Azure · kaylieee · May 29, 2026 · May 29, 2026
@@ -1,22 +1,39 @@
 # Release History
 
-## 2.1.0 (Unreleased)
+## 2.1.0 (2026-05-29)
 
 ### Features Added
 
 - Added protocol-style methods on `ResponsesClient` and `ResponsesAsyncClient` that accept a raw JSON request body (`BinaryData`) and a `com.openai.core.RequestOptions`, and return the openai-java raw HTTP response. These mirror the existing `createAzureResponse` and `createStreamingAzureResponse` typed surface: `createResponseWithResponse` (returns `HttpResponseFor<Response>`) and `createResponseStreamWithResponse` (returns `HttpResponseFor<StreamResponse<ResponseStreamEvent>>`). They delegate to the underlying openai-java `ResponseService.withRawResponse()` surface and continue to flow through the Azure HTTP pipeline.
-
-### Other Changes
-
-- Enabled `ResponsesTests` and `ResponsesAsyncTests` (previously `@Disabled`) with create/retrieve/delete/input-items and background-cancel coverage for the typed (`ResponseService` / `ResponseServiceAsync`) surface, plus coverage for the new protocol-method surface. Recordings published to `Azure/azure-sdk-assets` and referenced from `assets.json`.
+- Added preview support for external agents via `ExternalAgentDefinition`, `AgentKind.EXTERNAL`, and `AgentDefinitionOptInKeys.EXTERNAL_AGENTS_V1_PREVIEW`.
+- Added preview code-based hosted agent operations on `AgentsClient` and `AgentsAsyncClient`, including `createAgentVersionFromCode`, `updateAgentFromCode`, and `downloadAgentCode`, plus related code package models such as `CreateAgentVersionFromCodeContent`, `CodeFileDetails`, and `CodeDependencyResolution`. `CodeConfiguration` now exposes the service-computed code package hash via `getContentSha256()`.
+- Added preview agent optimization job and candidate management operations on `AgentsClient` and `AgentsAsyncClient`, including creating, listing, retrieving, canceling, and deleting optimization jobs, listing and inspecting candidates, downloading candidate files, and promoting candidates.
+- Added `stopSession` and `stopSessionWithResponse` to stop hosted-agent sessions.
+- Added `force` query parameter support for hosted-agent `deleteAgentWithResponse` and `deleteAgentVersionWithResponse` requests through `RequestOptions`, allowing active sessions to be cascade-deleted.
+- Added individual memory item operations to `MemoryStoresClient` and `MemoryStoresAsyncClient`: `createMemory`, `updateMemory`, `listMemories`, `getMemory`, and `deleteMemory`, with new `ListMemoriesOptions`, `DeleteMemoryResponse`, and `MemoryItemKind.PROCEDURAL` support.
+- Added new preview tools `FabricIqPreviewTool` and `ToolboxSearchPreviewTool`, plus related tool call/output models for Azure tools.
+- Added optional per-tool configuration via `ToolConfig` and `toolConfigs` accessors on supported tool classes.
+- Added `getComparisonFilter()` and `getCompoundFilter()` convenience getters on `FileSearchTool` for retrieving OpenAI filter types.
+- Added new feature-flag values, including `AgentDefinitionOptInKeys.CODE_AGENTS_V1_PREVIEW`, `AgentDefinitionOptInKeys.EXTERNAL_AGENTS_V1_PREVIEW`, and `FoundryFeaturesOptInKeys.AGENTS_OPTIMIZATION_V1_PREVIEW`.
+- Added hosted-agent, Fabric IQ, Toolbox Search, and async toolbox samples.
 
 ### Breaking Changes
 
+- `AgentEndpoint` renamed to `AgentEndpointConfig`.
+- Session file listing methods on `AgentSessionFilesClient` and `AgentSessionFilesAsyncClient` were renamed from `getSessionFiles` to `listSessionFiles` and now return paged `SessionDirectoryEntry` results. `SessionDirectoryListResponse` was removed.
+- Hosted-agent session methods no longer take a required `isolationKey` argument. Use overloads that accept the optional `userIsolationKey` value, or set the `x-ms-user-isolation-key` header through `RequestOptions`.
+- `AgentDefinitionOptInKeys.CONTAINER_AGENTS_V1_PREVIEW` was removed. Use the applicable hosted-agent, code-agent, agent-endpoint, workflow-agent, or external-agent opt-in key instead.
+- `HostedAgentDefinition` no longer exposes top-level `image` or `containerProtocolVersions` accessors. Use `ContainerConfiguration` for container images and `protocolVersions` for ingress protocol configuration.
+- `CodeConfiguration` constructor now requires `CodeDependencyResolution` in addition to runtime and entry point.
+- `WorkIqPreviewTool` now takes the Work IQ project connection ID directly. `WorkIQPreviewToolParameters` was removed.
+
 ### Bugs Fixed
 
 ### Other Changes
 
+- Enabled `ResponsesTests` and `ResponsesAsyncTests` (previously `@Disabled`) with create/retrieve/delete/input-items and background-cancel coverage for the typed (`ResponseService` / `ResponseServiceAsync`) surface, plus coverage for the new protocol-method surface. Recordings published to `Azure/azure-sdk-assets` and referenced from `assets.json`.
 - Re-enabled `SessionLogSyncTest` and `SessionLogAsyncTest`; both tests are recordable via `@RecordWithoutRequestBody` and run live against the configured Foundry project.
+- Regenerated client from the updated TypeSpec specification.
 
 ## 2.1.0-beta.1 (2026-05-12)
 

@@ -57,9 +57,9 @@ AgentsAsyncClient agentsAsyncClient = new AgentsClientBuilder()
 ``` 
 
 The Agents client library has the following sub-clients which group the different operations that can be performed: 
-- `AgentsClient` / `AgentsAsyncClient`: Perform operations related to agents, such as creating, retrieving, updating, and deleting agents. Also includes agent-session operations (`createSession`, `getSession`, `deleteSession`, `listSessions`, `getSessionLogStream`, `getSessionLogStreamWithResponse`).
+- `AgentsClient` / `AgentsAsyncClient`: Perform operations related to agents, such as creating, retrieving, updating, and deleting agents. Also includes hosted-agent sessions, code package operations and preview agent optimization operations.
 - `ResponsesClient` / `ResponsesAsyncClient`: Handle responses operations. See the [OpenAI's Responses API documentation][openai_responses_api_docs] for more information.
-- `MemoryStoresClient` / `MemoryStoresAsyncClient` **(preview)**: Manage memory stores for agents. This operation group requires the `MemoryStores=V1Preview` feature opt-in flag and is automatically set by the SDK on every request.
+- `MemoryStoresClient` / `MemoryStoresAsyncClient` **(preview)**: Manage memory stores and individual memory items for agents. This operation group requires the `MemoryStores=V1Preview` feature opt-in flag and is automatically set by the SDK on every request.
 - `ToolboxesClient` / `ToolboxesAsyncClient` **(preview)**: Manage toolboxes and toolbox versions. This operation group requires the `Toolboxes=V1Preview` feature opt-in flag and is automatically set by the SDK on every request.
 - `AgentSessionFilesClient` / `AgentSessionFilesAsyncClient` **(preview)**: Work with files in an agent session, including uploading, downloading, listing, and deleting session files.
 
@@ -121,21 +121,25 @@ The SDK supports a variety of tools that can be attached to agent definitions. S
 | `BingCustomSearchPreviewTool` | Bing custom search |
 | `BrowserAutomationPreviewTool` | Browser automation |
 | `ComputerUsePreviewTool` | Computer use |
+| `FabricIqPreviewTool` | Fabric IQ |
 | `McpTool` | Model Context Protocol (MCP) |
 | `MemorySearchPreviewTool` | Memory search |
 | `MicrosoftFabricPreviewTool` | Microsoft Fabric |
 | `SharepointPreviewTool` | SharePoint grounding |
+| `ToolboxSearchPreviewTool` | Toolbox search |
 | `WebSearchPreviewTool` | Web search |
 | `WorkIqPreviewTool` | Work IQ |
 
+Supported tool classes may also expose optional `name`, `description`, and `toolConfigs` properties for user-defined labels and per-tool configuration.
+
 ### Experimental features and opt-in flags
 
 Some features require an opt-in via the `Foundry-Features` HTTP header. The SDK provides two enums for these flags:
 
-- **`AgentDefinitionOptInKeys`** — Used when creating or updating agents. Passed as a parameter to `createAgent`, `updateAgent`, `createAgentVersion`, and related methods. Available keys: `HOSTED_AGENTS_V1_PREVIEW`, `WORKFLOW_AGENTS_V1_PREVIEW`, `CONTAINER_AGENTS_V1_PREVIEW`, `AGENT_ENDPOINT_V1_PREVIEW`.
-- **`FoundryFeaturesOptInKeys`** — Defines all known opt-in keys, including: `EVALUATIONS_V1_PREVIEW`, `SCHEDULES_V1_PREVIEW`, `RED_TEAMS_V1_PREVIEW`, `INSIGHTS_V1_PREVIEW`, `MEMORY_STORES_V1_PREVIEW`, `TOOLBOXES_V1_PREVIEW`, `SKILLS_V1_PREVIEW`.
+- **`AgentDefinitionOptInKeys`** — Used when creating or updating agents. Passed as a parameter to `createAgent`, `updateAgent`, `createAgentVersion`, and related methods. Available keys: `HOSTED_AGENTS_V1_PREVIEW`, `WORKFLOW_AGENTS_V1_PREVIEW`, `AGENT_ENDPOINT_V1_PREVIEW`, `CODE_AGENTS_V1_PREVIEW`, `EXTERNAL_AGENTS_V1_PREVIEW`.
+- **`FoundryFeaturesOptInKeys`** — Defines all known opt-in keys, including: `EVALUATIONS_V1_PREVIEW`, `SCHEDULES_V1_PREVIEW`, `RED_TEAMS_V1_PREVIEW`, `INSIGHTS_V1_PREVIEW`, `MEMORY_STORES_V1_PREVIEW`, `ROUTINES_V1_PREVIEW`, `TOOLBOXES_V1_PREVIEW`, `SKILLS_V1_PREVIEW`, `DATA_GENERATION_JOBS_V1_PREVIEW`, `MODELS_V1_PREVIEW`, `AGENTS_OPTIMIZATION_V1_PREVIEW`.
 
-> **Note:** The `MemoryStoresClient` automatically sets the `MemoryStores=V1Preview` opt-in flag on every request. The `ToolboxesClient` automatically sets the `Toolboxes=V1Preview` opt-in flag on every request.
+> **Note:** The `MemoryStoresClient` automatically sets the `MemoryStores=V1Preview` opt-in flag on every request. The `ToolboxesClient` automatically sets the `Toolboxes=V1Preview` opt-in flag on every request. Agent optimization methods accept `FoundryFeaturesOptInKeys.AGENTS_OPTIMIZATION_V1_PREVIEW`; code-based hosted agents and external agents use the corresponding `AgentDefinitionOptInKeys` values.
 
 ```java
 // OpenAI SDK ResponseService accessed from ResponsesClient
@@ -147,6 +151,23 @@ OpenAIClient openAIClient = builder.buildOpenAIClient();
 ConversationService conversationService = openAIClient.conversations();
 ```
 
+### Preview hosted-agent capabilities
+
+Hosted-agent previews are exposed on `AgentsClient` and `AgentsAsyncClient`. The following capabilities require the corresponding opt-in flag when you create or modify preview resources:
+
+| Capability | APIs and models | Opt-in flag |
+|---|---|---|
+| Code-based hosted agents | `createAgentVersionFromCode`, `updateAgentFromCode`, `downloadAgentCode`, `CodeConfiguration`, `CodeDependencyResolution` | `AgentDefinitionOptInKeys.CODE_AGENTS_V1_PREVIEW` |
+| External agents | `ExternalAgentDefinition`, `AgentKind.EXTERNAL` | `AgentDefinitionOptInKeys.EXTERNAL_AGENTS_V1_PREVIEW` |
+| Agent endpoints and sessions | `AgentEndpointConfig`, `createSession`, `listSessions`, `stopSession`, `AgentSessionFilesClient` | `AgentDefinitionOptInKeys.AGENT_ENDPOINT_V1_PREVIEW` |
+| Agent optimization | `createOptimizationJob`, `listOptimizationJobs`, `listOptimizationCandidates`, `promoteOptimizationCandidate` | `FoundryFeaturesOptInKeys.AGENTS_OPTIMIZATION_V1_PREVIEW` |
+
+For code-based hosted agents, `CodeConfiguration.getContentSha256()` returns the service-computed SHA-256 hash of the uploaded code package. Session APIs that need per-user isolation can use overloads that accept `userIsolationKey`, or set the `x-ms-user-isolation-key` header through `RequestOptions`. To delete hosted agents or agent versions that still have active sessions, add the `force=true` query parameter through `RequestOptions` when calling the corresponding `deleteAgentWithResponse` or `deleteAgentVersionWithResponse` method.
+
+### Memory item management
+
+`MemoryStoresClient` and `MemoryStoresAsyncClient` manage memory stores and individual memory items. In addition to store-level operations, use `createMemory`, `updateMemory`, `listMemories`, `getMemory`, and `deleteMemory` to manage individual memories. `ListMemoriesOptions` supports filtering by scope and `MemoryItemKind`, including `MemoryItemKind.PROCEDURAL`.
+
 ### Using OpenAI's official library
 
 If you prefer using the [OpenAI official Java client library][openai_java_sdk] instead, you can do so by including that dependency in your project instead and following the instructions in the linked repository. Additionally, you will have to set up your `OpenAIClient` as shown below:
@@ -492,6 +513,22 @@ See the full sample in [FabricSync.java](https://github.com/Azure/azure-sdk-for-
 
 ---
 
+##### **Fabric IQ (Preview)**
+
+Connect agents to Fabric IQ project connections for enterprise data grounding:
+
+```java com.azure.ai.agents.define_fabric_iq
+FabricIqPreviewTool fabricIqTool = new FabricIqPreviewTool(fabricIqConnectionId)
+    .setServerLabel("fabric_iq")
+    .setRequireApproval("never")
+    .setName("fabric_iq_lookup")
+    .setDescription("Use FabricIQ to answer questions grounded in enterprise data.");
+```
+
+See the full sample in [FabricIQSync.java](https://github.com/Azure/azure-sdk-for-java/tree/main/sdk/ai/azure-ai-agents/src/samples/java/com/azure/ai/agents/tools/FabricIQSync.java).
+
+---
+
 ##### **Microsoft SharePoint (Preview)** ([documentation](https://learn.microsoft.com/azure/foundry/agents/how-to/tools/sharepoint?pivots=java))
 
 Search through SharePoint documents for grounding:
@@ -576,6 +613,38 @@ See the full sample in [OpenApiWithConnectionSync.java](https://github.com/Azure
 
 ---
 
+#### Toolbox Tools
+
+Toolbox tools are defined in toolbox versions and managed through `ToolboxesClient` / `ToolboxesAsyncClient`.
+
+##### **Toolbox Search (Preview)**
+
+Use `ToolboxSearchPreviewTool` inside a toolbox version to let an agent search the available toolbox tools at runtime:
+
+```java com.azure.ai.agents.toolboxes.ToolboxSearchToolboxSample.createToolboxSearchToolbox
+ToolboxSearchPreviewTool toolboxSearchTool = new ToolboxSearchPreviewTool()
+    .setName("search_tools")
+    .setDescription("Search over available toolbox tools at runtime.");
+
+ToolboxVersionDetails version = toolboxesClient.createToolboxVersion(
+    toolboxName,
+    Collections.singletonList(toolboxSearchTool),
+    "Toolbox version with a Toolbox Search preview tool.",
+    null,
+    null,
+    null);
+
+System.out.printf("Created toolbox: %s%n", version.getName());
+System.out.printf("Toolbox version: %s%n", version.getVersion());
+for (Tool tool : version.getTools()) {
+    System.out.printf("Tool type: %s%n", tool.getType());
+}
+```
+
+See the full sample in [ToolboxSearchToolboxSample.java](https://github.com/Azure/azure-sdk-for-java/tree/main/sdk/ai/azure-ai-agents/src/samples/java/com/azure/ai/agents/toolboxes/ToolboxSearchToolboxSample.java).
+
+---
+
 ### Streaming responses
 
 The `ResponsesClient` and `ResponsesAsyncClient` support streaming, which allows you to process response events as they arrive rather than waiting for the full response. This is useful for displaying text to users in real time and observing tool execution progress.

@@ -1,15 +1,30 @@
 # Release History
 
-## 2.1.0 (Unreleased)
+## 2.1.0 (2026-05-29)
 
 ### Features Added
 
+- Added new preview `DataGenerationJobsClient` and `DataGenerationJobsAsyncClient` sub-clients for creating, retrieving, listing, canceling, and deleting data generation jobs, with related data generation source, output, token usage, and options models.
+- Added new preview `ModelsClient` and `ModelsAsyncClient` sub-clients for registering and managing model weights as Foundry `ModelVersion` resources, including pending upload, asynchronous model-version creation, listing, retrieving, updating, deleting, and credential retrieval operations.
+- Added new preview `RoutinesClient` and `RoutinesAsyncClient` sub-clients for creating or updating routines, retrieving, enabling, disabling, listing, deleting, listing runs, and dispatching routines, with related routine trigger, action, dispatch payload, and run models.
+- Added evaluator generation job operations to `EvaluatorsClient` and `EvaluatorsAsyncClient`, including create, get, list, cancel, and delete operations, plus evaluator pending-upload and credential helper operations.
+- Added versioned skill management on `SkillsClient` and `SkillsAsyncClient`, including `createSkillVersion`, `createSkillVersionFromFiles`, `listSkillVersions`, `getSkillVersion`, `getSkillContent`, `getSkillVersionContent`, and `deleteSkillVersion`.
+- Added `EvaluationsHelper` and Azure evaluator model types, such as `TestingCriterionAzureAIEvaluator` and `GraderAzureAIEvaluator`, to adapt Azure evaluator definitions to OpenAI evaluation request types.
+- Added protocol-style `WithResponse` overloads with `RequestOptions` for connection retrieval and dataset file/folder upload helpers.
+- Added new feature-flag values to `FoundryFeaturesOptInKeys`: `ROUTINES_V1_PREVIEW` (`Routines=V1Preview`), `DATA_GENERATION_JOBS_V1_PREVIEW` (`DataGenerationJobs=V1Preview`), `MODELS_V1_PREVIEW` (`Models=V1Preview`), and `AGENTS_OPTIMIZATION_V1_PREVIEW` (`AgentsOptimization=V1Preview`).
+
 ### Breaking Changes
 
+- The Skills API is now versioned. `SkillDetails` was removed and replaced by `Skill` and `SkillVersion`. Use `createSkillVersion` instead of `createSkill`, `createSkillVersionFromFiles` instead of `createSkillFromPackage`, `getSkillContent`/`getSkillVersionContent` instead of `downloadSkill`, and `deleteSkillVersion` for deleting a specific skill version. `updateSkill` now updates the default version only.
+- The `body` parameter on `EvaluationTaxonomiesClient.createEvaluationTaxonomy`, `EvaluationTaxonomiesClient.updateEvaluationTaxonomy`, and their async counterparts was renamed to `taxonomy`.
+
 ### Bugs Fixed
 
 ### Other Changes
 
+- Regenerated client from the updated TypeSpec specification.
+- Added samples for data generation jobs, model management, evaluator helper usage, and packaged skill upload/download.
+
 ## 2.1.0-beta.1 (2026-05-12)
 
 ### Features Added