[adk-java-releases-bot]

🤖 I have created a release beep boop

1.15.0 (2026-11-02)

Features

• add @experimental annotation for unstable ADK APIs (69914f4)
• add a builder for workflow graph edges (4721491)
• add a context caching demo sample (91a1838)
• add a Gemini context cache manager (e5083e3)
• add a GeminiLiveTransport seam to GeminiLlmConnection (e8b1c20)
• add a Spring Boot chat app backed by an ADK agent (bca0591)
• add a Spring Boot chat that streams replies and mixes ADK and Spring AI tools (7391fa8)
• Add ADK Issue Monitoring (Spam Detection) Agent sample for Java (fd45dda)
• Add ADK Java Issue Triaging Agent sample (fa94438)
• Add ADK PR Triaging Agent for google/adk-java (f14f644)
• add an injectable scheduler for ADK's internal fan-out (1b89d23)
• add App.toBuilder() (2e8dbea)
• add avatar config support to the live streaming flow (fb9274e)
• add bidirectional conversion between ADK and Spring AI (9949265)
• add cacheIntervals and createHttpOptions to ContextCacheConfig, as in ADK Python (3afe4e5)
• Add chat-completions API support to ApigeeLlm (df73784)
• Add ChatCompletionsHTTPClient and support for non-streaming requests (9529c1a)
• Add ChatCompletionsRequest object (88eb0f5)
• Add ChatCompletionsResponse to LlmResponse conversion (ec88c64)
• Add ClassPathSkillSource to load skills from the Java classpath (587073a)
• add Context as the common base of CallbackContext and ToolContext (1309d38)
• Add conversion from LlmRequest to ChatCompletionsRequest (d37f6ee)
• add Event.isolationScope (86f3130)
• add EventActions.agentState for session-resumability checkpoints (8505a01)
• Add GcsOffloader for asynchronously uploading content to Google Cloud Storage (51c9d1a)
• Add GcsOffloader for asynchronously uploading content to Google Cloud Storage (5bad20a)
• Add GcsOffloader for asynchronously uploading content to Google Cloud Storage (a1d2c1c)
• add GitHub release-docs analyzer (Java) (792d2f4)
• add injectable time and UUID providers for timestamps and ids (caccc7c)
• Add JSON cycle detection (1685a4e)
• add NodeInfo.name, runId, and parentRunId, and Event.nodeName (52f2be7)
• add one-way ADK Java to Kotlin engine interop module (baff584)
• Add onRunErrorCallback to ADK Plugin and Runner (3e6b915)
• add opt-in ResumabilityConfig flag for plain-text continuation auto-resume (f0fb722)
• add README for ADK LangChain4j integration library (f861ef9)
• add rewindBeforeInvocationId to EventActions and carry it through the Kotlin interop (4ff229b)
• add RunConfig.dollarEscapedPlaceholders to leave escaped placeholders in instructions as written (72b4af8)
• Add SkillSource interface and implementations for loading skills (509c4aa)
• Add streaming support for ChatCompletionsHTTPClient (384a0c5)
• add support for Gemma models in LlmRegistry (9d6cc80)
• Add support for refusal content using "[[REFUSAL]]:" prefix (e9184c9)
• Add telemetry and metrics recording capabilities (cc3b9ce)
• Add telemetry headers (4009905)
• add the experimental Edge and Start workflow types (a7b317a)
• add the experimental JoinNode workflow node (5137604)
• add the experimental Node and BaseNode workflow types (189d463)
• add the experimental Workflow type for workflow graphs (4c77cdf)
• add the workflow fields Event.output, Event.nodeInfo and EventActions.route (e823fba)
• Add thought signature support for chat completions (287987a)
• add thread-safe immutableEvents, addEvent, and addEvents helpers to Session (950b869)
• Add tools and toolset to use SkillSource in ADK agents (198b2fb)
• Add tools and toolset to use SkillSource in ADK agents (5ee51fd)
• Add tools and toolset to use SkillSource in ADK agents (83a4b71)
• add transcription in event (cb9d2e3)
• Add unique event_id to BigQuery agent analytics events (15c70de)
• Adding functionality to support customer content formating (52323b4)
• Adds the ADK Stale Issue Auditor sample (b6bd2dd)
• advance SequentialAgent to later sub-agents after a HITL resume when resumability is enabled (407478b)
• Allowing McpAsycToolset Builder to take in a McpSessionManager (78766c1)
• BQAA Java preview-readiness fixes (redaction, table bootstrap, drop stats) (c685ece)
• bridge ADK Java plugin onRunErrorCallback to the Kotlin engine (4ee155b)
• bump google-genai dependency to 1.58.0 (3abcf4f)
• carry code-execution, media resolution and more config across the Java/Kotlin interop (4f3b02d)
• carry context cache config and metadata on LlmRequest, LlmResponse and Event (187a4e1)
• carry the continuation token across the Kotlin engine interop (b3f0665)
• configure similarity top-K in the VertexAiRagRetrieval tool (71bd1cd)
• default message content capture in spans to false (81a6a04)
• dev: add the /version and /health endpoints (849e1f6)
• drop model-name checks from GoogleSearchTool and UrlContextTool (887da0e)
• Enhance BigQuery Agent Analytics Plugin with new event types (ec93f50)
• expose turnComplete for live content (857ee63)
• fire ADK Java plugin tool callbacks for native Kotlin tools (fda5a10)
• flows: add RunConfig.groupFunctionResponsesInHistory to group function calls before responses (1b9b395)
• flows: enable forced FC reordering based on gemini-3 model name (fc95ce7)
• Forward state delta from all events to parent session instead of just the last event (f4cd1b7)
• honor the plain-text continuation flag and allow only one resumability flag (f5c7561)
• Implement BigQuery auto-schema upgrade and view creation (14027d1)
• Implement Trace management, add HITL support (7407e37)
• Introduce max span limit to ApiServerSpanExporter (ae13073)
• langchain4j: make modelName optional when ChatModel or StreamingChatModel is provided (03036c3)
• let the ADK Java <-> Kotlin interop run on a caller-supplied dispatcher (29481a9)
• load apps, agents and workflows through AppLoader (9c3d550)
• make BaseAgent extend BaseNode so every agent is a workflow node (dbc973b)
• Make BigQueryAgentAnalyticsPlugin state per-invocation (629c390)
• mcp: allow rejecting config-supplied stdio MCP servers (28810cd)
• pause a workflow for user input (8578068)
• plugins: port DebugLoggingPlugin from adk-python (3c3633e)
• plugins: port ReflectAndRetryToolPlugin from adk-python (33e9f8b)
• plugins: port SaveFilesAsArtifactsPlugin from adk-python (272ea6c)
• Propagate A2A metadata to RunConfig for request-scoped access (285547b)
• rebuild a workflow's node progress from session events (7096019)
• Refactor BigQueryAgentAnalyticsPlugin for async in preparation for GCS offloading (d837ef0)
• refactor OpenTelemetry (OTel) instrumentation within the ADK core, moving from manual span management to structured helper classes (e6fe9aa)
• reject node names that corrupt node paths (08d4b62)
• resume a paused workflow from its session events (2e88a6d)
• resume Gemini generations paused with a continuation token (e8dfcb0)
• run a node once per item of a list input (528a99d)
• run a single workflow node (28f2606)
• run a tool as a workflow node (8b560fc)
• run a workflow graph (cecad8f)
• run a workflow node as the invocation root (66eab8c)
• run an ADK Kotlin-engine runner behind the ADK Java Runner API (6ff8241)
• run child nodes from a workflow node with Context.runNode (e293898)
• serve an app's structure, including workflow graphs (6fd21b5)
• sessions: add an opt-in switch for rejecting a duplicate session id (e6a46b1)
• share a single OkHttpClient with injectable daemon threads across the ADK (2394a95)
• share one run per Context.runNode run ID (a313ce4)
• spring-ai: pass ADK context through tool callbacks (f09aada)
• Support ChatCompletionChunk to LlmResponse conversion (589328e)
• support optional types in function tool parameters (9a06dd3)
• support pausing and resuming a session invocation (8361c5f)
• Support plugins in Java AgentTool similar to Python's implementation (02a08a1)
• Support Sub-agent Escalation event in Parallel Agent (Issue #561) (88c8b0e)
• support the resume runAsync overload when adapting a Kotlin runner to the Java Runner API (4e67b3d)
• Update 'gen_ai.usage.input_tokens' to include tool used tokens to match python ADK (ba23601)
• Update token usage reporting to include thoughts and cache tokens (436b802)
• Updated Spring AI to 2.0.0, ECJ, build works with Java 25 (3f6665b)
• validate workflow graphs at construction time (7674c2b)
• workflow: add NodeExecutionException as the base of NodeTimeoutException (a4c8ac1)
• workflow: let RetryConfig match exceptions by class (fb4c916)

Bug Fixes

• a2a: drop unparseable A2A metadata instead of aborting conversion (b75c916)
• a2a: fail the A2A stream in the handler, not via the transport (faa3482)
• a2a: guard null DataPart metadata in ResponseConverter (fcfd9bd)
• a2a: require explicit adk_type metadata to convert A2A DataParts (b704c5f)
• a2a: skip an unconvertible remote A2A part instead of failing the turn (9debe96)
• Account for nulls in EventActions and State (582cf7c)
• adjust default ToolExecutionMode to SEQUENTIAL as it was actual and widely used behavior for all ADK Java users before parallel tool execution fix (fe88217)
• agents: only stop ParallelAgent on direct sub-agent escalation (fc3e352)
• agents: run LoopAgent zero times when maxIterations is zero or negative (5feaae4)
• agents: warn when AgentTool config_path escapes agent base directory (7a4113e)
• Allow -latest model aliases in GoogleSearchTool (9181ea6)
• Allow BuiltInCodeExecutor for Gemini 3 models (1a3dd61)
• artifacts: do not derive a save's version number from a failed listing (3aaaeb1)
• artifacts: honor the "user:" namespace in InMemoryArtifactService (0f46d5c)
• artifacts: honor user: namespace, prevent phantom keys, sort listing (01e1e41)
• avoid setting Part.partMetadata when metadata is empty (c05206f)
• avoid StackOverflowError in PersistBarrier.awaitPersisted for large steps (a38b824)
• bigquery: BQAA Java P1 preview-readiness fixes (tracing, lifecycle, redaction, HITL) (2027a4b)
• Bypass redundant getSession read in ADK Runner (aaedcaf)
• chat: expose non-success HTTP status via typed exception (052f31a)
• close the Gemini response stream after a read error (461f7d6)
• codeexecutors: add opt-in strict sandbox to ContainerCodeExecutor (8049f7e)
• confine config-driven dynamic class loading to intended types (3967cfa)
• convert unsupported artifact MIME types to text (a60c246)
• core: align Gemini placeholder user turns with ADK Python (234d7dc)
• core: allow Long values to match INTEGER schema type (a6d41cf)
• core: fallback to name when Agent description is missing (233b83b)
• core: only resume tool confirmations for calls this agent emitted (e5aba3a)
• core: retry invalid structured model responses (aa37bfa)
• core: stop re-applying a tool confirmation after a later user turn (a87ee95)
• correctly reassemble streamed function-call arguments in Gemini streaming (6bae658)
• create and use Gemini context caches when ContextCacheConfig is set (68914c5), closes #1530
• dev: allow only same-origin requests by default for CORS and /run_live (5667d38)
• dev: ignore an unusable adk.web.backend-url instead of serving it (f7d7a7c)
• dev: keep '' CORS default, drive WebSocket origins from config, warn on '' (5029081)
• dev: let the dev UI's public address be configured (2c5bd9e)
• dev: reject DNS-rebinding requests that reach the server over loopback (1ae5078)
• dev: restrict default CORS origins and WebSocket allowed origins to localhost (de85300)
• dev: serve the dev UI and its assets under a /dev-ui path (6866919)
• dev: use localhost port wildcard for default CORS/WebSocket origins (cb73317)
• drop a delegating node's own output, as Python ADK does (5177269)
• Enable Jackson serialization for ToolConfig (3ec8fde)
• enforce appName ownership check in VertexAiSessionService.listEvents (6cb6557)
• enforce node retry policies and timeouts (367b402)
• events: accumulate endOfAgent in EventActions.merge to preserve parallel stop requests (03b04fa)
• events: treat an error event without function calls as a final response (ab67dc5)
• fail a timed-out node with NodeTimeoutException (0c661e1)
• fence relayed agent content so it cannot pose as instructions (45b5c34)
• firestore-session-service: honor appName in getSession/deleteSession/listEvents (a370034)
• Fix A2A protocol chunk streaming and task completion states (c95f669)
• Fix ADK Runner race condition for sequential tool execution (69680bb)
• Fix ADK Runner race condition for sequential tool execution (9031cad)
• fix Claude MCP tool inputSchema by falling back to parametersJsonSchema (760c8da)
• Fix critical race condition in ADK Runner (51f4d1f)
• Fix critical race condition in ADK Runner (3091156)
• Fix race condition and stale session in ADK Runner (7964e93)
• flows: avoid StackOverflowError after many steps in BaseLlmFlow (31ac2df)
• flows: compare branch paths on segment boundaries, not raw string prefixes (9219277)
• flows: end invocation on a deferred long-running tool call (6dd4594)
• flows: propagate downstream cancellation to the model stream (cce02ae)
• gemini: align streaming function-call handling with ADK Python (37bb5e6)
• harden disabled mode, config checks and error capture (64ce48c)
• ignore usage-only responses outside bidi (a6cb87a)
• initialize event ID when creating compaction events (fc480ec)
• inject Dev UI tracer into core engine for embedded telemetry (8bccc3b)
• introduce PARALLEL_SUBSCRIBE ToolExecutionMode; restore previous PARALLEL semantics (d3e7f31)
• keep context cache metadata in events and model responses (2c84291)
• keep event isolation scopes in an adapted Java session service (c2ab2e3)
• keep structuredContent and non-text content in MCP tool results (9b128df)
• keep thought signature and tool call parts through streaming and history (d7355a7)
• let a timeout cancel the node a workflow is starting (a017879)
• live: replay stored transcriptions as conversation history (9a15855)
• Make dry_run configurable for ADK Java PR triage, spam detection, and issue triage workflows (4225b07)
• make Java session appends on the Kotlin engine update the caller's session in place (650e950)
• Make stale issue workflow configurable for dry runs (d0edd41)
• map ChatResponse usage metadata to LlmResponse (71f6929)
• map token usage metadata for Anthropic Claude model (f76c5f9)
• mcp: emit valid JSON Schema when converting ADK tools to MCP tools (60250a9)
• mcp: guard empty tool parameters in adkToMcpToolType (66fa921)
• mcp: honor custom URL sub-paths in StreamableHttpServerParameters (a0c4b7b)
• mcp: honor stdioServerParams in McpToolset.fromConfig (cf71d7b)
• memory: format search timestamps as epoch millis (5269b87)
• memory: guard parts without text in searchMemory (e5380f9)
• memory: tokenize query and events alike in InMemoryMemoryService (008d409)
• models: set additionalProperties on schemas for OpenAI strict structured outputs (4b058cd)
• models: treat a part-less candidate as an error, not a success (e0e9425)
• Move @jsoncreator inside LiveRequest Builder (d667db8)
• never retry a WorkflowConfigurationException (fd8c360)
• omit caller values from VertexAiSessionService error messages (dc56f11)
• plugins: prevent BQAA offload data loss and preserve binary MIME (94b5bac)
• pre-merge stateDelta before onUserMessageCallback in Runner (f1155ec)
• preserve all parallel function calls on the live (BIDI) connection (edc330d)
• preserve non-client function call IDs in GeminiUtil (971abb4)
• preserve non-text output in streaming responses (b8be90d)
• preserve provider ChatOptions type to prevent ClassCastException (5c3d328)
• preserve tool responses and thought signatures in Spring AI message conversion (6eef478)
• prevent cross-user session data disclosure in VertexAiSessionService (d1b1d92)
• prevent dropping grounding-only responses in BaseLlmFlow (4de0d8c)
• propagate A2A request metadata into the run config in AgentExecutor (410ff81)
• record the thinking budget and level on the call_llm telemetry span (ed72ddb)
• recover compacted function calls and pair reused call IDs in LLM content assembly (ac542d6)
• Removing deprecated Optional methods (8ef99f9)
• Resolve IllegalArgumentException for text MIME types in LangChain4j adapter (6ad2043)
• Resolve NPE when McpTool description is null (8ed64ea)
• respect OpenTelemetry LIFO scoping in reactive streams (33e28e3)
• restrict live transfer_to_agent to the agent, its sub-agents, parent, and peers (f4d6de4)
• restrict turn-based transfer_to_agent to declared targets (9837e1b)
• resume Kotlin workflows behind an adapted Java session service (aab51ef)
• revert "Suppress empty-text-only chunks from streaming responses while preserving carried metadata" (69638df)
• round-trip thinking blocks in the Claude model (d1d9dc2)
• route HITL confirmation back to originating sub-agent in workflow agents (d608909)
• run approved tool confirmations for agents under a ParallelAgent (f866804)
• run tools concurrently in PARALLEL ToolExecutionMode (020499b)
• runner: build a new message when saving input blobs, instead of writing into the caller's Content (80c1a21)
• runner: keep callbackContextData on a resumed invocation (08876c4)
• Safely handle empty model in GoogleSearchTool request processor (e255192)
• scope ADK Java docs release analyzer to a single language (07a2ec9)
• scope the tool-confirmation resume scan to the current branch (b00e25d)
• send MCP tool input schemas in chat completions requests (ce88237)
• sessions: apply afterTimestamp and numRecentEvents together in VertexAiSessionService (24a4588)
• sessions: apply numRecentEvents and afterTimestamp together in InMemorySessionService (4d19f7d)
• sessions: forward the caller-supplied session id in VertexAiSessionService (77eae25)
• sessions: keep inlineData mimeType when encoding or decoding content (f7af931)
• sessions: reject a duplicate session id in FirestoreSessionService (9fe5830)
• sessions: reject negative numRecentEvents in GetSessionConfig (3adb8cf)
• sessions: validate HTTP status before parsing Vertex AI session responses (4192aca)
• SkillMdPath should be public (29d3203)
• skills: preserve sibling parts when loading binary resources (e1fbed9)
• skills: prevent path traversal in LocalSkillSource (55392f6)
• skills: remove unregistered script tool from default instructions (9d68053)
• skills: use forward slashes in LocalSkillSource.listResources (2a2ab5c)
• skip reserved framework names in MCP toolsets (509d69d)
• spring-ai: order SpringAIAutoConfiguration after Spring AI model auto-configurations (d342a95)
• spring-ai: preserve enum and items in tool schema conversion (7b1a5b8), closes #1609
• stop dropping the latest event(s) in VertexAiSessionService.getSession (987ef4e)
• stop losing event fields on Firestore session reloads (a161009)
• stop losing event fields on Vertex AI session reloads (887f9da)
• stop returning exception text to remote A2A peers (a3df463)
• store grounded events in an adapted Java session service (beca776)
• stream partial chunks, then one aggregated response (3a4742e)
• subscribe to appendEvent Single in runLive (fe6c683)
• support typed responseSchema in ChatCompletionsRequest (8c06011)
• Suppress empty-text-only chunks from streaming responses while preserving carried metadata (b4791ef)
• tools: run an AgentTool under the caller's RunConfig (18560ba)
• Update default BigQueryLoggerConfig table name and remove default dataset ID (723a2ef)
• update Java ADK ResponseConverter to propagate long-running tool IDs on AUTH_REQUIRED state (2d20d1b)
• update stream completion check in A2A SDK to handle all terminal and interrupted task states (2b87d65)
• upgrade Mockito and JaCoCo for Java 25 compatibility (8574fc5)
• use daemon threads in OkHttp dispatchers to allow graceful JVM shutdown (9b046b6)
• Use existing secrets and built-in token in ADK docs release analyzer workflow (456234f)
• wait for the Runner to persist a step's events before the ADK flow's next step (sequential-tool-execution race) (0a40557)
• wait for the transfer event to be persisted before running the transfer target (9f5d5ce)
• widen Integer to Long in castValue() for boxed Long parameter (bc32948)
• workflow: fix replay branches and shared runs on resume (a155d30)

Performance Improvements

• filter session events server-side by afterTimestamp in VertexAiSessionService.getSession (e12baa2)

Reverts

• keeping structuredContent and non-text content in MCP tool results (c897cdb)
• restrict live transfer_to_agent to the agent, its sub-agents, parent, and peers (18554a1)
• restrict turn-based transfer_to_agent to declared targets (84bae34)

Documentation

• add @deprecated guidance to two deprecated event setters (a88b093)
• add placeholder AGENTS.md, CLAUDE.md, and .gemini styleguide (e0d71c0)
• ask for the Java version in the bug report template (f1b8a66)
• attach orphaned javadoc to resolveToolInstance in ComponentRegistry (09a1496)
• clarify LlmAgent composition for workflow agents (49ff63b)
• fix inaccuracies in the tokt interop KDoc (1afd839)
• remove TODO regarding default InMemoryRunner.appName (99adf57)
• update firestore session service dependency versions and artifact identifiers... (5af3854)

---

This PR was generated with Release Please. See documentation.

[kvmilos]

Closing: 1.14.0 already landed on main as b45fae7 (imported from this PR at bc27833), but the PR head moved before the import was submitted, so it was never marked merged and release-please did not tag it. v1.14.0 is now tagged and released on that commit; release-please will open the next PR.