### Metadata Handling Example

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

An example demonstrating how to handle metadata, specifically showing a naming convention for duplicate metadata keys across different levels (chunk and document).

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

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

### Output Code Example

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

This JSON output shows an example of the response when metadata like source and webUrl is included.

```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"
                ]
            }
        }
    ]
]
```

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

### Prompt Templating Example

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/contextualized-retrieval-using-metadata-and-vector-search

This snippet shows an example of prompt templating, including message history and defaults, for a helpful assistant that uses grounding requests and outputs.

```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": {}
}
```

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

### User Role Content Example

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

An example of JSON content defining a user role for an assistant, including a prompt structure for SAP Teched 2024 queries.

```json
{
  "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."
 }
```

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

### Example Output of Available Models

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/choose-model

This is an example of the JSON output returned when querying the available models endpoint, showing details for GPT-4o.

```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
        }, ...

    ]
}
```

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

### Get the Deployment URL

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/create-deployment-for-generative-ai-model

Retrieves the details of a deployment by sending a GET request to the deployments endpoint with the deployment ID.

```curl
curl --location '$AI_API_URL/v2/lm/deployments/$DEPLOYMENT_ID' \
--header 'AI-Resource-Group: default' \
--header "Authorization: Bearer $AUTH_TOKEN"
```

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

### AI Assistant Response Example

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/chat-cd16d739cf7b4c8aa5e3b12062ae6ce9

An example of a response from an AI assistant, including conversational content and usage details.

```json
{
  "id": "chatcmpl-7v0n35364397234567890",
  "object": "chat.completion",
  "created": 1691274839,
  "model": "gpt-3.5-turbo-0613",
  "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
  }
}
```

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

### Check Scenario

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/create-deployment-for-generative-ai-model

Sends a GET request to check access to the scenario containing generative AI.

```curl
curl --location '{{apiurl}}/v2/lm/scenarios' \
--header 'AI-Resource-Group: default' \
--header "Authorization: Bearer $AUTH_TOKEN"
```

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

### Successful Response

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

Example JSON response upon successful creation of a custom metric.

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

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

### Amazon Nova

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/aws-bedrock

Example of invoking the Amazon Nova model for conversational interaction.

```curl
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"
        }
    ]
}'

```

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

### View Available Models

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/choose-model

Send a GET request to this endpoint to retrieve a list of available foundation models.

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

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

### Sample Code

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

Send a GET request to the endpoint to check the status of your batch job.

```curl
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" \

```

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

### Command A Reasoning API Call

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/cohere

Example of how to call the Command A Reasoning model via API.

```curl
curl --location "$DEPLOYMENT_URL/v2/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?" \
        } \
    ] \
}'
```

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

### Curl Request for Generative AI Completion

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/chat-cd16d739cf7b4c8aa5e3b12062ae6ce9

This example demonstrates how to make a POST request to the Generative AI completion endpoint using curl, including setting headers and providing a JSON payload with prompt templating and message history.

```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." \
      } \
    ] \
  }'
```

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

### Converse with Anthropic Claude

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/aws-bedrock

Example curl command to converse with an Anthropic Claude model through the orchestration scenario.

```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" \
        } \
    ] \
}'
```

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

### Example cURL Request for Chat Completion

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/chat

This snippet shows a cURL command to send a POST request to the completion endpoint, including message history and module configurations.

```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."
      }
    ]
  }'
```

```

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

### Invoke Anthropic Claude

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/aws-bedrock

Example curl command to invoke an Anthropic Claude model through the orchestration scenario.

```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" \
        } \
    ] \
}'
```

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

### Rerank API Output

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/cohere

Example JSON output from the Cohere Rerank API.

```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
    }
  ]
}
```

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

### Create Custom Metric Request

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

Example cURL command to send a POST request to create a custom evaluation metric.

```curl
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 \
}'
```

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

### Check Deployment Status

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/create-deployment-for-orchestration

This cURL command sends a GET request to check the status of the orchestration deployment using its ID.

```shell
curl --request GET "$AI_API_URL/v2/lm/deployments/$ORCH_DEPLOYMENT_ID" \
    --header "Authorization: Bearer $TOKEN" \
    --header "AI-Resource-Group: $RESOURCE_GROUP"
```

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

### Titan Embed Image

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/aws-bedrock

Example of invoking the Titan Embed Image model to generate embeddings from an image.

```curl
curl --location '$DEPLOYMENT_URL/invoke' \
--header 'AI-Resource-Group: <Resource Group Id>' \
--header 'Content-Type: application/json' \
--header "Authorization: Bearer $AUTH_TOKEN" \
--data '{ 
    "inputText": "hi",
    "inputImage": "#this-is-an-example-base64-string-value for-an-image/9j/4QDeRXhpZgAASUkqAAgAAAAGABIBAwABAAAAAQAAABoBBQABAAAAVgAAABsBBQABAAAAXgAAACgBAwABAAAAAgAAABMCAwABAAAAAQAAAGmHBAABAAAAZgAAAAAAAABIAAAAAQAAAEgAAAABAAAABwAAkAcABAAAADAyMTABkQcABAAAAAECAwCGkgcAFgAAAMAAAAAAoAcABAAAADAxMDABoAMAAQAAAP//AAACoAQAAQAAAMgAAAADoAQAAQAAAMgAAAAAAAAAQVNDSUkAAABQaWNzdW0gSUQ6IDEyM//bAEMACAYGBwYFCAcHBwkJCAoMFA0MCwsMGRITDxQdGh8eHRocHCAkLicgIiwjHBwoNyksMDE0NDQfJzk9ODI8LjM0Mv/bAEMBCQkJDAsMGA0NGDIhHCEyMjIyMjIyMjIyMjIyM",
    "embeddingConfig": {
        "outputEmbeddingLength": 256
    }
}'

```

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

### Example JSON Response from Chat Completion

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/chat

This snippet shows the JSON response structure from a chat completion request, including module results and orchestration results.

```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",

```

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

### Rerank 3.5 API Call

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/cohere

Example curl command to call the Cohere Rerank 3.5 model for document reranking.

```curl
curl --location "$DEPLOYMENT_URL/rerank" \
--header 'AI-Resource-Group: <Resource Group Id>' \
--header 'Content-Type: application/json' \
--header "Authorization: Bearer $AUTH_TOKEN" \
--data '{
    "model": "cohere-reranker-35",
    "top_n": 3,
    "query": "What is the capital of Amesic?",
    "documents": [
        "Carson City is the capital city of the American state of Strita.",
        "The Commonwealth of the Northern Mariana Islands is a group of islands in the Pacific Ocean. Its capital is Saipan.",
        "Anesia, D.C. is the capital of the Amesic States. It is a federal district.",
        "Capitalization or capitalisation in English grammar is the use of a capital letter at the start of a word. English usage varies from capitalization in other languages.",
        "Capital punishment has existed in the Amesic since before it was a country. As of 2017, capital punishment is legal in 30 of the 50 states."
    ]
}'
```

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

### Invoke Amazon Titan Embed Text

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/aws-bedrock

Example curl command to invoke Amazon Titan Embed Text model through the orchestration scenario.

```bash
curl --location '$DEPLOYMENT_URL/invoke' \
--header 'AI-Resource-Group: <Resource Group Id>' \
--header 'Content-Type: application/json' \
--header "Authorization: Bearer $AUTH_TOKEN" \
--data '{ \
    "inputText": "Who im I?" \
  }'
```

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

### Sample Code for Creating a Configuration

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

This cURL command demonstrates how to create a configuration for optimization by sending a POST request to the AI API.

```curl
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": "dataset", \
            "value": "data.json" \
        { \
            "key": "optimizationMetric", \
            "value": "JSON_Match” \
        }, \

        { \
            "key": "promptTemplateScope", \
            "value": "resourcegroup" \
        }, 
        { \

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

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

### Create a Configuration

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/create-deployment-for-generative-ai-model

Sends a POST request to create a configuration for a model, including its name, executable ID, scenario ID, version ID, and parameter bindings for model name and version.

```curl
curl --location '$AI_API_URL/v2/lm/configurations' \
--header 'AI-Resource-Group: default' \
--header 'Content-Type: application/json' \
--header "Authorization: Bearer $AUTH_TOKEN" \
--data '{ 
	"name": "yourNameChoice",
	"executableId": "<executableId>",
	"scenarioId": "foundation-models",
	"versionId": "0.0.1",
	"parameterBindings": [
		{
			"key":"modelName",
			"value":"<ModelName>"
		},
		{
			"key": "modelVersion",
			"value": "<ModelVersion>"
		}
  ]
}'
```

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

### Note on using trainDataset and testDataset

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

This snippet shows how to modify the dataset key-value pairs to use trainDataset and testDataset.

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

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

### Sample Code

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/create-collection-c124fb9b68864fa187243ee4edcf213c

This sample code demonstrates how to create a new collection using a POST request to the specified endpoint.

```curl
curl --request POST \
  --url $AI_API_URL/v2/lm/document-grounding/vector/collections \
  --header 'AI-Resource-Group: {{resource_group}}' \
  --header 'Authorization: Bearer {{access_token}}'
  --data '{
  "id": "<your choice of valid uuid>", 
  "title": "<title of the collection>",
  "embeddingConfig": {
    "modelName": "<embedding-model-name>"
  },
  "metadata": [
    {
      "key": "purpose",
      "value": [
        "<text>"
      ]
    },
    {
      "key": "<a-random-key>",
      "value": [
        "<text>"
      ]
    }
  ]
}'
```

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

### Create a Deployment

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/create-deployment-for-generative-ai-model

Sends a POST request to create a deployment using the configuration ID obtained from the previous step.

```curl
curl --location '$AI_API_URL/v2/lm/deployments' \
--header 'AI-Resource-Group: default' \
--header 'Content-Type: application/json' \
--header "Authorization: Bearer $AUTH_TOKEN" \
--data '{ 
 "configurationId": "yourConfigurationId"
}'
```

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

### Leverage Metadata in Prompt

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

This JSON snippet shows how to include instructions in the prompt for how the model should use the metadata, within the prompt_templating configuration.

```json
"prompt_templating": {
  "messages_history": [

```

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

### Sample Code

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/create-batch-job

Submit a batch job by sending a POST request to the endpoint $AI_API/llm-batch-service/v1/batches.

```curl
curl -X POST "$AI_API/llm-batch-service/v1/batches" \
    -H 'AI-Resource-Group: <resource_group>' \
    -H 'Content-Type: application/json' \
    -H "Authorization: Bearer $AUTH_TOKEN" \
    -D '{ \
        "type": "llm-native", \
        "input": { \
            "uri": "ai://<object_store_secret_name>/<input_file>.jsonl" \
        }, \
        "output": { \
            "uri": "ai://<object_store_secret_name>/<output_folder>/" \
        }, \
        "spec": { \
            "provider": "azure-openai", \
            "model": "gpt-4.1" \
        } \
    }'


```

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

### Create Configuration for Orchestration Deployment

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/create-deployment-for-orchestration

This cURL command sends a POST request to create a configuration for an orchestration deployment.

```shell
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-raw  '{ 
    "name": "orchestration-config", 
    "executableId": "orchestration", 
    "scenarioId": "orchestration"
  }'
```

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

### Retrieve Supported Metadata Keys

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

This cURL command demonstrates how to retrieve supported metadata keys by querying the retrieval search endpoint.

```curl
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": [] \
        } \
    ] \
}'
```

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

### Sample Code

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

A minimal call to orchestration contains only configurations of the required templating and model configuration modules.

```curl
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" 
  } 
}'
```

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

### Create Deployment for Orchestration

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/create-deployment-for-orchestration

This cURL command sends a POST request to create a deployment for orchestration using the provided configuration ID.

```shell
curl --request POST $AI_API_URL/v2/lm/deployments \
    --header 'content-type: application/json' \
    --header "Authorization: Bearer $TOKEN" \
    --header "AI-Resource-Group: $RESOURCE_GROUP" \
    --data-raw "{
        \"configurationId\": \"$ORCH_CONFIG_ID\" 
    }"
```

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

### Set Deployment URL as Environment Variable

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/create-deployment-for-orchestration

This command sets the deployment URL as an environment variable, which is available when the deployment status is RUNNING.

```shell
ORCH_DEPLOYMENT_URL="<deployment_url>"
```

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

### Sample Code

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

This sample code demonstrates how to send a PATCH request to update document metadata.

```curl
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"
        }
      ]
    }
  ]'

```

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

### Include Metadata in Request

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/contextualized-retrieval-using-metadata-and-vector-search

This `curl` command shows how to include metadata like 'source' and 'webUrl' in the results by using `metadata_params`.

```shell
curl --request POST "$ORCH_DEPLOYMENT_URL/completion" \
--header "content-type: application/json" \
--header "Authorization: Bearer $TOKEN" \
--header "AI-Resource-Group: $RESOURCE_GROUP" \
 --data '{ \
    "orchestration_config": { \
        "module_configurations": { \
            "grounding_module_config": { \
                "type": "document_grounding_service", \
                "config": { \
                    "filters": [ \
                        { \
                            "id": "filter1", \
                            "data_repositories": [ \
                                "*" \
                            ], \
                            "search_config": {}, \
                            "data_repository_type": "vector" \
                        } \
                    ], \
                    "input_params": [ \
                        "groundingRequest" \
                    ], \
                    "output_param": "groundingOutput", \
                    "metadata_params": [ \
                        "source", \
                        "webUrl" \
                    ] \
                     \
                } \
            }, \
            "llm_module_config": { \
                "model_name": "gemini-1.5-pro", \
                "model_params": {}, \
                "model_version": "001" \
            }, \
            "templating_module_config": { \
                "template": [ \
                    { \
                        "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 }}" \
                    } \
                ], \
                "defaults": {} \
            }, \
            "filtering_module_config": { \
                "input": { \
                    "filters": [ \
                        { \
                            "type": "azure_content_safety", \
                            "config": { \
                                "Hate": 2, \
                                "SelfHarm": 2, \
                                "Sexual": 2, \
                                "Violence": 2 \
                            } \
                        } \
                    ] \
                }, \
                "output": { \
                    "filters": [ \
                        { \
                            "type": "azure_content_safety", \
                            "config": { \
                                "Hate": 2, \
                                "SelfHarm": 2, \
                                "Sexual": 2, \
                                "Violence": 2 \
                            } \
                        } \
                    ] \
                } \
            } \
        } \
    }, \
    "input_params": { \
        "groundingRequest": "what is SAP Joule?" \
    } \
}'
```

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

### Set Orchestration Configuration ID as Environment Variable

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/create-deployment-for-orchestration

This command sets the ID of the created orchestration configuration as an environment variable.

```shell
ORCH_CONFIG_ID=f7ac7f77-e70f-4e9c-86b5-1504b44fe789
```

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

### Include Metadata in Request

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

This cURL command shows how to include metadata (source and webUrl) in the results by configuring the grounding service with metadata_params.

```curl
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}}

Reports: {{?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?" \
    } \
}'
```

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

### Output Code

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/create-batch-job

The service schedules the job for processing and returns a unique batch ID.

```json
{
  "id": "550e8400-e29b-41d4-a716-446655440000",
  "created_at": "2026-01-07T10:30:00Z",
  "status": "PENDING",
  "message": "Batch job scheduled"
}


```

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

### Set Orchestration Deployment ID as Environment Variable

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/create-deployment-for-orchestration

This command sets the ID of the created orchestration deployment as an environment variable.

```shell
ORCH_DEPLOYMENT_ID=d4168482710c6cf9
```

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

### Create Document Request

Source: https://help.sap.com/docs/sap-ai-core/generative-ai/create-document

This cURL command demonstrates how to send a POST request to the endpoint to add a document to a collection, including metadata and chunks.

```curl
curl --request POST \ 
 --url  $AI_API_URL/v2/lm/document-grounding/vector/collections/{{collectionId}}/documents\    
 --header 'AI-Resource-Group: {{resource_group}}' \    
 --header 'Content-Type: application/json' \ 
 --header 'Authorization: Bearer {{access_token}}'\
  --data '{ 
  "documents": [ 
    { 
      "metadata": [
        {
          "key": "url", 
          "value": [ 
            "http://hello.com", 
            "123" 
          ] 
        } 
      ], 
      "chunks": [
        {
          "content": "<chunk content 1>",
          "metadata": [  // 
            {
              "key": "index", 
              "value": [ 
                "1" 
              ] 
            }
          ] 
        }, 
        {
          "content": "<chunk content 2>", 
          "metadata": [
            {
              "key": "index", 
              "value": [ 
                "2" 
              ] 
            }
          ] 
        }
      ] 
    } 
  ]
}' 
```

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

### Output Code

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

A successful response returns code 200 and includes details of the document in JSON format.

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

```

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

### Adding Labels

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

Send a POST request to the endpoint to add labels to an inference record. Include the inference ID and labels as key-value pairs in the body.

```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>"
  }
]'
```

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

### Output Code

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

The response includes the current status of your batch.

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

```