|
8 | 8 |
|
9 | 9 | ### 🔧 Compatibility Notes |
10 | 10 |
|
11 | | -- The old OpenAI client (v1.0.0) is being deprecated in favor of the new OpenAI client (v1.4.0). |
12 | | - [See the documentation for more details](https://sap.github.io/ai-sdk/docs/java/foundation-models/openai/chat-completion) |
13 | | -- Generated classes for the following service specifications are subject to change: |
14 | | - - core |
15 | | - - openai |
16 | | - - orchestration |
17 | | - - document grounding |
18 | | - |
19 | | -- [Orchestration] Interfaces with only one implementation were reduced. |
20 | | - - As a result, the accessors for fields `OrchestrationModuleConfig.inputTranslationConfig` and `OrchestrationModuleConfig.outputTranslationConfig` now handle the implementing class explicitly. |
21 | | - - The same applies to helper methods `DpiMasking#createConfig()` and `MaskingProvider#createConfig()`. |
22 | | -- [Orchestration] `OrchestrationTemplate.withTemplate()` has been deprecated. Please use `OrchestrationTemplate.withTemplateMessages()` instead. |
23 | | -- [Orchestration] The method `createConfig()` is removed from `ContentFilter`, `AzureContentFilter` and `LlamaGuardFilter` and is replaced by `createInputFilterConfig()` and `createOutputFilterConfig()`. |
24 | | -- [Orchestration] Deprecated : `LLAMA3_1_70B_INSTRUCT`, `CLAUDE_3_SONNET`, `TITAN_TEXT_LITE`, `TITAN_TEXT_EXPRESS`, `GPT_4`, `GPT_4_0613`, `MIXTRAL_8X7B_INSTRUCT_V01`. |
25 | | - - `GPT_4` and `GPT_4_0613` are replaced by : `GPT_40`or `GPT_41`. |
26 | | - - `CLAUDE_3_SONNET` is replaced by `CLAUDE_4_SONNET`. |
27 | | - - `MIXTRAL_8X7B_INSTRUCT_V01` is replaced by `MISTRAL_SMALL_INSTRUCT`. |
28 | | -- [OpenAI] Deprecated : `GPT_4`. |
29 | | - - `GPT_4`is replaced by : `GPT_40`or `GPT_41`. |
30 | | - |
31 | | -- [Prompt Registry] Resource group has been added as a optional parameter to all endpoints. Set it to `"default"` if it was not set before. Examples: |
32 | | - - `client.importPromptTemplate(File)` --> `client.importPromptTemplate("default", File)`. |
33 | | - - `client.parsePromptTemplateById(id, false, inputParams)` --> `client.parsePromptTemplateById(id, "default", false, inputParams)`. |
34 | | - |
35 | | -- [Document Grounding] All classes with `Retrieval` have been renamed to fix the typo |
36 | | - - for example: `RetievalSearchResults` has been renamed to `RetrievalSearchResults` |
37 | | -- [Document Grounding] `PipelinesApi#getAllPipelines()` and `PipelinesApi#getPipelineById()` now any of these 3 classes implementing the `GetPipeline` interface: |
38 | | - - `MSSharePointPipelineGetResponse`, `S3PipelineGetResponse` and `SFTPPipelineGetResponse` |
| 11 | +- |
39 | 12 |
|
40 | 13 | ### ✨ New Functionality |
41 | 14 |
|
42 | | -- [Orchestration] Added support for [transforming a JSON output into an entity](https://sap.github.io/ai-sdk/docs/java/orchestration/chat-completion#json_schema) |
43 | | -- [Orchestration] Added `AzureContentFilter#promptShield()` available for input filtering. |
44 | | -- [Orchestration] Added new models for `OrchestrationAiModel`: `GEMINI_2_5_FLASH`, `GEMINI_2_5_PRO`, `ALEPHALPHA_PHARIA_1_7B_CONTROL`, `OPENAI_O4_MINI`, `CLAUDE_4_OPUS`, `CLAUDE_4_SONNET`, `OPENAI_O3`. |
| 15 | +- |
45 | 16 |
|
46 | 17 | ### 📈 Improvements |
47 | 18 |
|
48 | 19 | - |
49 | 20 |
|
50 | 21 | ### 🐛 Fixed Issues |
51 | 22 |
|
52 | | -- [Orchestration] Resolved duplicate JSON property issue, enabling Anthropic Claude chat completions. |
53 | | -- Remove logging of any request/response payloads to avoid potential exposure of sensitive data. |
| 23 | +- |
0 commit comments