| Name | Modified | Size | Downloads / Week |
|---|---|---|---|
| Parent folder | |||
| 1.18.0 and 1.18.0-beta28 source code.tar.gz | 2026-07-17 | 13.4 MB | |
| 1.18.0 and 1.18.0-beta28 source code.zip | 2026-07-17 | 16.3 MB | |
| README.md | 2026-07-17 | 11.0 kB | |
| Totals: 3 Items | 29.7 MB | 2 | |
Notable Changes
- Introduce Belief-Desire-Intention (BDI) agentic pattern by @mariofusco in https://github.com/langchain4j/langchain4j/pull/5730
- Add crash-resilient Human-in-the-Loop suspension and resume for agentic systems by @mariofusco in https://github.com/langchain4j/langchain4j/pull/5767
- Add OpenAI Text-to-Speech (TTS) support by @anunnakian in https://github.com/langchain4j/langchain4j/pull/4697
- Add
MistralAiBatchChatModelfor the Mistral Batch API by @subhashpolisetti in https://github.com/langchain4j/langchain4j/pull/5750 - EmbeddingModel: request/response API with per-call parameters, multimodal inputs, and observability by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/5735
Other Changes
- feat(anthropic): add midConversationSystemMessages option for mid-conversation system messages by @nickmaddren in https://github.com/langchain4j/langchain4j/pull/5604
- fix: Throw typed exception for null-valued required skill tool argument by @thswlsqls in https://github.com/langchain4j/langchain4j/pull/5612
- fix: mask apiKey in OvhAiEmbeddingModel builder toString by @thswlsqls in https://github.com/langchain4j/langchain4j/pull/5608
- fix: Include $defs in Vertex AI Anthropic tool input schema for tools using schema references by @thswlsqls in https://github.com/langchain4j/langchain4j/pull/5615
- fix: Map Vertex AI Gemini content-policy finish reasons to CONTENT_FILTER by @thswlsqls in https://github.com/langchain4j/langchain4j/pull/5616
- feat(anthropic): typed Skills support for document generation (#5530 Phase 1) by @PSchmitz-Valckenberg in https://github.com/langchain4j/langchain4j/pull/5605
- fix: Guard empty input and validate id/embedding sizes in VespaEmbeddingStore.addAll by @thswlsqls in https://github.com/langchain4j/langchain4j/pull/5632
- fix: Skip search results with unresolvable URIs in GoogleCustomWebSearchEngine by @thswlsqls in https://github.com/langchain4j/langchain4j/pull/5628
- fix: Remove leading space from DEEPSEEK_CODER_MATH_7B_AWQ model id by @thswlsqls in https://github.com/langchain4j/langchain4j/pull/5637
- fix: Override equals, hashCode and toString in WatsonxChatRequestParameters by @thswlsqls in https://github.com/langchain4j/langchain4j/pull/5636
- fix: Decode Voyage AI base64 embeddings as little-endian by @thswlsqls in https://github.com/langchain4j/langchain4j/pull/5638
- fix: honor root metadata mode in Weaviate removeAll(Filter) by @thswlsqls in https://github.com/langchain4j/langchain4j/pull/5631
- feat(gemini): support per-request imageConfig override in GoogleAiGeminiChatModel by @ZhangDT-sky in https://github.com/langchain4j/langchain4j/pull/4913
- fix: Wrap the full last expression when fixing single-line JavaScript in Judge0 by @thswlsqls in https://github.com/langchain4j/langchain4j/pull/5650
- fix: Preserve binary bytes in Azure dynamic sessions file download by @thswlsqls in https://github.com/langchain4j/langchain4j/pull/5652
- Workaround HHH-20500 for runtimes that haven't updated to the latest Hibernate ORM bug fixes yet by @beikov in https://github.com/langchain4j/langchain4j/pull/5655
- fix: Paginate ListObjectsV2 in AmazonS3DocumentLoader to load all objects by @thswlsqls in https://github.com/langchain4j/langchain4j/pull/5662
- feat: Add prefix overload to AzureBlobStorageDocumentLoader.loadDocuments by @thswlsqls in https://github.com/langchain4j/langchain4j/pull/5669
- Anthropic: add cache diagnostics (beta) support by @joshiste in https://github.com/langchain4j/langchain4j/pull/5659
- fix: Encode page content as UTF-8 in SeleniumDocumentLoader by @thswlsqls in https://github.com/langchain4j/langchain4j/pull/5671
- fix: Guard against null content type in GcsSource metadata by @thswlsqls in https://github.com/langchain4j/langchain4j/pull/5673
- feat(gemini): support per-request cachedContentName override by @PSchmitz-Valckenberg in https://github.com/langchain4j/langchain4j/pull/5645
- Propagate image detail level in OpenAI official chat completions by @smf-h in https://github.com/langchain4j/langchain4j/pull/5617
- docs: use clearly fictional example data in structured-outputs tutorial by @jialiuyang in https://github.com/langchain4j/langchain4j/pull/5620
- fix: advance the planner when an agent error handler returns a result by @subhashpolisetti in https://github.com/langchain4j/langchain4j/pull/5682
- feat(google-genai): support context cache creation and management by @subhashpolisetti in https://github.com/langchain4j/langchain4j/pull/5694
- fix: Include failing file URL in GitHubDocumentLoader load-failure exception message by @thswlsqls in https://github.com/langchain4j/langchain4j/pull/5684
- feat(google-genai): support per-request cachedContent override by @glaforge in https://github.com/langchain4j/langchain4j/pull/5676
- fix: Use UTF-8 when encoding Playwright page content for the document parser by @thswlsqls in https://github.com/langchain4j/langchain4j/pull/5686
- fix: Support all date functions when extracting metadata key in SqlFilterParser by @thswlsqls in https://github.com/langchain4j/langchain4j/pull/5699
- fix: mask apiKey in Cohere builder toString by @lixuanqun in https://github.com/langchain4j/langchain4j/pull/5691
- fix: guard against null function in streamed OpenAI tool-call deltas by @farzad-sedaghatbin in https://github.com/langchain4j/langchain4j/pull/5712
- fix: defensively copy input list in ChatMemory.set(...) before eviction by @farzad-sedaghatbin in https://github.com/langchain4j/langchain4j/pull/5714
- fix: align IsIn/IsNotIn numeric comparison with IsEqualTo/IsNotEqualTo by @farzad-sedaghatbin in https://github.com/langchain4j/langchain4j/pull/5717
- Improve visualization of execution history in agentic html report by @mariofusco in https://github.com/langchain4j/langchain4j/pull/5606
- Allow to use non-public and inherited methods as agentic methods by @mariofusco in https://github.com/langchain4j/langchain4j/pull/5704
- Allow to configure A2A transport by @mariofusco in https://github.com/langchain4j/langchain4j/pull/5678
- fix: ClassCastException when an MCP tool inputSchema contains anyOf alongside type "object" by @subhashpolisetti in https://github.com/langchain4j/langchain4j/pull/5726
- fix: avoid IllegalArgumentException in RetryPolicy jitter when bound is 0 by @Ayomipo18 in https://github.com/langchain4j/langchain4j/pull/5722
- fix: Anthropic honors cache_control on AiMessage and ToolExecutionResultMessage by @joshiste in https://github.com/langchain4j/langchain4j/pull/5729
- fix: Avoid StringIndexOutOfBoundsException when cleaning unclosed code fence in HibernateContentRetriever by @thswlsqls in https://github.com/langchain4j/langchain4j/pull/5732
- fix: Avoid StringIndexOutOfBoundsException when cleaning unclosed code fence in SqlDatabaseContentRetriever by @PSchmitz-Valckenberg in https://github.com/langchain4j/langchain4j/pull/5737
- Follow up on AgentsRegistry implementation by @mariofusco in https://github.com/langchain4j/langchain4j/pull/5719
- Allow a zero-args non-AI agent to also produce an output by @mariofusco in https://github.com/langchain4j/langchain4j/pull/5734
- fix: Paginate ListObjects in TencentCosDocumentLoader to load all obj… by @chensishang in https://github.com/langchain4j/langchain4j/pull/5709
- Surface async agent failures as AgentInvocationException by @subhashpolisetti in https://github.com/langchain4j/langchain4j/pull/5746
- feat(guardrails): add PromptInjectionGuardrail by @ArvindAkula in https://github.com/langchain4j/langchain4j/pull/5619
- fix: Skills: Declare shell tool timeout_seconds as integer and route bad values through typed exception by @thswlsqls in https://github.com/langchain4j/langchain4j/pull/5752
- fix: preserve precision of long/BigInteger/BigDecimal tool arguments by @farzad-sedaghatbin in https://github.com/langchain4j/langchain4j/pull/5755
- Fix race condition in AgentMonitor.onAgentInvocationError with parallel sub-agents by @mariofusco in https://github.com/langchain4j/langchain4j/pull/5759
- Allow to declaratively define beforeCall method on agentic patterns by @mariofusco in https://github.com/langchain4j/langchain4j/pull/5741
- Add jsonSchema() to Byte/Short/BigInteger/BigDecimal output parsers by @angus-guo in https://github.com/langchain4j/langchain4j/pull/5690
- Migrate Tavily and SearchApi web search engines from Retrofit/OkHttp to the HttpClient abstraction by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/5772
- Migrate Nomic, Jina, OVH AI, and Judge0 from Retrofit/OkHttp to the HttpClient abstraction by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/5773
- fix: preserve Ollama-specific parameters set via defaultRequestParameters by @wushiyuanmaimob in https://github.com/langchain4j/langchain4j/pull/5776
- Migrate Cohere, HuggingFace, and Workers AI from Retrofit/OkHttp to the HttpClient abstraction by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/5780
- fix: make enum tool-argument coercion locale-independent by @wushiyuanmaimob in https://github.com/langchain4j/langchain4j/pull/5778
- Allow to dinamically configure the A2A server URL by @mariofusco in https://github.com/langchain4j/langchain4j/pull/5766
- Fix error propagation and tracing in MonitoredExecution by @mariofusco in https://github.com/langchain4j/langchain4j/pull/5782
- Map provider errors to LangChain4j exceptions in Google GenAI, Vertex AI Gemini, Vertex AI Anthropic, Workers AI and Cohere by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/5785
- chore: upgrade google-genai dependency to 1.62.0 by @glaforge in https://github.com/langchain4j/langchain4j/pull/5748
- fix: pass builder skipCreateVectorExtension to PgVectorEmbeddingStore by @subhashpolisetti in https://github.com/langchain4j/langchain4j/pull/5788
- feat(google-genai): add generateContentConfigCustomizer to the chat models by @subhashpolisetti in https://github.com/langchain4j/langchain4j/pull/5646
- Updated Jackson to 2.22.1 by @dliubarskyi in https://github.com/langchain4j/langchain4j/pull/5793
- Apply output function also to supervisor result by @mariofusco in https://github.com/langchain4j/langchain4j/pull/5786
New Contributors
- @nickmaddren made their first contribution in https://github.com/langchain4j/langchain4j/pull/5604
- @jialiuyang made their first contribution in https://github.com/langchain4j/langchain4j/pull/5620
- @lixuanqun made their first contribution in https://github.com/langchain4j/langchain4j/pull/5691
- @Ayomipo18 made their first contribution in https://github.com/langchain4j/langchain4j/pull/5722
- @chensishang made their first contribution in https://github.com/langchain4j/langchain4j/pull/5709
- @wushiyuanmaimob made their first contribution in https://github.com/langchain4j/langchain4j/pull/5776
Full Changelog: https://github.com/langchain4j/langchain4j/compare/1.17.2...1.18.0