LangChain for Java Files

Notable Changes

• Add OllamaDecisionModel: Ollama decision models, including images by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6659
• Introduce Confidence Cascade agentic pattern by @mariofusco in https://github.com/langchain4j/langchain4j/pull/6765
• Observability: AiServiceInvocationSummaryEvent with all events of one AI Service invocation by @rayanahmed1215 in https://github.com/langchain4j/langchain4j/pull/5048
• langchain4j-test: LLM judges and assertions on chat requests by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6819
• Introduce decision agents by @mariofusco in https://github.com/langchain4j/langchain4j/pull/6820

Other Changes

• chore(deps): bump shell-quote from 1.10.0 to 1.12.0 in /docs by @dependabot[bot] in https://github.com/langchain4j/langchain4j/pull/6602
• Propagate the caller's context to work offloaded by AI Services by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6623
• Docs: non-blocking calls with the Spring Boot starters by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6621
• chore(deps): bump proxy-addr from 2.0.7 to 2.0.8 in /docs by @dependabot[bot] in https://github.com/langchain4j/langchain4j/pull/6591
• docs: Extend MCP server guide with Streamable HTTP server options by @kpavlov in https://github.com/langchain4j/langchain4j/pull/6622
• feat(micrometer-metrics): record chat operation duration by @Snow7-G in https://github.com/langchain4j/langchain4j/pull/6472
• Keep the conversation for output guardrail retries without chat memory by @edeandrea in https://github.com/langchain4j/langchain4j/pull/6627
• Add Azure DocumentDB embedding store, deprecate Azure Cosmos DB Mongo vCore by @SandraAhlgrimm in https://github.com/langchain4j/langchain4j/pull/4426
• Azure DocumentDB docs: Spring Boot starter behaviour, migration defaults and client settings by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6635
• TypeSafe: make the System One API path configurable by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6653
• JdkHttpClient: use HTTP/1.1 for plain http:// URLs by default by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6660
• Docs: HTTP version of SpringRestClient with the JDK client by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6668
• fix(guardrails): keep clean JSON a plain success in JsonExtractorOutputGuardrail by @fengyue-xve in https://github.com/langchain4j/langchain4j/pull/6600
• chore(google-genai): update Google GenAI SDK to 1.77.0 and expose new SDK features by @glaforge in https://github.com/langchain4j/langchain4j/pull/6680
• feat(google-genai): support raw/recursive JSON schemas, multimodal tool results, async chat, and streaming cancellation by @glaforge in https://github.com/langchain4j/langchain4j/pull/6681
• test(google-genai): migrate tests from gemini-2.5-flash to gemini-3.8-flash by @glaforge in https://github.com/langchain4j/langchain4j/pull/6682
• Use MeterProvider for performance reasons by @brunobat in https://github.com/langchain4j/langchain4j/pull/6704
• fix(agentic): make HTML report formatting locale-independent by @Mihir-Rabari in https://github.com/langchain4j/langchain4j/pull/6632
• OpenAI decision models: support images referenced by URL by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6715
• Streaming cancellation: no callbacks after cancel(), make StreamingHandle discoverable by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6717
• feat(decision-models): custom body parameters for all decision models by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6718
• Fix ITs hanging/flaking on slow providers; expose provider errors returned with HTTP 200 by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6716
• Allow TokenStream to process complete responses on a custom executor by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6724
• OpenAiStreamingChatModel: fail on provider errors sent inside the stream by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6722
• Docs: fix ClassPathSkillLoader snippet types and custom patterns count by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6729
• Require exactly one subagent for ParallelMapperService by @L00PLeGeNd in https://github.com/langchain4j/langchain4j/pull/6662
• fix(agentic): count every tool-loop model call in agent token usage by @subhashpolisetti in https://github.com/langchain4j/langchain4j/pull/6691
• Fixes [#6665] Untyped A2A client agent with a missing input fails with a NullPointerException instead of MissingArgumentExceptionUntyped A2A client agent with a missing input fails with a NullPointerException instead of MissingArgumentException by @kamisheng in https://github.com/langchain4j/langchain4j/pull/6666
• Allow to generate the agentic context summarization also using a StreamingChatModel by @mariofusco in https://github.com/langchain4j/langchain4j/pull/6721
• feat(bedrock): add guardContent placement support by @DragonFSKY in https://github.com/langchain4j/langchain4j/pull/5030
• feat(google-genai): support code execution in the chat models by @subhashpolisetti in https://github.com/langchain4j/langchain4j/pull/6709
• Docs: Spring Boot's HTTP client settings in the Spring Boot starters by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6740
• fix(agentic): always release the response an agent keeps per scope by @subhashpolisetti in https://github.com/langchain4j/langchain4j/pull/6693
• Fix the NullPointerException when an optional agent has no output key by @ryanmeowy in https://github.com/langchain4j/langchain4j/pull/6697
• Stop supervisor planning while a SuspendedResponse is pending by @L00PLeGeNd in https://github.com/langchain4j/langchain4j/pull/6636
• Decision models: conversation input for OpenAI and a typed DecisionInput by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6739
• Decision models: conversations for all providers, and conversation context for routers, tool filtering and output guardrails by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6741
• feat(google-genai): add GoogleGenAiTextToSpeechModel for Gemini text-to-speech by @subhashpolisetti in https://github.com/langchain4j/langchain4j/pull/6674
• Allow any agent or agentic pattern to declare an activation condition by @mariofusco in https://github.com/langchain4j/langchain4j/pull/6731
• Fail fast when a BlackboardPlanner or GoalOrientedPlanner has no output key by @ryanmeowy in https://github.com/langchain4j/langchain4j/pull/6720
• Add model name to generated agentic execution html report by @mariofusco in https://github.com/langchain4j/langchain4j/pull/6761
• chore(deps): bump follow-redirects from 1.16.0 to 1.16.1 in /docs by @dependabot[bot] in https://github.com/langchain4j/langchain4j/pull/6770
• Fix registration of inherited listeners for non-AI agents by @mariofusco in https://github.com/langchain4j/langchain4j/pull/6764
• fix(agentic): throw a clear error when a planner agent has no Planner by @zxc210 in https://github.com/langchain4j/langchain4j/pull/6781
• chore(deps): bump launch-editor from 2.14.1 to 2.14.2 in /docs by @dependabot[bot] in https://github.com/langchain4j/langchain4j/pull/6777
• Resolve SupplierParameterResolver parameters on every static method of declarative agents by @mariofusco in https://github.com/langchain4j/langchain4j/pull/6778
• AI Services: complete and consistent observability events by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6786
• Decision models: add from(...) factories, deprecate of(...) by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6803
• OpenAI decision models: add promptCacheKey by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6807
• Bump Jackson to 2.22.4 and 3.2.4 to fix DoS advisories by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6808
• Update dependency org.codehaus.mojo:build-helper-maven-plugin to v3.6.2 by @renovate[bot] in https://github.com/langchain4j/langchain4j/pull/6429
• Update dependency org.apache.pdfbox:pdfbox to v3.0.8 by @renovate[bot] in https://github.com/langchain4j/langchain4j/pull/6428
• Renovate: create Jackson update PRs first by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6811
• Update docker-java to 3.7.1 by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6813
• Update Google Cloud libraries by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6812
• Integration tests: use the project's Jackson version in the Quarkus and Spring test apps by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6814
• Renovate: group related updates, skip db2 hotfix builds, keep onnxrun… by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6816
• Add tool search strategy to agents by @subhashpolisetti in https://github.com/langchain4j/langchain4j/pull/6792
• Update onnxruntime to 1.23.2 by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6815
• Update dependency com.squareup.okhttp3:okhttp-bom to v4.12.1 by @renovate[bot] in https://github.com/langchain4j/langchain4j/pull/6817
• Update dependency org.apache.httpcomponents.client5:httpclient5 to v5.6.5 by @renovate[bot] in https://github.com/langchain4j/langchain4j/pull/6828
• Update dependency org.apache.httpcomponents.core5:httpcore5 to v5.4.4 by @renovate[bot] in https://github.com/langchain4j/langchain4j/pull/6829
• Update dependency com.qcloud:cos_api to v5.6.283 by @renovate[bot] in https://github.com/langchain4j/langchain4j/pull/6810
• Update dependency java-jdk to v17.0.21+9 by @renovate[bot] in https://github.com/langchain4j/langchain4j/pull/6826
• Update dependency io.projectreactor.tools:blockhound to v1.0.18.RELEASE by @renovate[bot] in https://github.com/langchain4j/langchain4j/pull/6825
• MCP: remove the deprecated Jackson 2 APIs so applications can exclude Jackson 2 by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6824
• Bump Azure, AWS, Infinispan, Tika and Vespa to fix Netty and Bouncy Castle advisories by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6849
• Agentic: don't report a token total of several models under one model name by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6861
• Fix the parallel mapper agent losing or mixing up the results of its items by @subhashpolisetti in https://github.com/langchain4j/langchain4j/pull/6841
• Update dependency org.jacoco:jacoco-maven-plugin to v0.8.16 by @renovate[bot] in https://github.com/langchain4j/langchain4j/pull/6853
• Update dependency org.infinispan.protostream:protostream-processor to v6.0.12 by @renovate[bot] in https://github.com/langchain4j/langchain4j/pull/6852
• Update dependency org.apache.pdfbox:pdfbox to v3.0.8 by @renovate[bot] in https://github.com/langchain4j/langchain4j/pull/6850
• Update dependency org.apache.opennlp:opennlp-tools to v2.5.12 by @renovate[bot] in https://github.com/langchain4j/langchain4j/pull/6847
• chore(deps): bump fast-uri from 3.1.8 to 3.1.9 in /docs by @dependabot[bot] in https://github.com/langchain4j/langchain4j/pull/6848
• chore(deps): bump body-parser from 1.20.6 to 1.20.9 in /docs by @dependabot[bot] in https://github.com/langchain4j/langchain4j/pull/6846
• chore(deps): bump postcss-selector-parser in /docs by @dependabot[bot] in https://github.com/langchain4j/langchain4j/pull/6860
• Notify agent listeners of the AI Service events of their agents by @subhashpolisetti in https://github.com/langchain4j/langchain4j/pull/6839
• MCP client: use HttpClient SPI by @danielkec in https://github.com/langchain4j/langchain4j/pull/6726
• A few MCP client fixes related to tracing by @jmartisk in https://github.com/langchain4j/langchain4j/pull/6703
• Do not request structured output when AI Service returns ChatResponse by @percyvega in https://github.com/langchain4j/langchain4j/pull/6642
• Skills: follow symbolic links only within the skill directory by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6876
• Report to agent listeners the model response of each invocation, also when the same agent runs concurrently by @mariofusco in https://github.com/langchain4j/langchain4j/pull/6875
• Also skip Vespa integration test when the flag -DskipITs is used by @mariofusco in https://github.com/langchain4j/langchain4j/pull/6879
• Update dependency org.mariadb.jdbc:mariadb-java-client to v3.5.11 by @renovate[bot] in https://github.com/langchain4j/langchain4j/pull/6864
• Update dependency org.mariadb.jdbc:mariadb-java-client to v3.5.11 by @renovate[bot] in https://github.com/langchain4j/langchain4j/pull/6863
• Update dependency org.postgresql:postgresql to v42.7.14 by @renovate[bot] in https://github.com/langchain4j/langchain4j/pull/6870
• Update dependency org.postgresql:postgresql to v42.7.14 by @renovate[bot] in https://github.com/langchain4j/langchain4j/pull/6871
• Update dependency com.google.guava:guava to v33 [SECURITY] by @renovate[bot] in https://github.com/langchain4j/langchain4j/pull/6877
• Bind metadata filter values as parameters (pgvector, MariaDB) and map them strictly (Milvus, MongoDB Atlas, Chroma) by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6895
• Validate the types of metadata filter comparison values by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/6896

New Contributors

• @fengyue-xve made their first contribution in https://github.com/langchain4j/langchain4j/pull/6600
• @Mihir-Rabari made their first contribution in https://github.com/langchain4j/langchain4j/pull/6632
• @L00PLeGeNd made their first contribution in https://github.com/langchain4j/langchain4j/pull/6662
• @kamisheng made their first contribution in https://github.com/langchain4j/langchain4j/pull/6666
• @ryanmeowy made their first contribution in https://github.com/langchain4j/langchain4j/pull/6697
• @rayanahmed1215 made their first contribution in https://github.com/langchain4j/langchain4j/pull/5048
• @percyvega made their first contribution in https://github.com/langchain4j/langchain4j/pull/6642

Full Changelog: https://github.com/langchain4j/langchain4j/compare/1.22.0...1.23.0