Skip to content

NEWS.md

Latest commit

 

History

History
125 lines (105 loc) · 5.18 KB

NEWS.md

File metadata and controls

125 lines (105 loc) · 5.18 KB

snowflakeR 0.2.0

New modules

Model Monitoring

Full model-monitoring lifecycle for deployed models:

  • sfr_monitor_source() / sfr_monitor_config() -- configure monitoring sources and settings
  • sfr_add_monitor() / sfr_get_monitor() / sfr_show_model_monitors() / sfr_delete_monitor() -- CRUD for monitors
  • sfr_monitor_drift() / sfr_monitor_performance() / sfr_monitor_stats() -- retrieve monitoring results
  • sfr_suspend_monitor() / sfr_resume_monitor() / sfr_describe_monitor() -- lifecycle management
  • sfr_add_monitor_segment() / sfr_drop_monitor_segment() -- segment drill-down
  • sfr_monitor_to_vetiver() / sfr_vetiver_to_metrics() -- vetiver integration bridge

Experiment Tracking

MLflow-style experiment tracking on Snowflake:

  • sfr_experiment() / sfr_start_run() / sfr_end_run() / sfr_delete_run() / sfr_delete_experiment() -- experiment lifecycle
  • sfr_exp_log_param() / sfr_exp_log_params() -- parameter logging
  • sfr_exp_log_metric() / sfr_exp_log_metrics() -- metric logging
  • sfr_exp_log_model() / sfr_exp_log_artifact() -- artifact logging
  • sfr_exp_list_artifacts() / sfr_exp_download_artifact() -- artifact retrieval
  • sfr_experiment_from_tune() / sfr_experiment_log_best() -- tune grid-search integration

New Feature Store functions

  • sfr_attach_feature_desc() -- attach human-readable descriptions to individual features
  • sfr_slice_feature_view() -- create a column-subset slice of a Feature View
  • sfr_fv_lineage() -- trace upstream/downstream lineage from a Feature View
  • sfr_list_fv_columns() -- list columns and types for a registered Feature View
  • sfr_fv_to_df() -- read a Feature View's data as a data.frame
  • sfr_fv_query() -- retrieve the underlying SQL query for a Feature View
  • sfr_fv_fqn() -- get the fully qualified name of a Feature View
  • sfr_load_fvs_from_dataset() -- recover Feature Views associated with a Dataset
  • sfr_update_default_warehouse() -- change the default warehouse for a Feature Store
  • sfr_storage_config() -- create Iceberg-backed storage configurations

New Model Registry functions

  • sfr_delete_model_version() -- delete a specific model version
  • sfr_get_model_metric() -- read a single metric by name
  • sfr_delete_model_metric() -- delete a single metric
  • sfr_model_description() -- get or set the version description
  • sfr_show_model_functions() -- list callable functions on a model version
  • sfr_model_lineage() -- trace upstream/downstream model lineage
  • sfr_export_model() -- export model artifacts to a local directory
  • sfr_get_model_task() -- get the task type of a model version
  • sfr_list_services() -- list active SPCS services for a model version
  • sfr_run_batch() -- run batch inference via SPCS
  • sfr_models() -- list Model objects in a registry (vs sfr_show_models() which returns a summary DataFrame)

New parameters on existing functions

Feature Store

  • sfr_feature_store(): default_iceberg_external_volume
  • sfr_feature_view() / sfr_create_feature_view(): initialize, refresh_mode, cluster_by, online_config
  • sfr_register_feature_view(): block
  • sfr_read_feature_view(): store_type, keys, feature_names
  • sfr_refresh_feature_view(): store_type
  • sfr_get_refresh_history(): store_type
  • sfr_generate_training_data(): exclude_columns, include_feature_view_timestamp_col, auto_prefix, join_method
  • sfr_generate_dataset(): exclude_columns, include_feature_view_timestamp_col, auto_prefix, join_method, output_type
  • sfr_retrieve_features(): exclude_columns, include_feature_view_timestamp_col, auto_prefix, join_method

Model Registry

  • sfr_log_model(): user_files, code_paths, resource_constraint, python_version
  • sfr_predict(): partition_column, strict_input_validation
  • sfr_deploy_model(): image_build_compute_pool, cpu_requests, memory_requests, gpu_requests, num_workers, max_batch_rows, block, build_external_access_integrations

Bug fixes

  • sfr_read_feature_view(): Fixed reticulate type conversion for keys (R character vectors now correctly converted to Python list of lists of strings) and feature_names parameters.
  • sfr_fv_to_df(): Worked around a reticulate TypeError on cluster_by field type inference by using fs.read_feature_view() instead of fv.to_df().
  • sfr_slice_feature_view(): Fixed AttributeError when the FeatureViewSlice object lacks name/version attributes; now falls back to the original request values.
  • sfr_models(): Fixed rbind failure when model comment is NULL by coercing to empty string.
  • sfr_add_monitor_segment() / sfr_drop_monitor_segment(): Added missing segment parameter in the Python bridge call.

Documentation

  • Regenerated NAMESPACE and man/ pages for all new exports.
  • Extended Feature Store and Model Registry vignettes with new sections covering online serving, aggregation, introspection, slicing, Iceberg, lineage, aliases, SQL-direct inference, batch inference, advanced deployment, and more.
  • Updated README.md module overview table and example notebooks table.
  • Added this NEWS.md changelog.