v2.7.0

Compatibility Notes

• [ai-api] Extract functionality around execution scheduling from the ExecutionApi to ExecutionScheduleApi. (8616d5e)
• [ai-api] Remove the BckndEvent type. (8616d5e)
• [ai-api] Split the BckndArgoCDApplicationData into BckndArgoCDApplicationDataResponse and BckndArgoCDApplicationDataRequest. (8616d5e)
• [core] Remove deprecated and retired models from model list.
  Remove retired model anthropic--claude-3-sonnet from model list, use suggested replacement anthropic--claude-4.5-sonnet instead.
  Remove retired model anthropic--claude-3-opus from model list.
  Remove deprecated model gpt-4o-mini, use suggested replacement gpt-5-mini instead.
  Remove deprecated models anthropic--claude-3.5-sonnet and anthropic--claude-3.7-sonnet. (b12626b)
• [orchestration] Support for constructing an OrchestrationStreamResponse without an HttpResponse has been deprecated, and will be removed in the next major release.
  Code directly instantiating this class should be updated to provide an HttpResponse object as the first parameter to allow reading from raw HTTP response. (6b49479)

New Features

• [core] Advertise AbortSignal support for HTTP request cancellation.
  This change adds typings, documentation and examples for using AbortSignal with the HTTP client to enable request cancellation. (b11b00c)
• [langchain] Support the withStructuredOutput() method in the Orchestration LangChain client. (4fce347)
• [orchestration] Add rawResponse property to OrchestrationStreamResponse.
  The new property exposes the raw HTTP response from the orchestration service for advanced use cases. (6b49479)
• [orchestration] Add getRequestId() method to OrchestrationResponse, OrchestrationStreamResponse and OrchestrationEmbeddingResponse.
  The new method allows retrieving the request ID from the orchestration service responses, which can be useful for debugging and tracking requests. (6b49479)

Improvements

• [core] Combine 'ai-client-type' headers if a custom 'ai-client-type' header is set. (56e9c3f)
• [rpt] Restrict the task_type property in the PredictionConfig type. (12b4129)

v2.6.0

New Features

• [rpt] Release Beta version of a client for the SAP-RPT-1 model. (790ad05)

Fixed Issues

• [orchestration] Export isConfigReference() method as a value instead of type-only export.
  This function can now be imported and used at runtime to check if a configuration is an orchestration configuration reference type. (f3b6dc5)

v2.5.0

Compatibility Notes

• [langchain] The @langchain/core package is now a peer dependency.
  You now have to install the @langchain/core package as a direct dependency of your project. (5b88f6f)

New Features

• [langchain] Support disabling streaming completely via the langchain option disableStreaming. (b91e0a7)
• [langchain] Support auto-streaming via the langchain option streaming. When enabled (e.g., transparently by LangGraph), responses are automatically streamed in invoke() calls. (b91e0a7)
• [orchestration] Add support for orchestration configuration references in the orchestration client. (2f19a40)

Fixed Issues

• [langchain] Pin @langchain/core to v1.1.8 to avoid a regression. (9ff7cec)

Improvements

• [langchain] Move the @langchain/core package from dependencies to peer dependencies.
  This reduces the chance of version incompatibilities between langchain-related packages. (5b88f6f)

v2.4.0

Compatibility Notes

• [prompt-registry] In the prompt-registry client schema, the Template type was renamed to PromptTemplate. (a3cbc6e)

New Features

• [orchestration] Added support for applyTo and translateMessagesHistory in order to enable selective input translation and automatic inference of target language for output translation parameters. (aa097da)
• [prompt-registry] Update prompt-registry specification adding support for resource group scoped prompt templates. (a3cbc6e)

Improvements

• [core] Added anthropic--claude-4.5-sonnet , anthropic--claude-4.5-haiku , gemini-2.5-flash-lite and sap-abap-1 to the available model list.
  Removed deprecated models gemini-2.0-flash and gemini-2.0-flash-lite scheduled for retirement. (2e1d2c2)

v2.3.0

Compatibility Notes

• [langchain, orchestration, prompt-registry] zod was upgraded to v4 (7c9605d)

New Features

• [langchain] Bump langchain to v1 (7c9605d)
• [orchestration] Added support for protected_material_code property to buildAzureContentSafetyFilter() function for output filter configuration to allow detecting protected code content from known github repositories. (485e21b)

v2.2.0

Compatibility Notes

• [document-grounding] MSSharePointConfigurationGetResponse now requires the sharePoint property (e2c34f3)
• [document-grounding] CommonConfiguration was replaced with backend-specific types: SFTPConfiguration and S3Configuration (e2c34f3)
• [document-grounding] Some types have been renamed to include endpoint-specific prefixes.
  Some instances of the prior names may still exist:
  • SearchResults was renamed to VectorSearchResults / RetrievalSearchResults
  • Chunk was renamed to VectorChunk / RetrievalChunk
  • SearchFilter was renamed to VectorSearchFilter
  • KeyValueListPair was renamed to VectorKeyValueListPair / RetrievalKeyValueListPair
  • DocumentKeyValueListPair was renamed to VectorDocumentKeyValueListPair / RetrievalDocumentKeyValueListPair
  • SearchConfiguration was renamed to VectorSearchConfiguration / RetrievalSearchConfiguration
  • SearchSelectOptionEnum was renamed to VectorSearchSelectOptionEnum / RetrievalSearchSelectOptionEnum
  • PerFilterSearchResult was renamed to RetrievalPerFilterSearchResult
  • PerFilterSearchError was renamed to RetrievalPerFilterSearchError
  • DataRepositorySearchResult was renamed to RetrievalDataRepositorySearchResult
  • SearchInput was renamed to RetrievalSearchInput (e2c34f3)
• [langchain] Remove structured ouput handling for deprecated gpt-4 & gpt-3 models (6100bca)

New Features

• [ai-api] Update ai-api package with the new specification (2509b). (58464e9)
• [core, orchestration] Introduce orchestration embedding client for consuming embedding feature of the orchestration service. (347eac1)
• [document-grounding] Update document-grounding specification (e2c34f3)

Fixed Issues

• [core] Replace active logging during streaming with error throwing to avoid logging the response payload. (5225275)

Improvements

• [core, orchestration] Add cohere--command-a-reasoning, mistralai--mistral-medium-instruct and perplexity-ai sonar and sonar-pro to model list (6100bca)

v2.1.0

Fixed Issues

• [orchestration] Fix JSDoc example of buildAzureContentSafetyFilter() function. (0cf7d80)

Improvements

• [core] Remove alephalpha-pharia-1-7b-control and deepseek-ai--deepseek-r1 from available model list. (0cf7d80)

v2.0.0

Compatibility Notes

• [foundation-models, orchestration] Change stream method parameter from AbortController to AbortSignal.
  The stream() method now accepts an AbortSignal instead of an AbortController as the second parameter in both Azure OpenAI and Orchestration clients. (4c00c27)
• [foundation-models, orchestration] Response object data property is renamed to _data.
  Use getter methods like getContent(), getTokenUsage(), getAssistantMessage() instead of direct data access. (5c52cb6)
• [foundation-models] Move generated types to internal exports while keeping frequently used types in main exports.
  • Generated types are no longer exported from @sap-ai-sdk/foundation-models and must be imported from @sap-ai-sdk/foundation-models/internal.js instead.
  • Frequently used types (AzureOpenAiChatCompletionTool, AzureOpenAiFunctionObject, AzureOpenAiChatCompletionRequestMessage, AzureOpenAiChatCompletionRequestSystemMessage, AzureOpenAiChatCompletionRequestUserMessage, AzureOpenAiChatCompletionRequestAssistantMessage, AzureOpenAiChatCompletionRequestToolMessage) remain available from main package exports.
  • Add new type AzureOpenAiChatCompletionParameters to replace AzureOpenAiCreateChatCompletionRequest which is no longer exported publicly. (5c52cb6)
• [langchain] Major breaking changes for LangChain orchestration v2:
  • Update LangChain orchestration configuration structure to use promptTemplating instead of separate llm and templating properties.
  • Replace llm.model_name with promptTemplating.model.name and llm.model_params with promptTemplating.model.params.
  • The templating.template property is now promptTemplating.prompt.template.
  • Rename inputParams parameter to placeholderValues in LangChain orchestration client methods.
  • Update message response property names from module_results to intermediate_results in additional kwargs. (86e6370)
• [langchain] Update imports to use new API facade from foundation-models package.
  • Some generated types now need to be imported from @sap-ai-sdk/foundation-models/internal.js instead of @sap-ai-sdk/foundation-models.
  • Update to use new AzureOpenAiChatCompletionParameters type to replace AzureOpenAiCreateChatCompletionRequest which is no longer exported publicly. (5c52cb6)
• [orchestration] buildTranslationConfig() function now requires type parameter to distinguish between input and output translation configuration. (740ba78)
• [orchestration] buildLlamaGuardFilter() function has been renamed to buildLlamaGuard38BFilter(). It now requires a type parameter to distinguish between input and output filter configurations, and accepts filter categories as an array. (740ba78)
• [orchestration] Move generated types to internal exports while keeping frequently used types in main exports.
  • Generated types are no longer exported from @sap-ai-sdk/orchestration and must be imported from @sap-ai-sdk/orchestration/internal.js instead.
  • Frequently used types (ChatMessage, SystemChatMessage, UserChatMessage, AssistantChatMessage, ToolChatMessage, DeveloperChatMessage, ChatCompletionTool, FunctionObject) remain available from main package exports. (5c52cb6)
• [orchestration] buildAzureContentSafetyFilter() function now requires type parameter to distinguish between input and output filter configuration. (997e8ec)
• [orchestration] Major breaking changes for orchestration v2:
  • Consolidate llm and templating modules into a single promptTemplating module.
  • The llm.model_name property is now promptTemplating.model.name and llm.model_params is now promptTemplating.model.params.
  • The templating.template property is now promptTemplating.prompt.template.
  • Rename inputParams parameter to placeholderValues in orchestration client methods.
  • Update response property names from orchestration_result to final_result and module_results to intermediate_results.
  • Replace top-level stream property with streamOptions.enabled and update streaming module options from llm to promptTemplating.
  • Update grounding configuration to use placeholders.input and placeholders.output instead of separate input_params and output_param.
  • Update Azure content filter property names to lowercase with underscores: Hate to hate, SelfHarm to self_harm, Sexual to sexual, and Violence to violence.
  • Remove deprecated buildAzureContentFilter() function and use buildAzureContentSafetyFilter() instead. (86e6370)

New Features

• [ai-api] Add resolveDeploymentUrl() function to resolve the deployment URL that matches the given criteria. (14745de)
• [foundation-models] Add getTokenUsage(), getFinishReason(), getContent(), getToolCalls(), getRefusal(), getAssistantMessage(), findChoiceByIndex() methods to Azure OpenAI chat completion response. (5c52cb6)
• [orchestration] Add prompt_shield property to buildAzureContentSafetyFilter() function for input filter configuration to allow enabling prompt attack detection. (997e8ec)
• [orchestration] Add getIntermediateResults() method to OrchestrationResponse, OrchestrationStreamResponse, OrchestrationStreamChunkResponse classes for accessing intermediate processing results from orchestration modules. (5c52cb6)
• [orchestration] Add deploymentId as the optional parameter for OrchestrationClient initialization. (0a418d0)
• [orchestration] Add findChoiceByIndex() method to find specific choices by index in streaming responses. (5c52cb6)

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. (500c0dd)
• [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. (9e1c43a)

v1.18.0

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)

v1.17.0

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)