic."
      },
      {
        "role": "assistant",
        "content": "1. My heart is filled with an intense passion for you.\n2. You are the light of my life, and my love for you knows no bounds.\n3. I cherish you more than words can express, my dear."
      },
      {
        "role": "user",
        "content": "Try sounding like shakespeare."
      },
      {
        "role": "assistant",
        "content": "1. Thou art the sun in my sky, and my love for thee doth burn with unyielding fervor.\n2. Within mine heart, thou hast ignited a flame of ardor that doth shine brightly for thee, my dearest.\n3. My love for thee doth surpass all measure, for thou art the true essence of my being."
      },
      {
        "role": "user",
        "content": "And now two more that sound like Goethe."
      }
    ],
    "llm": {
      "id": "chatcmpl-9kXqisJKnuNv1B4eXTUzqZEJSmzdC",
      "object": "chat.completion",
      "created": 1720880232,
      "model": "<ModelName>",
      "choices": [
        {
          "index": 0,
          "message": {
            "role": "assistant",
            "content": "1. In thy presence, my soul finds solace, for thou art the embodiment of love's sweetest melody.\n2. Like a gentle breeze upon a summer's eve, thy love doth caress my heart and fill it with eternal longing."
          },
          "finish_reason": "stop"
        }
      ],
      "usage": {
        "completion_tokens": 51,
        "prompt_tokens": 212,
        "total_tokens": 263
      }
    }
  },
  "orchestration_result": {
    "id": "chatcmpl-9kXqisJKnuNv1B4eXTUzqZEJSmzdC",
    "object": "chat.completion",
    "created": 1720880232,
    "model": "<ModelName>",
    "choices": [
      {
        "index": 0,
        "message": {
          "role": "assistant",
          "content": "1. In thy presence, my soul finds solace, for thou art the embodiment of love's sweetest melody.\n2. Like a gentle breeze upon a summer's eve, thy love doth caress my heart and fill it with eternal longing."
        },
        "finish_reason": "stop"
      }
    ],
    "usage": {
      "completion_tokens": 51,
      "prompt_tokens": 212,
      "total_tokens": 263
    }
  }
}
```

--------------------------------

### Example Output for Titan Embeddings

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/aws-bedrock?version=CLOUD

The expected JSON response structure containing the embedding vector and token count.

```json
{
  "embedding": [
    -0.0013275146,
    -0.024658203,
    ...
    0.068359375,
    0.014404297,
    0.025512695
  ],
  "inputTextTokenCount": 3
}
```

--------------------------------

### Chat Inference Response Structure

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/chat-cd16d739cf7b4c8aa5e3b12062ae6ce9?version=CLOUD

Example JSON response showing the chat history in the templating field and the final LLM completion result.

```json
{
  "request_id": "9a66158b-b019-951a-a5e7-3c3ec9d2662e",
  "intermediate_results": {
    "templating": [
      {
        "content": "Create 3 paraphrases of I love you.",
        "role": "user"
      },
      {
        "role": "assistant",
        "content": "1. I have deep affection for you.\n2. My feelings for you are strong and caring.\n3. You mean everything to me."
      },
      {
        "content": "Can you please make them more romantic.",
        "role": "user"
      },
      {
        "role": "assistant",
        "content": "1. My heart is filled with an intense passion for you.\n2. You are the light of my life, and my love for you knows no bounds.\n3. I cherish you more than words can express, my dear."
      },
      {
        "content": "Try sounding like shakespeare.",
        "role": "user"
      },
      {
        "role": "assistant",
        "content": "1. Thou art the sun in my sky, and my love for thee doth burn with unyielding fervor.\n2. Within mine heart, thou hast ignited a flame of ardor that doth shine brightly for thee, my dearest.\n3. My love for thee doth surpass all measure, for thou art the true essence of my being."
      },
      {
        "content": "And now two more that sound like Goethe.",
        "role": "user"
      }
    ],
    "llm": {
      "id": "chatcmpl-BuHIHDDcx0FPsOMVEXYp8eVKA4ulw",
      "object": "chat.completion",
      "created": 1752752545,
      "model": "gpt-4o-mini-2024-07-18",
      "system_fingerprint": "fp_efad92c60b",
      "choices": [
        {
          "index": 0,
          "message": {
            "role": "assistant",
            "content": "1. In the depths of my soul, I find an eternal longing for thee, a love that transcends the bounds of time and space.\n2. Thou art the muse of my existence, inspiring a passion within me that resonates with the very essence of nature itself."
          },
          "finish_reason": "stop"
        }
      ],
      "usage": {
        "completion_tokens": 55,
        "prompt_tokens": 212,
        "total_tokens": 267
      }
    }
  },
  "final_result": {
    "id": "chatcmpl-BuHIHDDcx0FPsOMVEXYp8eVKA4ulw",
    "object": "chat.completion",
    "created": 1752752545,
    "model": "gpt-4o-mini-2024-07-18",
    "system_fingerprint": "fp_efad92c60b",
    "choices": [
      {
        "index": 0,
        "message": {
          "role": "assistant",
          "content": "1. In the depths of my soul, I find an eternal longing for thee, a love that transcends the bounds of time and space.\n2. Thou art the muse of my existence, inspiring a passion within me that resonates with the very essence of nature itself."
        },
        "finish_reason": "stop"
      }
    ],
    "usage": {
      "completion_tokens": 55,
      "prompt_tokens": 212,
      "total_tokens": 267
    }
  }
}
```

--------------------------------

### Rerank 3.5 API Response

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/cohere?version=CLOUD

Example of the JSON response returned by the Rerank 3.5 model, containing relevance scores for the provided documents.

```json
{
  "id": "<id>",
  "meta": {
    "api_version": {
      "version": "1"
    },
    "billed_units": {
      "search_units": 1
    }
  },
  "results": [
    {
      "index": 2,
      "relevance_score": 0.85157084
    },
    {
      "index": 4,
      "relevance_score": 0.23100689
    },
    {
      "index": 1,
      "relevance_score": 0.056014005
    }
  ]
}
```

--------------------------------

### POST /v2/lm/configurations

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/create-configuration-for-optimization?version=CLOUD

Creates a new configuration for prompt optimization within the specified resource group.

```APIDOC
## POST /v2/lm/configurations

### Description
Creates a configuration for a prompt optimization, defining parameters, artifacts, and model settings required for execution.

### Method
POST

### Endpoint
$AI_API_URL/v2/lm/configurations

### Request Body
- **scenarioId** (string) - Required - Must be set to "genai-optimizations"
- **executableId** (string) - Required - Must be set to "genai-optimizations"
- **inputArtifactBindings** (object) - Required - Contains the artifactId for the registered dataset
- **parameterBindings** (object) - Required - Contains "basePrompt" (reference to prompt in registry) and optional "baseModel" and "modelParams"
- **dataset** (string) - Optional - Path to the JSON dataset in the input artifact
- **trainDataset** (string) - Optional - Path to the training dataset (required if testDataset is provided)
- **testDataset** (string) - Optional - Path to the test dataset (required if trainDataset is provided)
- **promptTemplateScope** (string) - Optional - Scope of the template (tenant or resourcegroup)
- **optimizationMetric** (string) - Optional - System defined optimization metric
- **customMetricId** (string) - Optional - ID of a custom metric
```

--------------------------------

### GET /llm-batch-service/v1/batches/$BATCH_ID/status

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/check-batch-status?version=CLOUD

Retrieves the current status of a specific batch job using its ID.

```APIDOC
## GET /llm-batch-service/v1/batches/$BATCH_ID/status

### Description
Retrieves the current status of an asynchronous batch job. Use this endpoint to poll for updates until the job reaches a terminal state.

### Method
GET

### Endpoint
$AI_API/llm-batch-service/v1/batches/$BATCH_ID/status

### Parameters
#### Path Parameters
- **$BATCH_ID** (string) - Required - The unique identifier of the batch job.

### Request Headers
- **Authorization** (string) - Required - Bearer $TOKEN
- **AI-Resource-Group** (string) - Required - The resource group used in the activation steps

### Request Example
curl -X GET "$AI_API/llm-batch-service/v1/batches/$BATCH_ID/status" \
    -H 'AI-Resource-Group: <resource_group>' \
    -H 'Content-Type: application/json' \
    -H "Authorization: Bearer $AUTH_TOKEN"

### Response
#### Success Response (200)
- **current_status** (string) - The current status of the batch (e.g., PENDING, IN_PROGRESS, COMPLETED, FAILED).
- **target_status** (string) - The intended final status of the batch.
- **updated_at** (string) - Timestamp of the last status update.
- **message** (string) - Descriptive message regarding the current status.

#### Response Example
{
  "current_status": "IN_PROGRESS",
  "target_status": "COMPLETED",
  "updated_at": "2026-01-07T10:35:00Z",
  "message": "Processing batch requests"
}
```

--------------------------------

### Include Metadata in Orchestration Completion Request

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/contextualized-retrieval-using-metadata-and-vector-search-911c1ca88d3243e4ab34509d08ceb962?version=CLOUD

Configure the grounding module with metadata_params to include specific fields like source and webUrl in the output.

```bash
curl --request POST \
  --url {{ORCH_DEPLOYMENT_URL}}/v2/completion \
  --header 'Authorization: Bearer <access_token>' \
  --header 'AI-Resource-Group: <resource_group>' \
  --header 'content-type: application/json' \
  --data '{
    "config": {
        "modules": {
            "grounding": {
                "type": "document_grounding_service",
                "config": {
                    "filters": [
                        {
                            "id": "filter1",
                            "data_repositories": [
                                "*"
                            ],
                            "search_config": {
                                "max_chunk_count": 100
                            },
                            "data_repository_type": "vector"
                        }
                    ],
                    "placeholders": {
                        "input": [
                            "groundingRequest"
                        ],
                        "output": "groundingOutput"
                    },
                    "metadata_params": [
                        "source",
                        "webUrl"
                    ]

                }
            },
            "prompt_templating": {
                "prompt": {
                    "template": [
                        {
                            "role": "user",
                            "content": "<prompt>: {{?groundingRequest}}\n\nReports: {{?groundingOutput}}"
                        }
                    ]
                },
                "model": {
                    "name": "<modelName>",
                    "version": "<modelVersion>",
                    "params": {
                        "max_completion_tokens": 50,
                        "temperature": 0.1,
                        "frequency_penalty": 0,
                        "presence_penalty": 0
                    }
                }
            }
        }
    },
    "placeholder_values": {
        "groundingRequest": "what is SAP Joule?"
    }
}'
```

--------------------------------

### Configure Train and Test Datasets

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/create-configuration-for-optimization?version=CLOUD

Snippet showing how to replace standard dataset keys with specific train and test dataset keys in the configuration.

```json
{
            "key": "trainDataset",
            "value": "train-dataset.json",
          }
          {
            "key": "testDataset",
            "value": "test-dataset.json",
          }
```

--------------------------------

### POST /v2/completion

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/consume-models-with-harmonized-api?version=CLOUD

Performs a minimal inference call by configuring the prompt templating and model modules.

```APIDOC
## POST /v2/completion

### Description
Performs a minimal inference call to the Harmonized API. This request configures the templating module with a user message and parameters, and specifies the model to be used for completion.

### Method
POST

### Endpoint
$ORCH_DEPLOYMENT_URL/v2/completion

### Request Body
- **config** (object) - Required - Contains the modules configuration including prompt_templating and model.
- **placeholder_values** (object) - Required - Contains the values for the parameters defined in the prompt template.

### Request Example
{
  "config": {
    "modules": {
      "prompt_templating": {
        "prompt": {
          "template": [
            {
              "role": "user",
              "content": "Reply with `{{?text}}` in {{?language}}"
            }
          ],
          "defaults": {
            "language": "English"
          }
        },
        "model": {
          "name": "<model>",
          "params": {
            "max_completion_tokens": 50,
            "temperature": 0.1,
            "frequency_penalty": 0,
            "presence_penalty": 0
          }
        }
      }
    }
  },
  "placeholder_values": {
    "text": "Orchestration is Working!",
    "language": "German"
  }
}

### Response
#### Success Response (200)
- **request_id** (string) - Unique identifier for the request.
- **intermediate_results** (object) - Results from each executed module.
- **final_result** (object) - The orchestration result, including the model response.

#### Response Example
{
  "request_id": "d8efd339-8ae1-9810-a076-22360875a2fe",
  "intermediate_results": {
    "templating": [
      {
        "content": "Reply with `Orchestration is Working!` in German",
        "role": "user"
      }
    ],
    "llm": {
      "id": "chatcmpl-D2VKMLhpxnszJKVXeNP0k6OcJTQDK",
      "choices": [
        {
          "message": {
            "role": "assistant",
            "content": "Die Orchestrierung funktioniert!"
          }
        }
      ]
    }
  },
  "final_result": {
    "id": "chatcmpl-D2VKMLhpxnszJKVXeNP0k6OcJTQDK",
    "choices": [
      {
        "message": {
          "role": "assistant",
          "content": "Die Orchestrierung funktioniert!"
        }
      }
    ]
  }
}
```

--------------------------------

### Metadata Naming Convention

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/contextualized-retrieval-using-metadata-and-vector-search-911c1ca88d3243e4ab34509d08ceb962?version=CLOUD

Demonstrates how metadata keys are resolved when present at both chunk and document levels.

```json
"metadata": {
  "webUrl": "<value from chunk metadata>",
  "document_webUrl": "<value from document metadata>"
}
```

--------------------------------

### Prompt Templating with Grounding Data

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/contextualized-retrieval-using-metadata-and-vector-search-911c1ca88d3243e4ab34509d08ceb962?version=CLOUD

Configuration for incorporating grounding requests and outputs into prompt templates.

```json
"prompt_templating": {
  "messages_history": [
    {
      "role": "user",
      "content": "You are a helpful assistant for any queries for SAP Teched 2024.\\nAnswer the grounding request by providing relevant answers that fit to the request. \\n\\nRequest: {{ ?groundingRequest }}\\n\\nReports:{{ ?groundingOutput }} Alongwith the original response, provide another response by making use of metadata received in the grounding output to provide more optimal response. Explain how the metadata was used."
     }
      ],
      "defaults": {}
  }
```

--------------------------------

### Configure Chat Orchestration via cURL

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/chat?version=CLOUD

Sends a POST request to the orchestration deployment URL with templating and LLM configurations, including message history for context.

```bash
curl --request POST $ORCH_DEPLOYMENT_URL/completion \
    --header 'content-type: application/json' \
    --header "Authorization: Bearer $TOKEN" \
    --header "AI-Resource-Group: $RESOURCE_GROUP" \
    --data-raw '{
    "orchestration_config": {
      "module_configurations": {
        "templating_module_config": {
          "template": [
            {
              "role": "user",
              "content": "And now two more that sound like Goethe."
            }
          ]
        },
        "llm_module_config": {
          "model_name": "<ModelName>",
          "model_params": {
            "max_tokens": 300,
            "temperature": 0.1,
            "frequency_penalty": 0,
            "presence_penalty": 0
          }
        }
      }
    },  
    "input_params": {
    },
    "messages_history": [
      {
        "role":"user",
        "content":"Create 3 paraphrases of I love you."
      },
      {
        "role":"assistant",
        "content":"1. I have deep affection for you.\n2. My feelings for you are strong and caring.\n3. You mean everything to me."
      },
      {
        "role":"user",
        "content":"Can you please make them more romantic."
      },
      {
        "role":"assistant",
        "content":"1. My heart is filled with an intense passion for you.\n2. You are the light of my life, and my love for you knows no bounds.\n3. I cherish you more than words can express, my dear."
      },
      {
        "role":"user",
        "content":"Try sounding like shakespeare."
      },
      {
        "role":"assistant",
        "content":"1. Thou art the sun in my sky, and my love for thee doth burn with unyielding fervor.\n2. Within mine heart, thou hast ignited a flame of ardor that doth shine brightly for thee, my dearest.\n3. My love for thee doth surpass all measure, for thou art the true essence of my being."
      }
    ]
  }'
```

--------------------------------

### Configure Chat Prompt Templating via cURL

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/chat-cd16d739cf7b4c8aa5e3b12062ae6ce9?version=CLOUD

Use this request to send a chat completion task to the orchestration deployment. Ensure the ORCH_DEPLOYMENT_URL, TOKEN, and RESOURCE_GROUP variables are correctly set in your environment.

```bash
curl --request POST $ORCH_DEPLOYMENT_URL/v2/completion \
    --header 'content-type: application/json' \
    --header "Authorization: Bearer $TOKEN" \
    --header "AI-Resource-Group: $RESOURCE_GROUP" \
    --data-raw '{
    "config": {
      "modules": {
        "prompt_templating": {
          "prompt": {
            "template": [
              {
                "role": "user",
                "content": "And now two more that sound like Goethe."
              }
            ]
          },
          "model": {
            "name": "gpt-4o-mini",
            "params": {
              "max_completion_tokens": 300,
              "temperature": 0.1,
              "frequency_penalty": 0,
              "presence_penalty": 0
            }
          }
        }
      }
    },  
    "placeholder_values": {
    },
    "messages_history": [
      {
        "role":"user",
        "content":"Create 3 paraphrases of I love you."
      },
      {
        "role":"assistant",
        "content":"1. I have deep affection for you.\n2. My feelings for you are strong and caring.\n3. You mean everything to me."
      },
      {
        "role":"user",
        "content":"Can you please make them more romantic."
      },
      {
        "role":"assistant",
        "content":"1. My heart is filled with an intense passion for you.\n2. You are the light of my life, and my love for you knows no bounds.\n3. I cherish you more than words can express, my dear."
      },
      {
        "role":"user",
        "content":"Try sounding like shakespeare."
      },
      {
        "role":"assistant",
        "content":"1. Thou art the sun in my sky, and my love for thee doth burn with unyielding fervor.\n2. Within mine heart, thou hast ignited a flame of ardor that doth shine brightly for thee, my dearest.\n3. My love for thee doth surpass all measure, for thou art the true essence of my being."
      }
    ]
  }'
```

--------------------------------

### POST /converse

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/aws-bedrock?version=CLOUD

Invokes a standard inference request to an Amazon Bedrock model.

```APIDOC
## POST /converse

### Description
Invokes a standard inference request to an Amazon Bedrock model using the specified deployment URL.

### Method
POST

### Endpoint
$DEPLOYMENT_URL/converse

### Parameters
#### Headers
- **AI-Resource-Group** (string) - Required - The Resource Group ID.
- **Content-Type** (string) - Required - application/json.
- **Authorization** (string) - Required - Bearer $AUTH_TOKEN.

#### Request Body
- **inferenceConfig** (object) - Optional - Configuration for inference including maxTokens, stopSequences, and temperature.
- **messages** (array) - Required - A list of message objects containing content and role.
```

--------------------------------

### Execute Minimal Inference Request

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/consume-models-with-harmonized-api?version=CLOUD

Use this curl command to send a POST request to the Harmonized API with prompt templating and model configuration.

```bash
curl --request POST $ORCH_DEPLOYMENT_URL/v2/completion \
--header 'content-type: application/json' \
--header "Authorization: Bearer $TOKEN" \
--header "AI-Resource-Group: $RESOURCE_GROUP" \
--data-raw '{
  "config": {
   "modules": {
     "prompt_templating": {
       "prompt": {
         "template": [
           {
             "role": "user",
             "content": "Reply with `{{?text}}` in {{?language}}"
           }
         ],
         "defaults": {
           "language": "English"
         }
       },
       "model": {
         "name": "<model>",
         "params": {
           "max_completion_tokens": 50,
           "temperature": 0.1,
           "frequency_penalty": 0,
           "presence_penalty": 0
         }
       }
     }
   }
  },
  "placeholder_values": {
   "text": "Orchestration is Working!",
   "language": "German"
  }
}'
```

--------------------------------

### POST /converse (Anthropic Claude)

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/aws-bedrock?version=CLOUD

Uses the converse interface to interact with Anthropic Claude models with specific inference configurations.

```APIDOC
## POST /converse

### Description
Sends a conversation request to an Anthropic Claude model deployment.

### Method
POST

### Endpoint
$DEPLOYMENT_URL/converse

### Parameters
#### Request Body
- **inferenceConfig** (object) - Optional - Configuration for inference including maxTokens and stopSequences.
- **messages** (array) - Required - A list of message objects containing role and content.
```

--------------------------------

### Retrieve Supported Metadata Keys via cURL

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/contextualized-retrieval-using-metadata-and-vector-search-911c1ca88d3243e4ab34509d08ceb962?version=CLOUD

Use this request to query the document grounding retrieval search endpoint to identify available metadata keys.

```bash
curl --request POST \
  --url {{AI_API_URL}}/v2/lm/document-grounding/retrieval/search \
  --header 'AI-Resource-Group: <resource_group>' \
  --header 'Authorization: Bearer <access_token>' \
  --header 'Content-Type: application/json' \
  --data '{
    "query": "what is SAP Joule",
    "filters": [
        {
            "id": "string",
            "searchConfiguration": {
                "maxChunkCount": 5
            },
            "dataRepositories": [
                "*"
            ],
            "dataRepositoryType": "vector",
            "dataRepositoryMetadata": [
                {
                    "key": "type",
                    "value": [
                        "custom"
                    ]
                }
            ],
            "documentMetadata": [],
            "chunkMetadata": []
        }
    ]
}'
```

--------------------------------

### POST /completion

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/chat?version=CLOUD

Sends a chat completion request to the orchestration deployment, utilizing specific templating and LLM configurations.

```APIDOC
## POST /completion

### Description
Sends a chat completion request to the orchestration deployment. This endpoint allows for configuring the templating module and LLM parameters to generate responses based on provided input and message history.

### Method
POST

### Endpoint
$ORCH_DEPLOYMENT_URL/completion

### Request Body
- **orchestration_config** (object) - Required - Configuration for the orchestration modules.
- **input_params** (object) - Optional - Input parameters for the request.
- **messages_history** (array) - Optional - The history of the conversation to provide context to the model.

### Request Example
{
  "orchestration_config": {
    "module_configurations": {
      "templating_module_config": {
        "template": [
          {
            "role": "user",
            "content": "And now two more that sound like Goethe."
          }
        ]
      },
      "llm_module_config": {
        "model_name": "<ModelName>",
        "model_params": {
          "max_tokens": 300,
          "temperature": 0.1,
          "frequency_penalty": 0,
          "presence_penalty": 0
        }
      }
    }
  },
  "input_params": {},
  "messages_history": [
    {
      "role": "user",
      "content": "Create 3 paraphrases of I love you."
    }
  ]
}
```

--------------------------------

### Invoke Amazon Nova models via REST API

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/aws-bedrock?version=CLOUD

Use the /converse endpoint to send inference requests to Amazon Nova models. Requires valid authentication tokens and resource group identifiers.

```bash
curl --location '$DEPLOYMENT_URL/converse' \
--header 'AI-Resource-Group: <Resource Group Id>' \
--header 'Content-Type: application/json' \
--header "Authorization: Bearer $AUTH_TOKEN" \
--data '{
    "inferenceConfig": {
        "maxTokens": 100,
        "stopSequences": [
            "blab"
        ],
        "temperature": 0.7
    },
    "messages": [
        {
            "content": [
                {
                    "text": "Perplexity means?"
                }
            ],
            "role": "user"
        }
    ]
}'
```

--------------------------------

### Invoke Cohere Command A Reasoning via cURL

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/cohere?version=CLOUD

Use this request to interact with the Cohere Command A Reasoning model. Ensure the DEPLOYMENT_URL, Resource Group Id, and AUTH_TOKEN are correctly configured.

```bash
curl --location "$DEPLOYMENT_URL/chat" \
--header 'AI-Resource-Group: <Resource Group Id>' \
--header 'Content-Type: application/json' \
--header "Authorization: Bearer $AUTH_TOKEN" \
--data '{
    "model": "cohere--command-a-reasoning",
    "stream": false,
    "frequency_penalty": 0.8,
    "thinking": {
        "type": "enabled"
    },
    "messages": [
        {
            "role": "user",
            "content": "Tell me about the reflection?"
        }
    ]
}'
```

--------------------------------

### Create a Custom Metric via cURL

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/create-custom-metric?version=CLOUD

Use this cURL command to register a new evaluation metric. Ensure the AI-Resource-Group, TOKEN, and AI_API_URL variables are correctly set in your environment.

```bash
curl --request POST \
  --url $AI_API_URL/v2/lm/evaluationMetrics \
  --header 'AI-Resource-Group: $RESOURCE_GROUP' \
  --header 'Content-Type: application/json' \ 
  --header 'Authorization: Bearer $TOKEN'
  --data '{
	  "scenario": "qa-evaluation",
	  "name": "factual-accuracy",
	  "version": "1.0",
	  "description": "Measures factual correctness of responses",
	  "evaluationMethod": "llm-as-a-judge",
	  "usageType": ["evaluation"],
	  "includeProperties": ["prompt", "reference"],
	  "spec": ...  // Custom Metric Definition
}'
```

--------------------------------

### Converse with Anthropic Claude Models

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/aws-bedrock?version=CLOUD

Use the /converse endpoint for multi-turn or structured conversational interactions with Claude models.

```bash
curl --location '$DEPLOYMENT_URL/converse' \
--header 'AI-Resource-Group: <Resource Group Id>' \
--header 'Content-Type: application/json' \
--header "Authorization: Bearer $AUTH_TOKEN" \
--data '{
    "inferenceConfig": {
        "maxTokens": 100,
        "stopSequences": [
            "blab"
        ]
    },
    "messages": [
        {
            "content": [
                {
                    "text": "Perplexity means?"
                }
            ],
            "role": "user"
        }
    ]
}'
```

--------------------------------

### POST /v2/completion

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/chat-cd16d739cf7b4c8aa5e3b12062ae6ce9?version=CLOUD

Sends a chat completion request to the orchestration deployment, utilizing the prompt templating module and message history.

```APIDOC
## POST /v2/completion

### Description
Sends a chat completion request to the orchestration deployment. This endpoint supports configuring prompt templates and providing message history for context.

### Method
POST

### Endpoint
$ORCH_DEPLOYMENT_URL/v2/completion

### Request Body
- **config** (object) - Required - Configuration for modules including prompt_templating.
- **placeholder_values** (object) - Optional - Values for placeholders defined in the template.
- **messages_history** (array) - Optional - A list of previous messages to maintain conversation context.

### Request Example
{
  "config": {
    "modules": {
      "prompt_templating": {
        "prompt": {
          "template": [
            {
              "role": "user",
              "content": "And now two more that sound like Goethe."
            }
          ]
        },
        "model": {
          "name": "gpt-4o-mini",
          "params": {
            "max_completion_tokens": 300,
            "temperature": 0.1,
            "frequency_penalty": 0,
            "presence_penalty": 0
          }
        }
      }
    }
  },
  "placeholder_values": {},
  "messages_history": [
    {"role": "user", "content": "Create 3 paraphrases of I love you."},
    {"role": "assistant", "content": "1. I have deep affection for you..."}
  ]
}
```

--------------------------------

### Invoke Anthropic Claude Models

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/aws-bedrock?version=CLOUD

Use the /invoke endpoint to send messages to Claude models. Requires a valid deployment URL, resource group ID, and authentication token.

```bash
curl --location '$DEPLOYMENT_URL/invoke' \
--header 'AI-Resource-Group: <Resource Group Id>' \
--header 'Content-Type: application/json' \
--header "Authorization: Bearer $AUTH_TOKEN" \
--data '{
    "anthropic_version": "bedrock-2023-05-31",
    "max_tokens": 100,
    "messages": [
        {
            "role": "user",
            "content": "Hello, Claude"
        }
    ]
}'
```

--------------------------------

### POST /chat

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/cohere?version=CLOUD

Sends a chat request to the Cohere Command A Reasoning model hosted on SAP AI Core.

```APIDOC
## POST /chat

### Description
Sends a prompt to the Cohere Command A Reasoning model to generate a response with thinking enabled.

### Method
POST

### Endpoint
$DEPLOYMENT_URL/chat

### Request Body
- **model** (string) - Required - The model identifier, set to "cohere--command-a-reasoning".
- **stream** (boolean) - Optional - Whether to stream the response.
- **frequency_penalty** (number) - Optional - Penalty for frequency.
- **thinking** (object) - Optional - Configuration for the thinking process.
- **messages** (array) - Required - List of message objects containing role and content.

### Request Example
{
    "model": "cohere--command-a-reasoning",
    "stream": false,
    "frequency_penalty": 0.8,
    "thinking": {
        "type": "enabled"
    },
    "messages": [
        {
            "role": "user",
            "content": "Tell me about the reflection?"
        }
    ]
}
```

--------------------------------

### POST /invoke-with-response-stream

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/aws-bedrock?version=CLOUD

Invokes a streaming inference request to an Amazon Bedrock model.

```APIDOC
## POST /invoke-with-response-stream

### Description
Invokes a streaming inference request to an Amazon Bedrock model where supported.

### Method
POST

### Endpoint
$DEPLOYMENT_URL/invoke-with-response-stream
```

--------------------------------

### POST /invoke (Amazon Titan)

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/aws-bedrock?version=CLOUD

Invokes Amazon Titan models for text or image embedding tasks.

```APIDOC
## POST /invoke

### Description
Invokes Amazon Titan models for text embedding or image embedding tasks.

### Method
POST

### Endpoint
$DEPLOYMENT_URL/invoke

### Parameters
#### Request Body
- **inputText** (string) - Required - The input text to process.
- **inputImage** (string) - Optional - Base64 encoded image string for image embedding tasks.
- **embeddingConfig** (object) - Optional - Configuration for embedding output, such as outputEmbeddingLength.
```

--------------------------------

### Add labels to an inference record

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/adding-to-inference-record?version=CLOUD

Use a POST request to attach key-value pair labels to a specific inference record.

```curl
curl --request POST \
  --url $AI_API/observability/inferences/{{inferenceId}}/labels \
  --header 'ai-resource-group: <resource_group>' \
  --header 'authorization: Bearer $TOKEN' \
  --header 'content-encoding: application/json' \
  --header 'content-type: application/json' \
  --data '[
  {
    "key": "ext.ai.sap.com/<key>",
    "value": "<value>"
  }'
```

--------------------------------

### POST /invoke (Anthropic Claude)

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/aws-bedrock?version=CLOUD

Invokes an Anthropic Claude model to generate text based on provided messages.

```APIDOC
## POST /invoke

### Description
Invokes an Anthropic Claude model deployment to generate text responses.

### Method
POST

### Endpoint
$DEPLOYMENT_URL/invoke

### Parameters
#### Request Body
- **anthropic_version** (string) - Required - The version of the Anthropic API (e.g., "bedrock-2023-05-31").
- **max_tokens** (integer) - Optional - The maximum number of tokens to generate.
- **messages** (array) - Required - A list of message objects containing role and content.
```

--------------------------------

### POST /observability/inferences/{{inferenceId}}/labels

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/adding-to-inference-record?version=CLOUD

Adds labels as key-value pairs to a specific inference record.

```APIDOC
## POST /observability/inferences/{{inferenceId}}/labels

### Description
Adds labels as key-value pairs to an existing inference record identified by the inferenceId.

### Method
POST

### Endpoint
$AI_API/observability/inferences/{{inferenceId}}/labels

### Parameters
#### Path Parameters
- **inferenceId** (string) - Required - The unique identifier of the inference record.

### Request Body
- **key** (string) - Required - The label key, typically prefixed with ext.ai.sap.com/.
- **value** (string) - Required - The label value.

### Request Example
[
  {
    "key": "ext.ai.sap.com/<key>",
    "value": "<value>"
  }
]

### Response
#### Success Response (200)
- **message** (string) - Confirmation message that the label has been stored.

#### Response Example
{
  "message": "inference label stored."
}
```

--------------------------------

### Label storage confirmation response

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/adding-to-inference-record?version=CLOUD

The JSON response returned upon successfully storing a label.

```json
{
  "message": "inference label stored."
}
```

--------------------------------

### Successful Response JSON Structure

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/batch-update-document-metadata?version=CLOUD

The expected JSON response format upon a successful batch update operation.

```json
{
"id": "document-uuid",
"metadata": [...]
}
```

--------------------------------

### Custom Metric Creation Response Format

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/create-custom-metric?version=CLOUD

The structure of the JSON response returned upon successful creation of a custom metric.

```json
{
"message": "<success message>"
"id": "<metric ID>" 
"scenario": "<scenario ID>"
"name": "<name>"
"version": "<version>"
}
```

--------------------------------

### Feedback storage confirmation response

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/adding-to-inference-record?version=CLOUD

The JSON response returned upon successfully storing feedback.

```json
{
  "message": "feedback stored."
}
```

--------------------------------

### Batch Update Document Metadata via cURL

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/batch-update-document-metadata?version=CLOUD

Sends a PATCH request to update document metadata. Ensure the AI-Resource-Group and Authorization headers are correctly configured.

```bash
curl --request PATCH \
  --url $AI_API_URL/v2/lm/document-grounding/pipelines/metadata/configurations/<metadataConfigId>/documents/<documentId> \
  --header 'AI-Resource-Group: <resource_group>' \
  --header 'Content-Type: application/json' \
  --header 'Authorization: Bearer $TOKEN' \
  --data '{
  "value": [
    {
      "id": "<documentId>",
      "metadata": [
        {
          "key": "contentObjectId",
          "value": [
            "12345"
          ],
          "matchMode": "ANY"
        }
      ]
    }
  ]
}'
```

--------------------------------

### Add feedback to an inference record

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/adding-to-inference-record?version=CLOUD

Use a POST request to submit feedback data for a specific inference record.

```curl
curl --request POST \
  --url $AI_API/observability/inferences/{{inferenceId}}/feedback \
  --header 'ai-resource-group: <resource_group>' \
  --header 'authorization: Bearer $TOKEN' \
  --header 'content-encoding: application/json' \
  --header 'content-type: application/json' \
  --data '[{ "content": {"stars": 5}}]'
```
Tokens:
Raw