[OpenAI] [Assistants] Assistant V2 support v2024-05-01-preview rele…

…ase (#40389) * Updated commit hash to latest assistantv2 commit * Using commit without the openAPI v3 annotations * Project compiles * Code regen after tooling update and commit cherry picking * Corrections to the readme from codegen * Code regen * Code regen * Modified return types for paginated endpoints * Migrated paged endpoints to use PageableList<> async * Added test skeleton class for vectorStore * Updated commit hash * WIP: introduced intermediate type for custom serialization logic * WIP: trying to add tests, running into issues * Removed AssistantFile related class * Fixed (de)serialization for assistantCreation response_format * Added docs for getters * Added customization for createRunOptions * Added tests for CreateAndRunThreadOptions * Added tests for updateAssistantOption * Added tests for getter methods * Added assertion for getters too * Removed some left over @generated * Code regen * Added missing responseFormat handling for ThreadRun * Added tests for tool_choice * Addd tests for createAndRunThreadOptions * Added tests createRunOptions * Removed imports * Added tests for null input JSON * code regen * Adjusted header value to use v2 for openAI * Updated test recordings * Removing recordings for removed tests * using a commit hash from main * Ajusted method * Re-recorded as much as possible. Adjusted some assertions. Marked tests that need more work * WIP: sanitizers behaivour * Renamed/removed retrieval related classes to file_search * WIP: deserialization is failing * Fle search test passing * Updated tests for file_search * Addressed 404 * renamed sample * Code regen * Various test improvements * WIP: style checks * Style checks * More style checks * [OpenAI-Assistant] Added vector store feature (#40467) * use the main repo's SHA * fixed the typo * Trying to make the CI happy * Second try * Reverting changes * WIP: sorting added features and breaking changes * Last details in the CHANGELOG * Typos and casing issues * CHANGELOG feedback * fixed the sanitizer CI failures * Added a few more notes in CHANGELOG.md --------- Co-authored-by: Shawn Fang <45607042+mssfang@users.noreply.github.com> Co-authored-by: Shawn Fang <shafang@microsoft.com>
Azure · Jun 6, 2024 · 85c1836 · 85c1836
1 parent d674de5
commit 85c1836
Show file tree

Hide file tree

Showing 133 changed files with 14,712 additions and 5,540 deletions.
diff --git a/sdk/openai/azure-ai-openai-assistants/CHANGELOG.md b/sdk/openai/azure-ai-openai-assistants/CHANGELOG.md
@@ -1,20 +1,130 @@
 # Release History
 
-## 1.0.0-beta.3 (Unreleased)
+## 1.0.0-beta.3 (2024-06-07)
 
 ### Features Added
 
-- Added `createRunStream` and `createThreadAndRunStream` methods to `AssistantsClient` and `AssistantsAsyncClient` classes. A suite of classes extending from `StreamUpdate` were added for users to be able to consume the incremental updates from the service
+#### Streaming
+
+- Operation updates:
+  - Added `createRunStream`, `createThreadAndRunStream` and `submitToolOutputsToRunStream` methods to `AssistantsClient` and `AssistantsAsyncClient` classes.
+    A suite of classes extending from `StreamUpdate` were added for users to be able to consume the incremental updates from the service.
+
+#### Assistants
+
+- Model updates:
+  - `Assistant`, `AssistantCreationOptions` and `UpdateAssistantOptions` models:
+    - Added new fields: `toolResources` , `temperature`, `topP` and `responseFormat`.
+
+#### Files
+
+- Model updates:
+  - `OpenAIFile` model changes:
+      - Added new fields: `status` of `FileState` type and `status_details` of `String` type.
+  - Added new enum `FileState` representing the type of the `status` field mentioned in the previous point.
+  - Added new possible values for `FilePurpose`: `batch`, `batch_output` and `vision`.
+
+#### Messages
+
+- Operation updates: 
+  - Updated `listMessages` to accept the filter `runId`.
+
+- Model updates:
+  - Added new model `MessageAttachment`.
+  - Updated docs renaming mentions of `retrieval` tool to `file_search`.
+  - Added new field, `startIndex` and `endIndex` to `MessageTextFileCitationAnnotation`, `MessageTextFilePathAnnotation` model.
+#### Run Step
+
+- Model updates:
+  - `ThreadRun` model updates:
+    - Added fields: `temperature`, `topP`, `maxPromptTokens`, `maxCompletionTokens`, `truncationStrategy`, `toolChoice` and `responseFormat`.
+  - Updated documentation for `RunCompletionUsage`.
+  - `CreateRunOptions` model updates:
+    - Added fields: `additionalMessages`, `temperature`, `topP`, `maxPromptTokens`, `maxCompletionTokens`, `truncationStrategy`, `toolChoice` and `responseFormat`.
+  - `CreateAndRunThreadOptions` model updates:
+    - Added fields: `toolResources`, `temperature`, `topP`, `maxPromptTokens`, `maxCompletionTokens`, `truncationStrategy`, `toolChoice` and `responseFormat`.
+  - Added new model for all the `truncationStrategy` fields called `TruncationObject`.
+
+#### Threads
+
+- Model updates:
+  - `AssistantThread` model now includes `toolResources` field as nullable.
+  - `AssistantThreadCreationOptions` updates include: `messages` type using renamed type `ThreadInitializationMessage` -> `ThreadMessageOptions`, `toolResources`.
+
+#### Tool Resources (new)
+
+- Model updates:
+  - There are 3 new models that were added: `ToolResources`, `CreateToolResourcesOptions` and `UpdateToolResourcesOptions`. As the name implies, wherever there is 
+    a field `toolResource` we use the appropriate type, depending on the model declaring it is meant to be a response object, create request object or update request object, respectively.
+
+#### Vector Stores (new)
+
+There are 3 main areas for vector stores into which its models and operations can be divided. That is vector stores themselves, vector store files and vector store file batches.
+
+- Model updates:
+  - new models (I will just list the top level response and request objects, but there are several subtypes describing more complex JSON object fields):  `VectorStore` and `VectorStoreOptions`,  `VectorStoreFile` and there is no request object, `VectorStoreFileBatch` and there is no request object.
+
+- Operation Updates:
+  - new operations `listVectorStores` , `createVectorStore`, `getVectorStore`, `modifyVectorStore` and `deleteVectorStore`.
+  - new vector store file operations: `listVectorStoreFiles`, `createVectorStoreFile`, `getVectorStoreFile` and `deleteVectorStoreFile`.
+  - new vector store file batch operations: `createVectorStoreFileBatch`, `getVectorStoreFileBatch`, `cancelVectorStoreFileBatch` and `listVectorStoreFileBatchFiles`.
 
 ### Breaking Changes
 
-- Removed method `uploadFile(FileDetails file, FilePurpose purpose)`. Use `uploadFile(FileDetails file, FilePurpose purpose, String fileName)` instead.
+#### Assistants
+
+- Model updates:
+  - Removed `AssistantFile` model.  
+  - `Assistant`, `AssistantCreateOptions` and `UpdateAssistantOptions` models:
+    - removed fields: `fileIds`.
+- Operation updates:
+  - Removed operations: `createAssistantFile` , `listAssistantFiles`, `getAssistantFile` and `deleteAssistantFile`.
+
+#### Files
+
+- Removed method `uploadFile(FileDetails file, FilePurpose purpose)`. Use `uploadFile(FileDetails file, FilePurpose purpose, String fileName)` instead
+
+#### Messages
+
+- Model updates:
+  - Renamed `ThreadInitializationMessage` to `ThreadMessageOptions`.
+  - Removed `MessageFile` model.
+  - Updated `ThreadMessage` model:
+    - The field `incomplete_details` was of the wrong type. Corrected from `MessageIncompleteDetailsReason` -> `MessageIncompleteDetails`.
+    - `assistantId` was marked as optional, but it was in fact nullable
+    - `runId` was marked as optional, but it was in fact nullable
+    - Removed field `fileIds`
+    - Added new field `attachments` a nullable array of `MessageAttachment`
+
+- Operation updates:
+  - Removed `MessageFile` related operations: `listMessageFiles` and `getMessageFile` 
+  - Updated `createMessage` to accept the `ThreadMessageOptions` model (also used in `AssistantThreadCreationOptions`) 
+
+#### Run Step
+
+- Model updates:
+  - `ThreadRun` model updates:
+    - Removed field `fileIds`
+
+#### Threads
+
+- Model updates:
+  - Extracted fields used in `updateThread` operation into model `UpdateAssistantThreadOptions` which now includes the new fields `toolResources`.
+
+- Operation updates:
+  - `updateThread` using extracted model `UpdatedAssistantThreadOptions` instead of parameters using the spread operator.
+
+#### Tools
+
+- Model updates: (mostly about renaming tool `retrieval` to `file_search`)
+  - Renamed model `RetrievalToolDefinition` to `FileSearchToolDefinition` and the associated discriminator value.
+  - Renamed model `RunStepDeltaRetrievalToolCall` to `RunStepDeltaFileSearchToolCall`.
+  - `RunStepToolCall` variant `RunStepRetrievalToolCall` renamed to `RunStepFileSearchToolCall` and the associated discriminator value.
 
 ### Bugs Fixed
 
 - A combination of inputs for `uploadFile` would allow users to not send `String filename` to the service resulting always in an error, as this is actually mandatory.
 
-### Other Changes
 
 ## 1.0.0-beta.2 (2024-02-13)
 
@@ -40,7 +150,7 @@ client.listFilesWithResponse(requestOptions)
 
 ### Other Changes
 
-- On `uploadFile` method in `AssistantsClient` and `AssistantsAsyncClient`, it is required to set the "filename" of the file, via `setFilename` method in `FileDetails` class. The double quote character `"`, the newline charactor `0x0A`, the return charactor `0x0D` in "filename" would be escaped by the client library.
+- On `uploadFile` method in `AssistantsClient` and `AssistantsAsyncClient`, it is required to set the "filename" of the file, via `setFilename` method in `FileDetails` class. The double quote character `"`, the newline character `0x0A`, the return character `0x0D` in "filename" would be escaped by the client library.
 
 
 ## 1.0.0-beta.1 (2024-02-07)

diff --git a/sdk/openai/azure-ai-openai-assistants/README.md b/sdk/openai/azure-ai-openai-assistants/README.md
@@ -103,7 +103,7 @@ String threadId = thread.getId();
 With a thread created, a message can be created on it:
 ```java readme-sample-createMessage
 String userMessage = "I need to solve the equation `3x + 11 = 14`. Can you help me?";
-ThreadMessage threadMessage = client.createMessage(threadId, MessageRole.USER, userMessage);
+ThreadMessage threadMessage = client.createMessage(threadId, new ThreadMessageOptions(MessageRole.USER, userMessage));
 ```
 
 As we have a thread and message, we can create a run:
@@ -115,7 +115,7 @@ There is also a convenience method to create a thread and message, and then run
 ```java readme-sample-createThreadAndRun
 CreateAndRunThreadOptions createAndRunThreadOptions = new CreateAndRunThreadOptions(assistantId)
         .setThread(new AssistantThreadCreationOptions()
-                .setMessages(Arrays.asList(new ThreadInitializationMessage(MessageRole.USER,
+                .setMessages(Arrays.asList(new ThreadMessageOptions(MessageRole.USER,
                         "I need to solve the equation `3x + 11 = 14`. Can you help me?"))));
 run = client.createThreadAndRun(createAndRunThreadOptions);
 ```
@@ -163,12 +163,19 @@ Once uploaded, the file ID can then be provided to an assistant upon creation. N
 an appropriate tool like Code Interpreter or Retrieval is enabled.
 
 ```java readme-sample-createRetrievalAssistant
+// Create Tool Resources. This is how we pass files to the Assistant.
+CreateToolResourcesOptions createToolResourcesOptions = new CreateToolResourcesOptions();
+createToolResourcesOptions.setFileSearch(
+    new CreateFileSearchToolResourceOptions(
+        new CreateFileSearchToolResourceVectorStoreOptionsList(
+            Arrays.asList(new CreateFileSearchToolResourceVectorStoreOptions(Arrays.asList(openAIFile.getId()))))));
+
 Assistant assistant = client.createAssistant(
     new AssistantCreationOptions(deploymentOrModelId)
         .setName("Java SDK Retrieval Sample")
         .setInstructions("You are a helpful assistant that can help fetch data from files you know about.")
-        .setTools(Arrays.asList(new RetrievalToolDefinition()))
-        .setFileIds(Arrays.asList(openAIFile.getId()))
+        .setTools(Arrays.asList(new FileSearchToolDefinition()))
+        .setToolResources(createToolResourcesOptions)
 );
 ```
 
@@ -283,7 +290,7 @@ run via the `SubmitRunToolOutputs` method so that the run can continue:
 
 ```java readme-sample-functionHandlingRunPolling
 do {
-    Thread.sleep(500);
+    Thread.sleep(1000);
     run = client.getRun(thread.getId(), run.getId());
 
     if (run.getStatus() == RunStatus.REQUIRES_ACTION

diff --git a/sdk/openai/azure-ai-openai-assistants/assets.json b/sdk/openai/azure-ai-openai-assistants/assets.json
@@ -2,5 +2,5 @@
   "AssetsRepo" : "Azure/azure-sdk-assets",
   "AssetsRepoPrefixPath" : "java",
   "TagPrefix" : "java/assistants/azure-ai-openai-assistants",
-  "Tag" : "java/assistants/azure-ai-openai-assistants_ceb2cd5662"
+  "Tag" : "java/assistants/azure-ai-openai-assistants_868e5c86d0"
 }