Skip to main content

Release Notes

1.18.0 - August 26, 2025

Improvements

  • [core] Add gpt-5,gpt-5-mini and gpt-5-nano to and remove gemini-1.5-flash, gemini-1.5-pro and ibm--granite-13b-chat from the available model list. (97711bc)
  • [core] Add anthropic--claude-4-opus, anthropic--claude-4-sonnet, amazon--nova-premier, gemini-2.5-flash and gemini-2.5-pro to and remove mistralai--mixtral-8x7b-instruct-v01, meta--llama3.1-70b-instruct, nvidia--llama-3.2-nv-embedqa-1b, amazon--titan-embed-text, gpt-4, amazon--titan-text-express and amazon--titan-text-lite from the available model list. (b2629b4)

1.17.0 - July 28, 2025

New Features

  • [langchain] Support withStructuredOutput() method in OpenAI LangChain client. (8bb24c3)

Fixed Issues

  • [langchain] Use isInteropZodSchema instead of isZodSchemaV4 in LangChain Azure OpenAI and Orchestration clients to support both Zod v3 and v4 schemas. (8d54af6)

Improvements

  • [orchestration] Add support for standard and custom DPI entities in buildDpiMaskingProvider() function. Allow configuration of masking strategies via replacement_strategy for both entity types. (fab35bc)
  • [orchestration] Add utility functions getContent(), getRefusal(), getAllMessages(), getAssistantMessage(), and getResponse() to stream response. (225f40c)

1.16.0 - July 10, 2025

Compatibility Notes

  • [document-grounding] Refactor Pipelines to GetPipelines, Pipeline to GetPipeline, PipelinePostRequest to CreatePipeline, PipelineStatus to GetPipelineStatus, and RetievalSearchResults to RetrievalSearchResults. (efffc16)
  • [langchain] Adjust content of chat result in LangChain Orchestration by removing finish_reason, index, function_call, request_id from additional_kwargs and adding tool_calls and request_id in generationInfo. (e9c19b4)
  • [orchestration] The type of logprobs in the response is corrected as a result of a bug fix from the Orchestration service. (9a0d6f7)
  • [orchestration] The ModuleResults and LlmModuleResult types are now only meant to be used for the non-streaming case. Previously, they represented both streaming and non-streaming response. For streaming, use ModuleResultsStreaming and LlmModuleResultStreaming types instead.
    • @sap-ai-sdk/ai-api@1.16.0
    • @sap-ai-sdk/core@1.16.0
    • @sap-ai-sdk/prompt-registry@1.16.0 (9a0d6f7)

New Features

  • [document-grounding] Support the new pipeline trigger API and the enhanced pipeline status API. (efffc16)
  • [foundation-models] Support Azure OpenAI streaming with tool calls. (e9c19b4)
  • [langchain] Support LangChain OpenAI streaming. (e9c19b4)

Fixed Issues

  • [langchain] Retry on timeout for non-streaming requests in LangChain Orchestration client. (268c38e)
  • [orchestration] The role property of the ResponseChatMessage type is now mandatory as it is always returned. (9a0d6f7)

1.15.0 - June 13, 2025

New Features

  • [langchain] Support bindTools() method in Orchestration LangChain client. (4474313)

Improvements

  • [core] Add o3, gpt-4.1, gpt-4.1-mini, gpt-4.1-nano, o4-mini and mistralai--mistral-small-instruct to available model list. (5307dd0)

1.14.0 - May 28, 2025

Compatibility Notes

  • [orchestration] Since ChatMessage type is now one of the many specific types such as SystemChatMessage and UserChatMessage with predefined roles in the orchestration service specification, always define type when creating objects of type OrchestrationModuleConfig and TemplatingModuleConfig to avoid role being any string. (84cfa07)

New Features

  • [document-grounding] Support document grounding with SharePoint and S3 data repositories.
    • @sap-ai-sdk/core@1.14.0 (e30f66f)
  • [langchain, orchestration] Support streaming in LangChain orchestration client. (7574dd1)
  • [langchain] Support bindTools() method in Azure OpenAI LangChain client. (30939f2)
  • [orchestration] Introduce convenience functions getAllMessages(), getAssistantMessage(), getToolCalls() and getRefusal() for orchestration. (77efa15)
  • [orchestration] Introduce support for tool calls, for both streaming and regular invocation. (9931f98)
  • [orchestration] Support document grounding with SharePoint and S3 data repositories.
    • @sap-ai-sdk/ai-api@1.14.0
    • @sap-ai-sdk/core@1.14.0
    • @sap-ai-sdk/prompt-registry@1.14.0 (e30f66f)

Improvements

  • [orchestration] Make templating and templating.template properties optional in the Prompt type and introduce a new messages property to allow sending messages directly without requiring a template. This enables users to reuse the same client instance without re-initialization by passing updated messages at runtime. (7ca8d88)

1.13.0 - May 05, 2025

Fixed Issues

  • [foundation-models] Remove incorrect error logging if finish reason is null in the streaming chunk. (8130838)

Improvements

  • [core] Add gemini-2.0-flash, gemini-2.0-flash-lite, anthropic--claude-3.7-sonnet and deepseek-ai--deepseek-r1 to and remove gemini-1.0-pro from the available model list. (a6ba3af)

1.12.0 - Apr 17, 2025

Compatibility Notes

  • [langchain] Use LangchainOrchestrationModuleConfig to type the configuration object passed to the LangChain orchestration client. (5360c8c)

New Features

  • [orchestration] Add support for passing a YAML specification as a string to the templating property in OrchestrationModuleConfig. (5360c8c)

1.11.0 - Apr 8, 2025

Compatibility Notes

  • [core] Removed deprecated models:
    • text-embedding-ada-002, use text-embedding-3-small or text-embedding-3-large instead.
    • meta--llama3-70b-instruct, use meta--llama3.1-70b-instruct instead.
    • gpt-35-turbo, use gpt-4o-mini instead.
    • gpt-35-turbo-16k, use gpt-4o-mini instead. - gpt-4-32k, use gpt-4o instead. (627a152)

New Functionality

  • [langchain] Support using data masking in LangChain orchestration client. (3f8bd81)
  • [langchain] Support using content filtering in LangChain orchestration client. (e1f2b5a)
  • [orchestration] Introduce buildDpiMaskingProvider() convenience function to build masking provider SAP Data Privacy Integration. (a8d05b0)
  • [orchestration] Support setting metadata_params property in buildDocumentGroundingConfig() convenience function. (264c5b0)

1.10.0 - Mar 19, 2025

New Functionality

  • [prompt-registry] Add a new package @sap-ai-sdk/prompt-registry for consuming APIs from prompt registry service. (01ffe58)

Improvements

  • [core] Add o1 , o3-mini and alephalpha-pharia-1-7b-control to the available model list. (6f28f47)

1.9.0 - Mar 6, 2025

Compatibility Notes

  • [core] Due to the introduction of ErrorWithCause , AxiosError is now wrapped inside the cause property. For example, use error.cause.response.data instead of error.response.data to access the error response from the server. (bfed500)

New Functionality

  • [langchain] Add LangChain Orchestration client. (a45dc06)
  • [orchestration] Introduce buildLlamaGuardFilter() convenience function to build Llama guard filters. (bc51f59)

Improvements

  • [orchestration] Update Orchestration to 2502b Release (84175fb)

Fixed Issues

  • [ai-api, core] Add missing cause in the error object for failing HTTP requests by using ErrorWithCause , providing more context for debugging.

1.8.0 - Feb 14, 2025

Compatibility Notes

  • [ai-api] Update AI API spec to 2501b release. The kubesubmitV4GenericSecretsGet function is renamed to kubesubmitV4GenericSecretsGetAll . (1731104)
  • [orchestration] Deprecate buildAzureContentFilter() function. Use buildAzureContentSafetyFilter() function instead. (54a9044)
  • [orchestration] Update Orchestration to 2502a Release (2c76a1f)

New Functionality

  • [orchestration] Support configuring data masking for grounding inputs. (16d97ab)
  • [orchestration] Support using prompt registry with orchestration client. (5c248a1)

Fixed Issues

  • [ai-api, foundation-models, orchestration] Consider destination when resolving deployment ids. (09b0d2d)
  • [langchain] Add tool_calls array to assistant messages only when it is not empty. (cd06f2a)

1.7.0 - Jan 30, 2025

Compatibility Notes

  • [orchestration] Update Orchestration API to the 2501a release. As part of this change, the modelsGet method of ScenarioAPI has been renamed to scenarioQueryModels . (0a62553)

New Functionality

  • [orchestration] Support using help.sap.com as data repository type in the grounding module. (f08ac2e)
  • [orchestration] Add streaming option in the OrchestrationModuleConfig type to set global streaming options. (bdaae7e)

Fixed Issues

  • [langchain] Fix the internal mapping of LangChain to Azure OpenAI and vice versa. (ccfa2eb)
  • [orchestration] Fix logging bug when using streaming with an Orchestration JSON client. (bdaae7e)

1.6.0 - Jan 14, 2025

New Functionality

  • [orchestration] Support image recognition for orchestration service. (1da2caa)
  • [orchestration] Add support for streaming in the orchestration client. (d836abf)

Improvements

  • [orchestration] Make model_params property in the LlmModuleConfig optional and refine the type definition to also include known properties. (1476584)
  • [orchestration] Add buildDocumentGroundingConfig() convenience function to create document grounding configuration in the Orchestration client. (a039890)

1.5.0 - Jan 7, 2025

New Functionality

  • [ai-api, core, foundation-models, langchain, orchestration] Add support for providing custom destination for AI Core besides using environment variable and service binding. (b4a5506)
  • [document-grounding] Add a new package @sap-ai-sdk/document-grounding for consuming vector, pipeline and retrieval APIs from document grounding service. This package is still in beta and is subject to breaking changes. Use it with caution. (4425bfd)
  • [orchestration] Add support for using a JSON configuration obtained from AI Launchpad to consume orchestration service. (17a1eea)

1.4.0 - Dec 10, 2024

New Functionality

  • [foundation-models] Add support for streaming chat completion in the Azure OpenAI client. (d79cee8)
  • [orchestration] Add support for grounding capabilities of the orchestration service. (d79cee8)

1.3.0 - Nov 15, 2024

Compatibility Notes

  • [orchestration] Switch some of the orchestration interfaces to types, as they were introduced by accident.
  • [orchestration] Remove grounding key from the type ModuleResults . (6fe32b8)
  • [orchestration] FilteringConfig type has been renamed to InputFilteringConfig and OutputFilteringConfig for future extensibility. (bf17e17)

Fixed Issues

  • [foundation-models, orchestration] Get choice via index by comparing the index property instead of using array index. @sap-ai-sdk/ai-api@1.3.0 @sap-ai-sdk/core@1.3.0 (91df549)
  • [orchestration] Fix input and output filters to disallow additional properties as part of the filter config.

1.2.0 - Oct 25, 2024

Compatibility Notes

  • [foundation-models] Adjust AzureOpenAiEmbeddingOutput type to include multiple embedding responses as opposed to one. (99498cd)

New Functionality

  • [foundation-models] Add convenience method to access all embeddings in an Azure OpenAI response ( AzureOpenAiEmbeddingResponse ). (99498cd)

Fixed Issues

  • [langchain] Fix performance issues when creating embeddings for split documents by sending all documents in one request instead of splitting it up in separate requests. (99498cd)

1.1.0 - Oct 4, 2024

Compatibility Notes

  • [ai-api] Move modelsGet() from ModelApi to ScenarioApi , and remove ModelAPI . (5bd2e4d)

New Functionality

  • [orchestration] Add support for the data masking capabilities of the orchestration service. (a1105d9)

Improvements

  • [ai-api] Add kubesubmitV4ResourceQuotaGetDeploymentQuota() function in DeploymentApi to get details about quota and usage for deployments. Additionally introduce two new types BckndDeploymentQuotaItem and BckndDeploymentResourceQuotaResponse . (5bd2e4d)

Fixed Issues

  • [ai-api] Fix AiExecutionModificationResponseList type to correctly represent an array of responses or errors. (5bd2e4d)
  • [ai-api, core, foundation-models, langchain, orchestration] Fix sending the correct resource group headers when custom resource group is set. (771f986)
  • [ai-api, core, foundation-models, langchain, orchestration] Fix missing and unused dependencies. (506a1e4)
  • [foundation-models] Fix index-based data access in embedding response. Previously, the 0th index data was always returned. (3cbfdc7)
  • [langchain] Fix auto completion for Azure OpenAI Embedding models. (5a977da)
  • [langchain] Fix LangChain types for proper IDE auto completion. (8cda2de)

1.0.0 - Sep 19, 2024

New Functionality

  • [ai-api] Provide an HTTP wrapper to consume the AI Core Rest APIs. It is powered by the SAP Cloud SDK OpenAPI generator. (ba9133b)
  • [foundation-models] Offer an OpenAI client to consume Azure OpenAI models for chat completion and embeddings via generative AI hub of SAP AI Core. (ba9133b)
  • [langchain] Support using the LangChain framework for consuming Azure OpenAI models from generative AI hub of SAP AI Core. (ba9133b)
  • [orchestration] Introduce orchestration client for consuming the following features of the orchestration service:
    • harmonized LLM access via orchestration
    • client side prompt templates
    • content filtering (ba9133b)