Announcements

We are excited to announce the release of the stable (1.2.0) version for the next batch of modules:

• langchain4j-anthropic
• langchain4j-azure-open-ai
• langchain4j-bedrock
• langchain4j-google-ai-gemini
• langchain4j-mistral-ai
• langchain4j-ollama

Many of the other LC4j modules are released under the 1.2.0-beta8 version and remain experimental/unstable.

The langchain4j-bom has been released with version 1.2.0 and includes the latest versions of all modules.

Notable Changes

• Support thinking/reasoning by @dliubarskyi in #3380
• Streaming partial tool calls by @dliubarskyi in #3303
• MCP: option to automatically expose resources as tools by @jmartisk in #3302
• OpenAI: allow setting custom chat request parameters and access raw HTTP responses and SSE events by @dliubarskyi in #2694
• AI Services: add possibility to transform a ChatRequest before it is sent to the LLM by @mariofusco in #3378
• AI Services: expose intermediate ChatResponses by @dliubarskyi in #3349
• Azure OpenAI: report actual TokenUsage when using streaming chat model by @ssmogos in #3412
• TokenStream: add callback handler before the tool is executed by @ChiaoGeek in #3179

Breaking Changes

• Vertex AI Gemini: fix streaming model: introduce Executor to make it async by @douglas-DS in #3297
• Azure OpenAI: removed TokenCountEstimator from streaming models by @dliubarskyi in #3431

Other Changes

• Ollama - Partial thinking support in Ollama by @bidek in #3135
• Fix: OpenAI, Azure OpenAI: throw ContentFilteredException in case of content filtered or refusal by @dliubarskyi in #3294
• Update CONTRIBUTING.md to mention spotless by @glaforge in #3230
• Gemini: add two boolean features by @neupanerabin in #3245
• Improve DefaultMcpClient's behavior after closing by @jmartisk in #3234
• Add mcp server health check by @Cooosin in #3241
• Fix typo in community BOM dependency declaration by @Cooosin in #3257
• Update documentation with Neo4j features: entity retrievers, conversation memory and graph construction feature examples by @vga91 in #3250
• Add some docs about using MCP without AI services by @jmartisk in #3263
• support pseduo sets in the tool schema by @jtnord in #3152
• Fix typo in test case names by @alfonsomunozpomer in #3239
• [minor fix] Remove unnecessary casts in RetryUtils by @vesense in #3237
• Update grpc-protobuf dependency version to 1.68.1 by @roryp in #3161
• Missing logprobs response parameter on Google AI Gemini's BaseGeminiChatModel class by @glaforge in #3254
• chore(deps): replace dependency com.vackosar.gitflowincrementalbuilder:gitflow-incremental-builder with io.github.gitflow-incremental-builder:gitflow-incremental-builder by @renovate[bot] in #2475
• Fix logging message to reflect correct number of transformed text segments by @minglu7 in #3160
• Fix broken MCP doc link by updating to archived repository URL by @harikrishna553 in #3243
• Use isNullOrEmpty and isNullOrBlank functions from Utils class by @TheJavaGuy in #3071
• Remove the need for Azure dependencies at compile time by @jdubois in #3287
• MCP: roots support by @jmartisk in #3283
• Improve JSON extraction from LLM response by @mariofusco in #3268
• Refactor Azure OpenAI tests to address flackiness by @kpavlov in #3280
• minor fix wrong example code in neo4j.md by @NeatGuyCoding in #3292
• Run integration tests on PR and main only on JDK 21 by @kpavlov in #3300
• Parametrize baseUrl for GoogleAi Gemini models. by @kpavlov in #3272
• Feat/OpenAI compatible docs by @LizeRaes in #3309
• Update dependency maven to v3.9.10 by @renovate[bot] in #3314
• Update dependency com.microsoft.onnxruntime:onnxruntime to v1.22.0 by @renovate[bot] in #3316
• Update ai.djl.version to v0.33.0 by @renovate[bot] in #3049
• Update dependency com.google.protobuf:protobuf-java-util to v3.25.8 by @renovate[bot] in #3313
• Update jlama.version to v0.8.4 by @renovate[bot] in #3048
• Reorder parameters in VespaEmbeddingStore by @timmhirsens in #3301
• Update mikepenz/action-junit-report action to v5 by @renovate[bot] in #3318
• feat: Playwright Support by @vesense in #3282
• add http1.1 for LMStudio in docs by @LizeRaes in #3324
• fix memory leak in load() of PlaywrightDocumentLoader by @NeatGuyCoding in #3327
• Fix for #3335 by @edeandrea in #3336
• Refactor moderation handling and improve exception details by @kpavlov in #3024
• [Doc] Fix extra parenthesis in logging section of documentation by @harikrishna553 in #3339
• Update dependency com.google.guava:guava to v32.1.3-jre by @renovate[bot] in #3341
• Fix #2711 by @dliubarskyi in #3346
• Refactor: Extract validation logic from build() into separate methods by @harikrishna553 in #3344
• Fixed some typo in the documentation by @sunyuhan1998 in #3342
• support api key authentication for Bedrock API by @0x-fang in #3332
• Fix anthropic cache to add cache only last item in systemMessage/Tool list by @Claudio-code in #3337
• fix：extended IT MCP rewiring wait time by @Cooosin in #3325
• Update OpenAI Java to v2.13.1 by @YSZhuoyang in #3366
• Fix bedrock streaming model hangs, when using tools. by @tomas1885 in #3397
• Apply AssertJ best practices as of July 2025 by @timtebeek in #3398
• Simplify equals() method in TableDefinition using Objects.equals by @harikrishna553 in #3400
• Fix npe in DefaultToolExecutor by @weiping-code in #3406
• Introduce SPI for building instances of the GuardrailService by @edeandrea in #3388
• Fix class-level Javadoc: use singular Document and update Tika format… by @harikrishna553 in #3387
• Bump on-headers and compression in /docs by @dependabot[bot] in #3376
• Add support for loading documents with custom DocumentParser by @harikrishna553 in #3386
• Add some unit tests for tool_calls arguments has null value by @weiping-code in #3407
• Add overload of load method to SeleniumDocumentLoader for raw content loading by @harikrishna553 in #3371
• Replaced usage of DefaultCredentialsProvider.create() with DefaultCredentialsProvider.builder().build() by @harikrishna553 in #3368
• Imorovized the doc and testcases for loadDocument to Support Branch, … by @harikrishna553 in #3367
• document JSON Codec SPI by @camilojc in #3305
• Update dependency org.apache.commons:commons-lang3 to v3.18.0 [SECURITY] by @renovate[bot] in #3352
• Remove unused imports from chat request JSON package by @harikrishna553 in #3350
• Add varargs overload for splitAll to DocumentSplitter by @harikrishna553 in #3348
• Update dependency com.google.cloud:libraries-bom to v26.63.0 by @renovate[bot] in #3340
• Update tinylog.version to v2.7.0 by @renovate[bot] in #3317
• Update dependency com.google.cloud:google-cloud-aiplatform to v3.68.0 by @renovate[bot] in #3315
• Fix broken link to self-query retriever documentation by @harikrishna553 in #3411
• minor fix findJsonEnd for json array by @NeatGuyCoding in #3307
• fix dependency convergence rule due to commons-lang3 version by @olamy in #3418
• Refactor equals method to use Objects.equals for improved readability by @harikrishna553 in #3416
• Add graph transformer section by @vga91 in #3110
• Update dependency com.fasterxml.jackson:jackson-bom to v2.19.2 by @renovate[bot] in #3422
• Update dependency com.openai:openai-java to v2.19.1 by @renovate[bot] in #3421
• Filtering Metadata with Infinispan (#720) by @karesti in #3399
• Langchain4j DefaultMcpClient resource leak on throwed constructor (#2977) by @Zalewa in #3343
• [Fix] Fix the testcase should_load_and_parse_html_document by @harikrishna553 in #3430
• Update dependency maven to v3.9.11 by @renovate[bot] in #3428
• Add varargs overload for transformAll in TextSegmentTransformer by @harikrishna553 in #3424
• Update ai-mocks.version to v0.4.6 by @renovate[bot] in #3425
• Introduce SPI for building instances of the InputGuardrailExecutorBuilder/OutputGuardrailExecutorBuilder instances by @edeandrea in #3434
• Ollama - Add support for tracking and retrieving capabilities in Ollama models by @bidek in #3134
• Google Vertex AI: add custom credentials by @mdii in #3393
• [Fix] fix potential multi-threading problems in InfinispanMetadataFilterMapper and add unit tests by @NeatGuyCoding in #3445

New Contributors

• @Cooosin made their first contribution in #3241
• @jtnord made their first contribution in #3152
• @alfonsomunozpomer made their first contribution in #3239
• @roryp made their first contribution in #3161
• @TheJavaGuy made their first contribution in #3071
• @timmhirsens made their first contribution in #3301
• @sunyuhan1998 made their first contribution in #3342
• @0x-fang made their first contribution in #3332
• @YSZhuoyang made their first contribution in #3366
• @tomas1885 made their first contribution in #3397
• @weiping-code made their first contribution in #3406
• @camilojc made their first contribution in #3305
• @douglas-DS made their first contribution in #3297
• @olamy made their first contribution in #3418
• @Zalewa made their first contribution in #3343
• @ssmogos made their first contribution in #3412
• @mdii made their first contribution in #3393
• @ChiaoGeek made their first contribution in #3179

Full Changelog: 1.1.0...1.2.0