1.2.3

What's Changed

• [1.2] Make OpenAI tests more predictable, when working with real LLM by @fedinskiy in #1867
• [1.2] Enable previously disabled tests, since the issues were fixed by @fedinskiy in #1871
• [1.2] Add an option to change the path to source documents for chatbot example by @fedinskiy in #1918
• [1.2] Backports by @fedinskiy in #1949
• Release 1.2.3 by @jmartisk in #2044

Full Changelog: 1.2.2...1.2.3

1.5.0.CR2

What's Changed

• Fix nightly build by @geoand in #2029
• Allow multiple McpClientName annotations on McpClientAuthProvider beans by @jmartisk in #2030
• Exclude JDK HTTP client from Ollama by @geoand in #2032
• Fix config mapping registration in gpu-llama3 runtime module by @orionpapadakis in #2034
• Bump gpu-llama3 to 0.3.3 and curate docs by @orionpapadakis in #2036

Full Changelog: 1.5.0.CR1...1.5.0.CR2

1.5.0.CR1

What's Changed

• pom.xml cleanup by @jmartisk in #1950
• feat: Added support to auto reinitialize session if not found by @frami in #1958
• Bump the Quarkus dependency to 3.27.1 by @jmartisk in #1963
• Guard against empty schema in Gemini requests by @geoand in #1962
• Fix Infinispan related Ecocystem CI issue by @geoand in #1966
• Add an All Configuration Options doc page by @gastaldi in #1967
• Update the filesystem MCP server version by @jmartisk in #1969
• Don't fail the MCP client Dev UI page when there's an enum parameter by @jmartisk in #1971
• Use proper Quarkus version when Ecosystem CI runs ITs by @geoand in #1972
• Add tool guardrails for function calling validation and transformation by @cescoffier in #1973
• Handle unexpected completion of stream from claude api by @ejstuart in #1975
• Use omni-moderation-latest as the OpenAI moderation model by @geoand in #1977
• Bump to LangChain4j 1.9.1 by @mariofusco in #1974
• WebSocket MCP transport integration by @jmartisk in #1980
• Set default event name to message in MCP SSE client by @yblazart in #1986
• Refactor guardrail metrics by @cescoffier in #1984
• Update Azure docs to link to advanced use cases by @hubert123490 in #1982
• Use Quarkus 3.27 in samples by @geoand in #1988
• Relax Chroma test to not compare the vectors by @cescoffier in #1990
• Add verfication for gen_ai metrics in tests by @fedinskiy in #1991
• Add observability and metrics support for tool guardrails by @cescoffier in #1993
• Anthropic stream improvements, add streaming handle implementation by @ejstuart in #1997
• Ignore null SSE mesages by @kevinferrare-1a in #2001
• Ensure that Guardrails work in native mode by @geoand in #2000
• Temporarily hard-code the versions of various testcontainers by @geoand in #2003
• Ensure that Guardrails work in native mode - again by @geoand in #2005
• Document embedding index usage and recommendations by @myfear in #2006
• Ensure that when returning a Multi, everything is run on in a blocking thread by @geoand in #2012
• Introduce allowContinuousForcedToolCalling into @RegisterAiService by @geoand in #2013
• Implement removeIds and removeAll by @karesti in #2015
• Align with upstream LangChain4j error handling by @geoand in #2014
• Re-implement evaluation framework by @cescoffier in #2018
• Update GPULlama3 version by @orionpapadakis in #2020
• Add optional preview profile for GPU Llama3 by @geoand in #2024
• Add Guardrails functionality and tests to chatbot sample by @fedinskiy in #2007

New Contributors

• @frami made their first contribution in #1958
• @yblazart made their first contribution in #1986
• @hubert123490 made their first contribution in #1982
• @kevinferrare-1a made their first contribution in #2001

Full Changelog: 1.4.2...1.5.0.CR1

1.4.2

What's Changed

• Use a separate moderation model for moderation requests by @fedinskiy in #1942
• Ignore empty SSE messages, accept \r\n and \n for line separators by @jmartisk in #1944
• Fix seed parameter overwriting max tokens for AzureOpenAi by @JanHuege in #1946

Full Changelog: 1.4.1...1.4.2

1.4.1

What's Changed

• Fix GPU resources utilization by @orionpapadakis in #1937
• Refactor multi-modality handling by @geoand in #1940

Full Changelog: 1.4.0...1.4.1

1.4.0

What's Changed

• Switch to the upstream ChromaEmbeddingStore impl, support V2 API by @jmartisk in #1915
• Introduce seed parameter for azure openai chat models by @JanHuege in #1916
• Automatically sort segments for embedding with gemini into batches of 100 by @jmartisk in #1922
• Add tracing configuration about tool invocation by @geoand in #1921
• Fire intermediate chat event at the correct spot by @ejstuart in #1923
• Temporarily disable a failing MCP registry client test by @jmartisk in #1924
• Support custom headers for OpenAI module by @geoand in #1929
• Add google search capability to quarkus vertex ai by @ramesh-spendflo in #1925
• Add doc page for GPULlama3 by @orionpapadakis in #1930
• Scope google credentials to cloud platform by default by @ramesh-spendflo in #1926
• Vertex AI Response deserialization - Ignore unknown attributes by @ramesh-spendflo in #1931

Full Changelog: 1.4.0.CR2...1.4.0

1.4.0.CR2

What's Changed

• Bump quarkus-neo4j.version from 5.6.0 to 5.7.0 by @dependabot[bot] in #1776
• Fix build time validation for @SupervisorAgent related agentic system by @geoand in 7119089

Full Changelog: 1.4.0.CR1...1.4.0.CR2

1.4.0.CR1

What's Changed

• Make McpRootsStdioTransportTest work in the Platform CI by @jmartisk in #1855
• Introduce compile time safety to the agentic module by @geoand in #1858
• Allow some integration tests to run with platform bits by @fedinskiy in #1859
• Remove warnings and possible failures related to quarkus-plugin by @fedinskiy in #1861
• Make OpenAI tests more predictable, when working with real LLM by @fedinskiy in #1866
• Enable previously disabled tests, since the issues were fixed by @fedinskiy in #1868
• Add intermediate chat event by @ejstuart in #1876
• Introduce support for Video in AI services by @geoand in #1881
• Add interleaved thinking and cache prompting by @ejstuart in #1885
• Use upstream LangChain4j mixins for Jackson codec by @ejstuart in #1884
• Remove Quarkus-specific auditing capabilities by @edeandrea in #1886
• Introduce reasoningEffort and serviceTier config support for OpenAI by @geoand in #1887
• Bump to LangChain4j 1.8.0 by @mariofusco in #1888
• Add an option to change the path to source documents for chatbot example by @fedinskiy in #1890
• Add RAG test using QueryRouter and ContentInjector by @mariofusco in #1902
• added mcpClientAuthProvider to Builder by @KaiSuchomel in #1892
• Add GPULlama3.java as model provider to run on GPUs by @mikepapadim in #1793
• Fix exception rethrow in tool executor by @ejstuart in #1904
• Introduce type safety to outputKey handling by @geoand in #1903
• Add streaming support for GPULlama3.java integration. by @orionpapadakis in #1906
• Make sure that @ModelName works for streaming responses in AI services by @geoand in #1905

New Contributors

• @mikepapadim made their first contribution in #1793
• @orionpapadakis made their first contribution in #1906

Full Changelog: 1.3.1...1.4.0.CR1

1.2.2

What's Changed

• [1.2] Allow some integration tests to run with platform bits by @fedinskiy in #1860
• [1.2] Remove warnings and possible failures related to quarkus-plugin by @fedinskiy in #1862
• [1.2] Movie sample backports by @fedinskiy in #1864

Full Changelog: 1.2.1...1.2.2

1.3.1

What's Changed

• fix: disable topK for anthropic streaming model by @Wal8800 in #1834
• Disable McpRegistryClientTest for now by @jmartisk in #1835
• Trace tool calls by @sgdesmet in #1829
• Use more precise guide link for Ollama by @holly-cummins in #1836
• InvocationParameters and ToolExecution.resultObject() support by @patriot1burke in #1837
• enhance: allow anthropic recorder to add chat model listener by @Wal8800 in #1840
• Support MicroProfile Health Checks for MCP Clients by @sberyozkin in #1648
• Add missing Gemini FinishReason values by @sberyozkin in #1841
• Deprecate existing auditing events. by @edeandrea in #1843
• Fix tool call instrumentation regression by @sgdesmet in #1848
• Fix the default value of Tavily's log-responses by @jmartisk in #1850
• Default empty list for Tavily include and exclude domains by @Joshua-Hell in #1851
• Implement upstream LC4j observability capabilities by @edeandrea in #1846
• Add check for tool providers when streaming by @ejstuart in #1853
• Release 1.3.1 by @geoand in #1854

New Contributors

• @Wal8800 made their first contribution in #1834
• @sgdesmet made their first contribution in #1829
• @Joshua-Hell made their first contribution in #1851

Full Changelog: 1.3.0...1.3.1