Adding _copy_file_or_tree tests

* Preliminary Java support

* Renames

* Package restructuring

* Add mleap python flavor - preliminary

* Implement javafunc with correct loader module

* Input parsing

* Add missing files

* Failed attempt at package installation

* Fix up naming semantics, loader module impl

* Server impl

* Fixes - serving works

* Benchmarking results

* Benchmarking script

* SparkML critical path lats

* add batching to bmark

* Remove runID from SageMakerServer

* ..

* Add method that allows deserialization of MLeap pipeline in native format

* Move JavaFunc into protected sagemaker package

* Modify pyfunc container to handle mleap

* Python API changes progress

* Use package protection to hide modules that we don't yet wish to expose to users

* Fixes

* Container mods

* Remove mlflow-java directory

* Serving is functional. Add sparkml module

* Exception message fixes

* Throw unsupportedop from attempting to load a model from runId

* Remove nonfunctional PackageInstaller

* Package restructuring, remove unnecessary lines

* Improved error handling

* Doc tweaks

* JavaFunc error coherency improvements

* Remove unused main method

* Fix more merge conflicts

* Remove invalid comments

* Attempt at MLeap serialization parameter validation

* Debugging

* Exception tweaks

* Add javafunc experimental comments

* Remove javafunc line

* Pom fixes

* More pom fixes, make container use jar, correct version

* Construct leapframe from pandas

* Add missing files

* Docs fixes

* Comment fix

* Docs fix

* pom fixes

* New tests

* Test improvements

* Fix print statement omission

* Docker build, spark lib fixes

* Renaming, new yaml/json utilities, prediction column selection from MLeap

* Container emits json-serialized predictions column

* lambda syntax fix

* MLeap predictor return json

* Merge docker tests

* Use

* Add test for correct handling of invalid inputs

* helper function documentation

* Pom changes, docs

* Revert MLeap, sagemaker container

* Remove leapframe, sagemaker java infra

* Revert tests

* Revert helper functions for tests

* Fix spacing

* revert setup py

* Remove commented pom

* Remove javafunc

* Remove py

* Add style checker to pom with correct configuration

* Docs

* Add tests - FOR REF LATER

* Remove tests

* Remove unused tracking utils

* Revert "Remove tests"

This reverts commit c88d6b2.

* Remove pipeline deserialization test

* Serialization utils docs

* Add unit tests

* Error message fix

* Revert py changes

* Revert py

* Remove unnecessary test resources

* Doc link fixes

* Change package name in java file

* Change package name in directory structure

* Pom updates

* Print stack trace on modeltest fail

* Remove trailing  from test case names

* Rename tests to comply with style guide, add class argument to serialization  utility

* Format code

* Format tests

* Rename  to

* Update travis yml with Java tests

* Remove superfluous files

* Test fix

* Remove swn files

* Remove swm

* begin R service API refactoring

* fix tracking uri getting

* remove old mlflow_create_experiment()

* reorg files

* local server registration

* experiment adding

* mlflow client server url attribute

* delete/restore experiments

* get run

* log metric

* fix log metric docs

* set tag

* log param

* get param

* get metric

* get metric history

* list and get experiment

* set terminated

* reorg files

* refactor mlflow_log_artifact()

* reorg files

* remove unused mlflow_ensure_run_id

* refactor mlflow_end_run()

* delete old cold

* fix get experiment

* options for view_type for list experiments

* revise tests

* fix mlflow_get_run dispatch

* deregister local servers in tests

* fix tests

* pass client to rest calls

* more fixes

* get or start run for all fluent

* actually refresh state for tests

* remove mlflow_connect()

* rename mlflow_tracking_uri to mlflow_get_tracking_uri

* remove mlflow_disconnect()

* remove active connection code

* rename mlflow_connection class to mlflow_server

* remove unused mlflow_connection_url

* remove mentions of connection

* cosmetics

* typo

* remove dead code

* with.mlflow_active_run condition handling

* error message when client isn't specified for tracking service API

* active experiment id

* setting tracking uri shouldn't change other globals

* refactor entities run

* test server -> validate server

* reorg files

* reorg test files

* remove run_uuid parameter from mlflow_log_model

* clean up create/list experiment rest wrapper

* remove unnecessary assignments in rest wrappers

* set experiment id independent of run

* consistent naming getters and setters

* some r cmd check fixes

* fix r cmd check

* rename mlflow_get_active_run -> mlflow_active_run and make private mlflow_set_active_run

* unexport experiment id getting/setting

* add mlflow_get_experiment_by_name() and mlflow_set_experiment()

* typo

* with.mlflow_run

* refactor create experiment

* refactor list experiments

* refactor get run

* refactor create run

* refactor get experiment

* refactor get experiment by name

* refactor delete experiment

* refactor restore experiment

* log metric

* refactor client api

* fluent api refactor

* reorg files

* tracking client docs

* fluent docs

* rbuildignore internal

* typos

* default end_time of mlflow_client_set_terminated to current time

* fix mlflow_end_run()

* default user

* document tracking client api

* shorten function name

* add mlflow_client_list_artifacts()

* ignore noisy lintr warnings (#1)

Update MLflow hosted docs to include latest updates from MLflow master

Proto cleanup

Signed-off-by: B-Step62 <yuki.watanabe@databricks.com>

Signed-off-by: B-Step62 <yuki.watanabe@databricks.com>

Signed-off-by: B-Step62 <yuki.watanabe@databricks.com>

Signed-off-by: B-Step62 <yuki.watanabe@databricks.com>
Signed-off-by: juntai-zheng <juntai.zheng@gmail.com>

This commit implements PR mlflow#1 of the multi-turn scorer implementation plan:

1. Add MLFLOW_ENABLE_MULTI_TURN_EVALUATION environment variable:
   - Defaults to False (feature is opt-in)
   - Controls whether multi-turn evaluation is enabled
   - Follows existing pattern for experimental features

2. Update Scorer base class for multi-turn support:
   - Document session_traces parameter in Scorer.__call__() docstring
   - Update Scorer.run() to accept and handle session_traces parameter
   - session_traces contains list of traces from same conversation session
   - Note: _is_multi_turn_scorer attribute already exists from previous commit

These changes provide the foundational infrastructure for multi-turn
evaluation without changing any evaluation logic or behavior.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: Avesh Singh <aveshcsingh@gmail.com>

Signed-off-by: B-Step62 <yuki.watanabe@databricks.com>

#20001)

Signed-off-by: B-Step62 <yuki.watanabe@databricks.com>
Signed-off-by: NJAHNAVI2907 <jahnavinelavelli29@gmail.com>

Issue mlflow#1: Improve test to verify timeout mechanism
- Simplified test_shutdown_timeout_prevents_hang to verify timeout exists
- Added environment cleanup to all tests to prevent interference
- All 11 tests now pass

Issue mlflow#2: Add activation checks to _at_exit_callback
- Added checks for queue activation state before cleanup
- Added checks for thread/threadpool existence to prevent AttributeError
- Handles edge case where shut_down_async_logging() called before atexit
- Applied to both async_artifacts_logging_queue.py and async_logging_queue.py

Issue mlflow#3: Register environment variable properly
- Added MLFLOW_ASYNC_LOGGING_SHUTDOWN_TIMEOUT to environment_variables.py
- Type: float, default: 30.0
- Updated both queue files to import and use registered variable
- Removed manual os.environ.get() calls
- Removed unused 'os' import from async_artifacts_logging_queue.py

All changes improve code quality, consistency, and maintainability.

Issue mlflow#1: Improve test to verify timeout mechanism
- Simplified test_shutdown_timeout_prevents_hang to verify timeout exists
- Added environment cleanup to all tests to prevent interference
- All 11 tests now pass

Issue mlflow#2: Add activation checks to _at_exit_callback
- Added checks for queue activation state before cleanup
- Added checks for thread/threadpool existence to prevent AttributeError
- Handles edge case where shut_down_async_logging() called before atexit
- Applied to both async_artifacts_logging_queue.py and async_logging_queue.py

Issue mlflow#3: Register environment variable properly
- Added MLFLOW_ASYNC_LOGGING_SHUTDOWN_TIMEOUT to environment_variables.py
- Type: float, default: 30.0
- Updated both queue files to import and use registered variable
- Removed manual os.environ.get() calls
- Removed unused 'os' import from async_artifacts_logging_queue.py

All changes improve code quality, consistency, and maintainability.

Signed-off-by: SF2 Orchestrator <orchestrator@sf2.local>