0.22.0

What's Changed

• Properly warn about missing javac -parameters flag by @geoand in #1113
• Bump quarkus-neo4j.version from 4.4.0 to 5.0.0 by @dependabot in #1115
• Allow rewriting of user messages from input guardrails by @mariofusco in #1083
• Bump io.quarkiverse.wiremock:quarkus-wiremock-test from 1.3.3 to 1.4.0 by @dependabot in #1116
• Integrate Mistral AI moderation models by @jmartisk in #1117

Full Changelog: 0.22.0.CR2...0.22.0

0.22.0.CR2

What's Changed

• Update version README during release by @geoand in #1105
• Bump com.github.tjake:jlama-core from 0.8.2 to 0.8.3 by @dependabot in #1107
• Initial implementation of the response augmenter idea by @cescoffier in #1106
• Enable resolution of AI services by bean name by @aldettinger in #1110
• Release 0.22.0.CR2 by @geoand in #1111
• Release 0.22.0.CR2 by @geoand in #1112

Full Changelog: 0.22.0.CR1...0.22.0.CR2

0.22.0.CR1

What's Changed

• Bump org.apache.maven.plugins:maven-surefire-plugin from 3.5.0 to 3.5.1 by @dependabot in #976
• Register all models from configuration as beans regardless of the existance of injection point by @manovotn in #1061
• Prepare the arrival of the bot by @gsmet in #1062
• Show how OIDC ModelAuthProvider can be used with Azure OpenAI by @sberyozkin in #1056
• Prioritize Jackson over Jsonb in extension that use quarkus-rest-client-jackson by @geoand in #1067
• Make Llama3.2 the default for Ollama by @geoand in #1066
• Introduce a global temperature property by @geoand in #1068
• Properly handle Ollama streaming by @geoand in #1070
• Fix MismatchedInputException in Ollama streaming by @jpohlmeyer in #1072
• Provide an abstract output guardrails for json data extraction by @mariofusco in #1060
• Use latest version of the Llama3.java code by @geoand in #1076
• Easy RAG extension: don't depend on the upstream Easy RAG module by @jmartisk in #1078
• Add AbstractJsonExtractorOutputGuardrail to guardrails docs by @mariofusco in #1082
• Add a Secure SQL ChatBot demo by @sberyozkin in #1073
• Update Quarkus to 3.15.2 and update documentation by @gsmet in #1084
• Move JsonGuardrailsUtilsTest to the deployment module by @jmartisk in #1086
• Detect when models to not support tools when using streaming by @cescoffier in #1085
• Drop references to resteasy-reactive artifacts by @gsmet in #1088
• Bump to Jlama 0.8.2 by @geoand in #1090
• Prevent Jlama inference to block vertx event loop by @mariofusco in #1092
• When creating fastJar don't try to include models more than once by @edeandrea in #1093
• Remove duplicate invocation from handleMessageStop. by @dennysfredericci in #1094
• static keyword is not allowed there. by @lordofthejars in #1098
• Upgrade to LangChain4j 0.36.2 by @jmartisk in #1097
• Bump io.smallrye.certs:smallrye-certificate-generator-junit5 from 0.8.1 to 0.9.2 by @dependabot in #1101
• Add chatbot codestart by @iocanel in #1103
• Migrate to the JsonSchemaElement API by @edeandrea in #1100

New Contributors

• @manovotn made their first contribution in #1061

Full Changelog: 0.21.0...0.22.0.CR1

0.21.0

What's Changed

• Moving samples updates to release-prepare.yml by @gastaldi in #1035
• Update Llama3 and Jlama ITs in CI by @geoand in #1037
• Add prompt template and variables to input / output guardrails by @dennysfredericci in #992
• Allow to rewrite LLM result in an OutputGuardrail by @mariofusco in #1021
• Adjusted method signature mapping to json schema, to allow collections in tool arguments by @Tarjei400 in #1039
• Bring ModelAuthProvider support to the OpenAI extension by @geoand in #1041
• Bump jlama to version 0.8.1 by @mariofusco in #1042
• Improve model download progress logging by @geoand in #1045
• Fix warning about missing model-id in Ollama by @geoand in #1048
• Add basic docs for Llama3.java by @geoand in #1049
• Allow using generic ModelAuthProvider with named models by @sberyozkin in #1050
• Document rewriting output guardrails by @mariofusco in #1052
• Enable Tools to Define Execution Model by @cescoffier in #1023
• Allow for Ollama configuration to make quarkus.langchain4j.foo.chat-model.provider unnecessary by @geoand in #1057

New Contributors

• @Tarjei400 made their first contribution in #1039

Full Changelog: 0.21.0.CR4...0.21.0

0.21.0.CR4

What's Changed

• Switch config doc generation to the new plugin introduced in 3.14 by @gsmet in #1017
• Fix incorrect model-id parameter in WatsonxRecorder by @andreadimaio in #1024
• Improve web-search sample by @jmartisk in #1025
• Add description for neo4j extension by @holly-cummins in #1026
• Easy-rag: allow to specify minScore through configuration by @edeandrea in #1027
• Add Jlama documentation by @mariofusco in #1018
• Add missing llama3-java module by @geoand in #1028

Full Changelog: 0.21.0.CR1...0.21.0.CR4

0.21.0.CR1

What's Changed

• Remove BAM module by @andreadimaio in #968
• Add support for BGE 1.5 (regular and quantized) in-process embedding by @cescoffier in #966
• Refactor CI workflows (only pull requests) by @cescoffier in #972
• Bump org.apache.maven.plugins:maven-failsafe-plugin from 3.5.0 to 3.5.1 by @dependabot in #977
• Add clarification about WebSearchRetrievalAugmentor by @jmartisk in #982
• Enable /chat and /chat_stream in watsonx.ai by @andreadimaio in #981
• Split and Reuse the Release workflow by @gastaldi in #986
• Anthropic chat model not available when enabled by @dennysfredericci in #985
• Release workflow names are inverted by @gastaldi in #987
• Enable ScoringModel in watsonx.ai by @andreadimaio in #994
• Use custom integration with Jlama by @geoand in #998
• Enable watsonx.ai to process ImageContent in UserMessage by @andreadimaio in #1000
• Move inner classes to top level classes for API concision by @holly-cummins in #1003
• Allow including configured models in the built artifact for Jlama by @geoand in #1004
• [FEATURE] ToolProvider | Select tools dynamically on incoming message by @MiggiV2 in #989
• Add optional logging to jlama requests and responses by @mariofusco in #999
• Add support for TLS configuration name by @geoand in #1006
• Don't include tools from removed beans in the ToolsMetadata by @jmartisk in #1010
• Properly support TokenStream return type in AiService by @geoand in #1005
• Implement support for output guardrail on streamed responses by @cescoffier in #1011
• Introduce llama3-java module by @geoand in #1008
• Update watsonx module: Add Text Extraction APIs, Configuration changes, and Pretty-Print logging by @andreadimaio in #1013

New Contributors

• @dennysfredericci made their first contribution in #985
• @MiggiV2 made their first contribution in #989

Full Changelog: 0.20.3...0.21.0.CR1

0.20.3

What's Changed

• Fix local embedding tokenizer location by @cescoffier in #960
• Shadows api key from tavily logs by @lordofthejars in #961
• Bump io.quarkiverse:quarkiverse-parent from 16 to 18 by @dependabot in #959
• Move OpenWebUI Dev UI Actions to deployment classpath by @phillip-kruger in #867
• Fix ONNX Runtime Execution in Native Executable by @cescoffier in #964

Full Changelog: 0.20.1...0.20.3

0.20.1

What's Changed

• Extend the chatbot sample to use streaming by @cescoffier in #953
• Upgrade to Quarkus Antora - fix issues on Arm Macs by @ppalaga in #913
• Avoid duplicating info for AiService implementation constructors by @geoand in #955

Full Changelog: 0.20.0...0.20.1

0.20.0

What's Changed

• Fix issue with Image generation in AiService by @geoand in #951

Full Changelog: 0.20.0.CR1...0.20.0

0.20.0.CR1

What's Changed

• Bump to LangChain4j 0.35 by @geoand in #948

Full Changelog: 0.19.0...0.20.0.CR1