3.0.0.0

Version 3.0.0.0 Release Notes

Compatible with OpenSearch 3.0.0.0

Breaking Changes

• Use _list/indices API instead of _cat/index API in CatIndexTool (#3243)
• Deprecate the restful API of batch ingestion (#3688)

Features

• Onboard MCP (#3721)
• Plan, Execute and Reflect Agent Type (#3716)
• Support custom prompts from user (#3731)
• Support MCP server in OpenSearch (#3781)

Enhancements

• Support sentence highlighting QA model (#3600)
• Add parser for ModelTensorOutput and ModelTensors (#3658)
• Function calling for openai v1, bedrock claude and deepseek (#3712)
• Update highlighting model translator to adapt new model (#3699)
• Implement async mode in agent execution (#3714)

Bug Fixes

• Fix building error due to a breaking change from core (#3617)
• Fixing the circuit breaker issue for remote model (#3652)
• Fix compilation error (#3667)
• Revert CI workflow changes (#3674)
• Fix config index masterkey pull up for multi-tenancy (#3700)
• Fix error message when input map and output map length not match (#3730)
• Agent Framework: Handle model response when toolUse is not accompanied by text (#3755)
• Allow user to control react agent max_interations value to prevent empty response (#3756)
• Agent framework: Fix SearchIndexTool to parse special floating point values and NaN (#3754)
• Directly return Response objects from metadata client responses (#3768)
• Remove opensearch-ml-2.4.0.0.zip file that was added by random mistake (#3763)
• Replace null GetResponse with valid response and not exists (#3759)
• Fix ListIndexTool and SearchIndexTool (#3720)
• Support MCP session management (#3803)
• Support customized message endpoint and addressing comments (#3810)
• Excluding circuit breaker for Agent (#3814)

Maintenance

• Update CB setting to 100 to bypass memory check (#3627)
• Use model type to check local or remote model (#3597)
• Fixing security integ test (#3646)
• Remove forcing log4j version to 2.24.2 (#3647)
• Improve test coverage for MLHttpClientFactory.java (#3644)
• Improve test coverage for MLEngineClassLoader class (#3679)
• Typo: MLTaskDispatcher _cluster/settings api (#3694)
• Add more logs to troubleshoot flaky test (#3543)
• Add package for security test (#3698)
• Add sdk implementation to the connector search (#3704)
• Sdk client implementation for search connector, model group and task (#3707)
• Add Feature Flag for MCP connectors Feature ((#3738)
• Support phasing off SecurityManager usage in favor of Java Agent (#3729)
• Add java-agent gradle plugin (#3727)

Documentation

• Add tutorial for RAG of openai and bedrock (#2975)
• Fix template query link (#3612)
• Add standard blueprint for vector search (#3659)
• Add blueprint for Claude 3.7 on Bedrock (#3584)
• Add standard blueprint for azure embedding ada2 (#3725)

3.0.0.0-beta1

Version 3.0.0.0-beta1 Release Notes

Compatible with OpenSearch 3.0.0.0-beta1

Breaking Changes

• Deprecate the restful API of batch ingestion (#3688)[https://github.com//pull/3688]

Enhancements

• Add parser for ModelTensorOutput and ModelTensors (#3658)[https://github.com//pull/3658]
• Function calling for openai v1, bedrock claude and deepseek (#3712)[https://github.com//pull/3712]
• Update highlighting model translator to adapt new model (#3699)[https://github.com//pull/3699]
• Plan, Execute and Reflect Agent Type (#3716)[https://github.com//pull/3716]
• Implement async mode in agent execution (#3714)[https://github.com//pull/3714]

Bug Fixes

• Fixing the circuit breaker issue for remote model (#3652)[https://github.com//pull/3652]
• Fix compilation error (#3667)[https://github.com//pull/3667]
• Revert CI workflow changes (#3674)[https://github.com//pull/3674]
• Fix config index masterkey pull up for multi-tenancy (#3700)[https://github.com//pull/3700]

Maintenance

• Remove forcing log4j version to 2.24.2 (#3647)[https://github.com//pull/3647]
• Improve test coverage for MLHttpClientFactory.java (#3644)[https://github.com//pull/3644]
• Improve test coverage for MLEngineClassLoader class (#3679)[https://github.com//pull/3679]
• Typo: MLTaskDispatcher _cluster/settings api (#3694)[https://github.com//pull/3694]
• Add more logs to toubleshot flaky test (#3543)[https://github.com//pull/3543]
• Add package for security test (#3698)[https://github.com//pull/3698]
• Add sdk implementation to the connector search (#3704)[https://github.com//pull/3704]
• Sdk client implementation for search connector, model group and task (#3707)[https://github.com//pull/3707]

Documentation

• Add standard blueprint for vector search (#3659)[https://github.com//pull/3659]
• Add blueprint for Claude 3.7 on Bedrock (#3584)[https://github.com//pull/3584]

3.0.0.0-alpha1

Version 3.0.0.0-alpha1 Release Notes

Compatible with OpenSearch 3.0.0.0-alpha1

Breaking Changes

• Use _list/indices API instead of _cat/index API in CatIndexTool (#3243)[https://github.com//pull/3243]

Bug Fixes

• Fix building error due to a breaking change from core (#3617)[https://github.com//pull/3617]

Documentation

• Add tutorial for RAG of openai and bedrock (#2975)[https://github.com//pull/2975]
• Fix template query link (#3612)[https://github.com//pull/3612]

Enhancements

• Support sentence highlighting QA model (#3600)[https://github.com//pull/3600]

Maintenance

• Update CB setting to 100 to bypass memory check (#3627)[https://github.com//pull/3627]
• Use model type to check local or remote model (#3597)[https://github.com//pull/3597]
• Fixing security integ test (#3646)[https://github.com//pull/3646]

2.19.1.0

Version 2.19.1.0 Release Notes

Compatible with OpenSearch 2.19.1

Bug Fixes

• fix ignoreFailure flag in ml inference search processors (#3570)[https://github.com//pull/3570]
• updating bulk update to use sdkclient (#3546)[https://github.com//pull/3546]
• Add interface for no postprocessing function case (#3553)[https://github.com//pull/3553]
• add edge case for models that are marked as not found in cache (#3523)[https://github.com//pull/3523]
• applying sdkclient changes to config index (#3521)[https://github.com//pull/3521]
• remainig sdk client changes for search (#3522)[https://github.com//pull/3522]
• adding tenantId for inline connector in model registration (#3531)[https://github.com//pull/3531]
• Remove the integ test case cause cohere classify is deprecated in its llm service (#3513)[https://github.com//pull/3513]
• fix null decrypted credential while getting batch predict job status (#3518)[https://github.com//pull/3518]

2.19.0.0

Version 2.19.0.0 Release Notes

Compatible with OpenSearch 2.19.0

Enhancements

• adding multi-modal pre-processor for cohere (#3219)[https://github.com//pull/3219]
• [Enhancement] Fetch system index mappings from json file instead of string constants (#3153)[https://github.com//pull/3153]
• Retrieve remote model id from registration response in IT to avoid flaky (#3244)[https://github.com//pull/3244]
• [Refactor] Remove bloat due to unnecessary setup in test and add retry for potential flaky behavior due to timeout (#3259)[https://github.com//pull/3259/files]
• [Enhancement] Enhance validation for create connector API (#3260)[https://github.com//pull/3260]
• Add application_type to ConversationMeta; update tests (#3282)[https://github.com//pull/3282]
• Enhance Message and Memory API Validation and storage (#3283)[https://github.com//pull/3283]
• Use Adagrad optimiser for Linear regression by default (#3291)[https://github.com//pull/3291/files]
• [Enhancement] Add schema validation and placeholders to index mappings (#3240)[https://github.com//pull/3240]
• add action input as parameters for tool execution in conversational agent (#3200)[https://github.com//pull/3200]
• Remove ignore decorator for testCohereClassifyModel (#3324)[https://github.com//pull/3324]
• refactor: modifying log levels and adding more logs to display error details (#3337)[https://github.com//pull/3337]
• Primary setup for Multi-tenancy (#3307)[https://github.com//pull/3307]
• apply multi-tenancy and sdk client in Connector (Create + Get + Delete) (#3382)[https://github.com//pull/3382]
• adding multi-tenancy + sdk client related changes to model, model group and connector update (#3399)[https://github.com//pull/3399]
• applying multi-tenancy to task apis, deploy, predict apis (#3416)[https://github.com//pull/3416]
• adding tenantID to the request + undeploy request (#3425)[https://github.com//pull/3425]
• Check before delete (#3209)[https://github.com//pull/3209]
• multi-tenancy + sdk client related changes in agents (#3432)[https://github.com//pull/3432]
• Introduce Ml Inference Search Request Extension (#3284)[https://github.com//pull/3284]
• Cherry-pick BWC fix for system prompt and user instructions (#3437)[https://github.com//pull/3437]
• Add deepseek as a trusted endpoint. (#3440)[https://github.com//pull/3440]
• applying multi-tenancy in search [model, model group, agent, connector] (#3433)[https://github.com//pull/3433]
• Added amazon rekognition as a trust endpoint (#3419)[https://github.com//pull/3419]
• adding multi-tenancy to config api and master key related changes (#3439)[https://github.com//pull/3439]
• Undeploy models with no WorkerNodes (#3380)[https://github.com//pull/3380]
• support batch task management by periodically bolling the remote task via a cron job (#3421)[https://github.com//pull/3421]
• Add pre and post process functions for Bedrock Rerank API #3254 (#3339)[https://github.com//pull/3339]
• [Backport 2.19] [BACKPORT 2.x] applying multi-tenancy in search [model, model group, agent, connector] (#3433) (#3469)[https://github.com//pull/3469]
• [Backport 2.19] fixing connector validation (#3471)[https://github.com//pull/3471]
• [BACKPORT 2.x] applying multi-tenancy in search [model, model group, agent, connector] (#3433) (#3443) (#3469)[https://github.com//pull/3469]

Bug Fixes

• getFirst is not allowed in java 17 (#3226)[https://github.com//pull/3226]
• Fix: ml/engine/utils/FileUtils casts long file length to int incorrectly (#3198)[https://github.com//pull/3198]
• fix for sync up job not working in 2.17 when upgraded from previous versions (#3241)[https://github.com//pull/3241]
• Fix remote model with embedding input issue (#3289)[https://github.com//pull/3289]
• Adds preset contentRegistry for IngestProcessors (#3281)[https://github.com//pull/3281]
• Revert "Add application_type to ConversationMeta; update tests (#3282)" (#3315)[https://github.com//pull/3315]
• Revert "Filter out remote model auto redeployment (#2976)" and related commits (#3104, #3214) (#3368)[https://github.com//pull/3368]
• Fix JsonGenerationException error in Local Sample Calculator and Anomaly Localization Execution Response (#3434)[https://github.com//pull/3434]
• [Backport 2.19] Fix guardrail it for 2.19 (#3468)[https://github.com//pull/3468]
• addressing client changes due to adding tenantId in the apis (#3474) (#3480)[https://github.com//pull/3480]

Maintenance

• Bump guava version to 32.1.3 (#3300)[https://github.com//pull/3300]
• Update Gradle to 8.11.1 (#3139)[https://github.com//pull/3139/files]
• Force version 3.29.0 of org.eclipse.core.runtime to mitigate CVE vulnerabilities (#3313)[https://github.com//pull/3313]
• Upgraded software.amazon.awssdk from 2.25.40 to 2.29.0 to address CVE… (#3320)[https://github.com//pull/3320]
• Adding back Mingshi as Maintainer. (#3367)[https://github.com//pull/3367]
• updating sdk client version (#3392)[https://github.com//pull/3392]
• downgrading codecov action (#3409) (#3410)[https://github.com//pull/3410]
• fix CVE from ai.djl dependency (#3478) (#3482)[https://github.com//pull/3482]

Infrastructure

• Enable custom start commands and options to resolve GHA issues (#3223)[https://github.com//pull/3223]
• Add Spotless Check to maintain consistency (#3386)[https://github.com//pull/3386]
• Add runs-on field to Spotless Check step in CI (#3400)[https://github.com//pull/3400]
• Checkout code from pull request head for spotless (#3422)[https://github.com//pull/3422]
• Fixes spotless on Java 11 (#3449)[https://github.com//pull/3449]
• add spotless to all build.gradle files (#3453)[https://github.com//pull/3453]
• Fixes Two Flaky IT classes RestMLGuardrailsIT & ToolIntegrationWithLLMTest (#3253)[https://github.com//pull/3253]
• Improve test coverage for RemoteModel.java (#3205)[https://github.com//pull/3205]
• Revert Text Block changes from "Enhance validation for create connector API" #3260 (#3329)[https://github.com//pull/3329]

Documentation

• Fix: typo in MLAlgoParams (#3195)[https://github.com//pull/3195/files]
• Add tutorial for bge-reranker-m3-v2, multilingual cross-encoder model on SageMaker (#2848)[https://github.com/opensearch-project/ml-commons/commit/https://github.com//pull/2848]
• [Backport main] adding blue print doc for cohere multi-modal model (#3232)[https://github.com//pull/3232]
• Tutorial for using Asymmetric models (#3258)[https://github.com//pull/3258]
• add tutorials for cross encoder models on Amazon Bedrock (#3278)[https://github.com//pull/3278]
• fix typo (#3234)[https://github.com//pull/3234]
• Tutorial for ml inference with cohere rerank model (#3398)[https://github.com//pull/3398]
• Add DeepSeek connector blueprint (#3436)[https://github.com//pull/3436]
• fix post_process_function on rerank_pipeline_with_bge-rerank-m3-v2_model_deployed_on_Sagemaker.md (#3296)[https://github.com//pull/3296]

2.18.0.0

Version 2.18.0.0 Release Notes

Compatible with OpenSearch 2.18.0

Enhancements

• Filter out remote model auto redeployment (#2976)[https://github.com//pull/2976]
• Allow llmQuestion to be optional when llmMessages is used. (#3072)[https://github.com//pull/3072]
• Enhance batch job task management by adding default action types (#3080)[https://github.com//pull/3080]
• use connector credential in offline batch ingestion (#2989)[https://github.com//pull/2989]
• change to model group access for batch job task APIs (#3098)[https://github.com//pull/3098]
• add rate limiting for offline batch jobs, set default bulk size to 500 (#3116)[https://github.com//pull/3116]
• Support ML Inference Search Processor Writing to Search Extension (#3061)[https://github.com//pull/3061]
• Enable pass query string to input_map in ml inference search response processor (#2899)[https://github.com//pull/2899]
• add config field in MLToolSpec for static parameters (#2977)[https://github.com//pull/2977]
• Add textract and comprehend url to trusted enpoints (#3154)[https://github.com//pull/3154]

Bug Fixes

• Fix ml inference ingest processor always return list using JsonPath (#2985)[https://github.com//pull/2985]
• populate time fields for connectors on return (#2922)[https://github.com//pull/2922]
• Fix for rag processor throwing NPE when optional parameters are not provided (#3057)[https://github.com//pull/3057]
• Fix PR #2976 bug due to missing adding function_name and algorithm in querying models (#3104)[https://github.com//pull/3104]
• Gracefully handle error when generative_qa_parameters is not provided (#3100)[https://github.com//pull/3100]
• Fix error log to show the right agent type (#2809)[https://github.com//pull/2809]
• fix model stuck in deploying state during node crash/cluster restart (#3137)[https://github.com//pull/3137]
• increase the wait timeout to fetch the master key (#3151)[https://github.com//pull/3151]
• Handle BWC for bedrock converse API (#3173)[https://github.com//pull/3173]

Maintenance

• Bump protobuf version to 3.25.5 to patch potential DOS (#3083)[https://github.com//pull/3083]
• removing api keys from the integ test log (#3112)[https://github.com//pull/3112]
• allowing backport prs to skip approval (#3132)[https://github.com//pull/3132]
• updating the approval requirement (#3148)[https://github.com//pull/3148]
• unblocking the integ test pipeline for release (#3159)[https://github.com//pull/3159]

Infrastructure

• Support index.auto_expand_replicas 0-all for .plugins-ml-config (#3017)[https://github.com//pull/3017]
• Add Test Env Require Approval Action (#3005)[https://github.com//pull/3005]
• upgrading upload artifact to v4 (#3162)[https://github.com//pull/3162]
• Bump actions/download-artifact from 3 to 4.1.7 in /.github/workflows (#2881)[https://github.com//pull/2881]

Documentation

• add tutorial for cross-account model invocation on amazon managed cluster (#3064)[https://github.com//pull/3064]
• support role temporary credential in connector tutorial (#3058)[https://github.com//pull/3058]
• connector blueprint for amazon bedrock converse (#2960)[https://github.com//pull/2960]
• Updates dev guide to inform the workflow approval step (#3062)[https://github.com//pull/3062]
• tune titan embedding model blueprint for v2 (#3094)[https://github.com//pull/3094]
• Add bedrock multimodal build-in function usage example in doc (#3073)[https://github.com//pull/3073]

2.17.0.0

Version 2.17.0.0 Release Notes

Compatible with OpenSearch 2.17.0

Features

• Offline batch ingestion API actions and data ingesters (#2844)[https://github.com//pull/2844]
• Support get batch transform job status in get task API (#2825)[https://github.com//pull/2825]

Enhancements

• Adding additional info for memory metadata (#2750)[https://github.com//pull/2750]
• Support skip_validating_missing_parameters in connector (#2830)[https://github.com//pull/2830]
• Support one_to_one in ML Inference Search Response Processor (#2801)[https://github.com//pull/2801]
• Expose ML Config API (#2850)[https://github.com//pull/2850]

Bug Fixes

• Fix delete local model twice quickly get 500 response issue (#2806)[https://github.com//pull/2806]
• Fix cohere model input interface cannot validate cohere input issue (#2847)[https://github.com//pull/2847]
• Add processed function for remote inference input dataset parameters to convert it back to its original datatype (#2852)[https://github.com//pull/2852]
• Use local_regex as default type for guardrails (#2853)[https://github.com//pull/2853]
• Agent execution error in json format (#2858)[https://github.com//pull/2858]
• Fix custom prompt substitute with List issue in ml inference search response processor (#2871)[https://github.com//pull/2871]
• Fix breaking changes in config index fields (#2882)[https://github.com//pull/2882]
• Output only old fields in get config API (#2892)[https://github.com//pull/2892]
• Fix http dependency in CancelBatchJobTransportAction (#2898)[https://github.com//pull/2898]

Maintenance

• Applying spotless to common module (#2815)[https://github.com//pull/2815]
• Fix Cohere test (#2831)[https://github.com//pull/2831]

Infrastructure

• Test: recover search index tool it in multi node cluster (#2407)[https://github.com//pull/2407]

Documentation

• Add tutorial for Bedrock Guardrails (#2695)[https://github.com//pull/2695]

Refactoring

• Code refactor not to occur nullpointer exception (#2816)[https://github.com//pull/2816]

2.16.0.0

Version 2.16.0.0 Release Notes

Compatible with OpenSearch 2.16.0

Features

• Add initial MLInferenceSearchResponseProcessor (#2688)[https://github.com//pull/2688]
• Add initial search request inference processor (#2731)[https://github.com//pull/2731]
• Add Batch Prediction Mode in the Connector Framework for batch inference (#2661)[https://github.com//pull/2661]

Enhancements

• Automated model interface generation on aws llms (#2689)[https://github.com//pull/2689]
• Increase execute thread pool size (#2691)[https://github.com//pull/2691]
• Add multi modal default preprocess function (#2500)[https://github.com//pull/2500]
• Add model input validation for local models in ml processor (#2610)[https://github.com//pull/2610]
• Removing experimental from the Conversation memory feature(#2592)[https://github.com//pull/2592]
• Pass all parameters including chat_history to run tools (#2714)[https://github.com//pull/2714]
• Feat: add bedrock runtime agent for knowledge base (#2651)[https://github.com//pull/2651]
• change disk circuit breaker to cluster settings (#2634)[https://github.com//pull/2634]

Bug Fixes

• Add stashcontext to connector getter (#2742)[https://github.com//pull/2742]
• Excluding remote models from max node per node setting (#2732)[https://github.com//pull/2732]
• Add logging for throttling and guardrail in connector (#2725)[https://github.com//pull/2725]
• Add acknowledge check for index creation in missing places (#2715)[https://github.com//pull/2715]
• Update config index mappings to use correct field types (#2710)[https://github.com//pull/2710]
• Fix yaml test issue (#2700)[https://github.com//pull/2700]
• Fix MLModelTool returns null if the response of LLM is a pure json object (#2675)[https://github.com//pull/2675]
• Bump ml config index schema version (#2656)[https://github.com//pull/2656]
• Fix final answer with extra meaningless symbol (#2676)[https://github.com//pull/2676]
• Add XContentType to wrap the CreateIndexRequest mappings in _doc key to fix v1 templates issue (#2759)[https://github.com//pull/2759]
• Remove ignoreFailure and fix JsonArray Parsing Issue (#2770)[https://github.com//pull/2770]
• Merge the existing parameters when updating connectors (#2784)[https://github.com//pull/2784]

Maintenance

• Upgrade djl version to 0.28.0 (#2578)[https://github.com//pull/2578]
• Register system index descriptors through SystemIndexPlugin.getSystemIndexDescriptors (#2586)[https://github.com//pull/2586]

Infrastructure

• Enable tests with mockStatic in MLEngineTest (#2582)[https://github.com//pull/2582]
• Fix GA workflow that publishes Apache Maven artifacts (#2625)[https://github.com//pull/2625]
• Temp use of older nodejs version before moving to Almalinux8 (#2628)[https://github.com//pull/2628]
• Add more logs for automated model interface creation (#2778)[https://github.com//pull/2778]

Documentation

• Add amazon textract blueprint (#2562)[https://github.com//pull/2562]
• Make all Bedrock model blueprints in a tidier format (#2642)[https://github.com//pull/2642]
• Fix remote inference blueprints (#2692)[https://github.com//pull/2692]
• Add connector blueprint for cohere embedding models in bedrock (#2667)[https://github.com//pull/2667]
• Update tutorials for caching secrets for non-aws models (#2637)[https://github.com//pull/2637]
• Add tutuorial for cross-encoder model on sagemaker (#2607)[https://github.com//pull/2607]
• Add offline batch inference connector blueprints (#2768)[https://github.com//pull/2768]

Refactoring

• Change multimodal connector name to bedrock multimodal connector (#2672)[https://github.com//pull/2672]

2.15.0.0

Version 2.15.0.0 Release Notes

Compatible with OpenSearch 2.15.0.0

Features

• Add connector tool (#2516)
• guardrails model support (#2491)

Enhancements

• hanlde the throttling error in the response header (#2442)
• Implementing retry for remote connector to mitigate throttling issue (#2462)
• ml inference ingest processor support for local models (#2508)
• add setting to allow private IP (#2534)
• add IMMEDIATE refresh policy (#2541)

Bug Fixes

• fix memory CB bugs and upgrade UTs to compatible with core changes (#2469)
• fix error of ML inference processor in foreach processor (#2474)
• fix error message with unwrapping the root cause (#2458)
• adding immediate refresh to delete model group request (#2514)
• Fix model still deployed after calling undeploy API (#2510)
• Fix bedrock embedding generation issue (#2495)
• Fix init encryption master key (#2554)

Documentation

• add a connector blueprint for Amazon Comprehend APIs (#2470)
• add titan embeeding v2 to blueprint (#2480)
• tutorial: generate embedding for arrays of object (#2477)
• Small fix in blueprint docs (#2501)
• Titan Embedding Connector Blueprint content referenced by users of OpenSearch 2.11 version (#2519)

Infrastructure

• add IT for flow agent with CatIndexTool (#2425)
• Remove strict version dependency to compile minimum compatible version (#2486)
• add IT flow agent with search index tool (#2460)
• fix flaky IT (#2530)
• disable jvm memory circuit breaker for IT (#2540)
• fix flaky test of PredictionITTests and RestConnectorToolIT (#2437)

Maintenance

• Updating security reachout email (#2445)

2.14.0.0

Version 2.14.0.0 Release Notes

Compatible with OpenSearch 2.14.0.0

Features

• Initiate MLInferencelngestProcessor (#2205)
• Add TTL to un-deploy model automatically (#2365)
• ML Model Interface (#2357)

Enhancements

• Change httpclient to async (#1958)
• Migrate RAG pipeline to async processing. (#2345)
• Filtering hidden model info from model profiling for users other than superadmin (#2332)
• check model auto deploy (#2288)
• restrict stash context only for stop words system index (#2283)
• Add a flag to control auto-deploy behavior (#2276)

Bug Fixes

• fix stopwords npe (#2311)
• guardrails npe (#2304)
• not sending failure message when model index isn't present (#2351)
• fix guardrails mapping (#2279)
• fix no model group index issue in connector helper notebook (#2336)
• Fixes #2317 predict api not working with asymmetric models (#2318)
• fixing isHidden null issue (#2337)
• fix remote register model / circuit breaker 500(#2264)
• guardrails bug fixes and IT for creating guardrails (#2269)
• Added missing result filter to inference (#2367)
• making Boolean type for isHidden (#2341
• Clear planningWorkerNodes when model auto-deploys again after undeploy (#2396)
• Avoid race condition in syncup model state refresh (#2405)
• Add a flag to distinguish remote model auto deploy and transport deploy (#2410)
• Add deploySetting in registering local models (#2415)

Refactoring

• feat: Add search index tool (#2356)
• Move visualization tool to ml-commons (#2363)

Documentation

• Add connector blueprint for VertexAI Embedding endpoint (#2268)

Infrastructure

• remove checkstyle (#2312)
• Increase rounding delta from 0.005% to 0.5% on RestMLInferenceIngestProcessorIT (#2372)
• add agent framework security it tests by (#2266)
• Add IT for interface (#2394)
• Fix local build failure for RestMLInferenceIngestProcessorIT (#2402)

Maintenance

• fix CVE for org.eclipse.core.runtime (#2378)