podstack 1.3.14__tar.gz → 1.3.16__tar.gz

{podstack-1.3.14 → podstack-1.3.16}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: podstack
-Version: 1.3.14
+Version: 1.3.16
 Summary: Official Python SDK for Podstack GPU Notebook Platform
 Author-email: Podstack <support@podstack.ai>
 License-Expression: MIT

{podstack-1.3.14 → podstack-1.3.16}/podstack/registry/__init__.py

@@ -47,18 +47,22 @@ __all__ = [
     "set_experiment",
     "get_experiment",
     "list_experiments",
+    "archive_experiment",
     "start_run",
     "end_run",
+    "get_run",
+    "list_runs",
     "log_params",
     "log_metrics",
     "log_artifact",
     "set_tag",
+    "update_run_notes",
     "register_model",
     "get_model",
     "list_models",
     "set_model_stage",
     "set_model_alias",
-    # New methods
+    # MLOps helpers
     "log_model",
     "load_model",
     "log_dataset",
@@ -66,6 +70,32 @@ __all__ = [
     "get_metric_history",
     "download_artifact",
     "search_runs",
+    "get_run_datasets",
+    "get_model_lineage",
+    "autolog",
+    # HPO Sweeps
+    "create_sweep",
+    "get_sweep",
+    "list_sweeps",
+    "suggest_trial_params",
+    "create_trial",
+    "complete_trial",
+    "list_trials",
+    "stop_sweep",
+    # Alerts
+    "create_alert",
+    "list_alerts",
+    "delete_alert",
+    # Approvals
+    "list_pending_approvals",
+    "approve_promotion",
+    "reject_promotion",
+    # Schedules
+    "create_schedule",
+    "get_schedule",
+    "update_schedule",
+    "delete_schedule",
+    "list_schedules",
     # Classes
     "Experiment",
     "Run",
@@ -317,25 +347,47 @@ def load_model(model_name: str, version: int = None, stage: str = None, framewor
 def log_dataset(
     name: str,
     path: str = None,
+    source_path: str = None,
+    df=None,
+    context: str = "training",
+    split: str = None,
+    digest: str = None,
+    source_type: str = "local",
+    tags: dict = None,
     version: str = None,
     description: str = None,
-    digest: str = None,
     num_rows: int = None,
-    num_features: int = None
+    num_features: int = None,
 ):
     """
-    Log dataset metadata for the active run.
+    Log a dataset to the active run.
 
     Args:
         name: Dataset name.
-        path: Dataset path or URI.
-        version: Dataset version string.
-        description: Dataset description.
-        digest: Hash/digest of the dataset.
-        num_rows: Number of rows/samples.
-        num_features: Number of features/columns.
-    """
-    _get_client().log_dataset(name, path, version, description, digest, num_rows, num_features)
+        path: Local file path or URI.
+        source_path: Alias for path.
+        df: Optional pandas DataFrame for auto schema/profile computation.
+        context: One of "training", "validation", "test".
+        split: Alias for context ("train", "val", "test").
+        digest: SHA-256 hex digest (auto-computed from path if not provided).
+        source_type: One of "local", "s3", "gcs", "url".
+        tags: Optional dict of string tags.
+    """
+    return _get_client().log_dataset(
+        name=name,
+        path=path,
+        source_path=source_path,
+        df=df,
+        context=context,
+        split=split,
+        digest=digest,
+        source_type=source_type,
+        tags=tags,
+        version=version,
+        description=description,
+        num_rows=num_rows,
+        num_features=num_features,
+    )
 
 
 def compare_runs(run_ids: list, metric_keys: list = None) -> dict:
@@ -400,3 +452,227 @@ def search_runs(
         List of matching Run objects.
     """
     return _get_client().search_runs(experiment_id, status, max_results, offset)
+
+
+def get_run(run_id: str):
+    """Get a run by ID."""
+    return _get_client().get_run(run_id)
+
+
+def list_runs(
+    experiment_id: str = None,
+    status: str = None,
+    limit: int = 20,
+    offset: int = 0
+) -> list:
+    """List runs, optionally filtered by experiment or status."""
+    return _get_client().list_runs(experiment_id, status, limit, offset)
+
+
+def archive_experiment(experiment_id: str):
+    """Archive an experiment."""
+    return _get_client().archive_experiment(experiment_id)
+
+
+def update_run_notes(run_id: str, notes: str):
+    """Update the free-form notes for a run."""
+    _get_client().update_run_notes(run_id, notes)
+
+
+def get_run_datasets(run_id: str) -> list:
+    """List datasets logged for a run."""
+    return _get_client().get_run_datasets(run_id)
+
+
+def get_model_lineage(model_id: str) -> dict:
+    """Get full lineage for a model (versions → runs → datasets)."""
+    return _get_client().get_model_lineage(model_id)
+
+
+def autolog(
+    framework: str = None,
+    log_every_n_steps: int = 1,
+    log_system_metrics: bool = True,
+    system_metrics_interval: float = 10.0,
+):
+    """
+    Enable automatic logging for ML training frameworks.
+
+    Supports pytorch_lightning, huggingface, and sklearn.
+    Auto-detects available frameworks when framework is None.
+    """
+    _get_client().autolog(framework, log_every_n_steps, log_system_metrics, system_metrics_interval)
+
+
+# ==================== HPO Sweeps ====================
+
+def create_sweep(
+    experiment_id: str,
+    name: str,
+    search_space: dict,
+    strategy: str = "random",
+    max_trials: int = 20,
+    metric=None,
+    direction: str = "minimize",
+) -> dict:
+    """
+    Create a hyperparameter optimization sweep.
+
+    Args:
+        experiment_id: Experiment to run trials in.
+        name: Sweep name.
+        search_space: Dict mapping param names to spec dicts.
+        strategy: "random" (default) or "grid".
+        max_trials: Maximum number of trials.
+        metric: Metric key to optimize (str), or dict with "name" and "direction" keys.
+        direction: "minimize" (default) or "maximize". Ignored if metric is a dict.
+
+    Returns:
+        Sweep dict with id, status, etc.
+    """
+    if isinstance(metric, dict):
+        direction = metric.get("direction", direction)
+        metric = metric.get("name", None)
+    return _get_client().create_sweep(
+        experiment_id, name, search_space, strategy, max_trials, metric, direction
+    )
+
+
+def get_sweep(sweep_id: str) -> dict:
+    """Get a sweep by ID."""
+    return _get_client().get_sweep(sweep_id)
+
+
+def list_sweeps(experiment_id: str) -> list:
+    """List all sweeps for an experiment."""
+    return _get_client().list_sweeps(experiment_id)
+
+
+def suggest_trial_params(sweep_id: str) -> dict:
+    """Get suggested hyperparameter values for the next trial."""
+    return _get_client().suggest_trial_params(sweep_id)
+
+
+def create_trial(sweep_id: str, run_id: str, params: dict) -> dict:
+    """Record a new trial linked to a sweep and run."""
+    return _get_client().create_trial(sweep_id, run_id, params)
+
+
+def complete_trial(sweep_id: str, trial_id: str, value: float) -> None:
+    """Mark a trial as completed with its objective metric value."""
+    _get_client().complete_trial(sweep_id, trial_id, value)
+
+
+def list_trials(sweep_id: str) -> list:
+    """List all trials for a sweep."""
+    return _get_client().list_trials(sweep_id)
+
+
+def stop_sweep(sweep_id: str) -> None:
+    """Stop a running sweep."""
+    _get_client().stop_sweep(sweep_id)
+
+
+# ==================== Alerts ====================
+
+def create_alert(
+    run_id: str,
+    metric_key: str,
+    condition: str,
+    threshold: float,
+    notify_email: str = None,
+    notify_slack: str = None,
+) -> dict:
+    """
+    Create a metric threshold alert for a run.
+
+    Args:
+        run_id: Run to monitor.
+        metric_key: Metric name to watch.
+        condition: One of gt, lt, gte, lte, eq.
+        threshold: Trigger threshold value.
+        notify_email: Email address to notify.
+        notify_slack: Slack webhook URL to notify.
+
+    Returns:
+        Alert dict with id.
+    """
+    return _get_client().create_alert(run_id, metric_key, condition, threshold, notify_email, notify_slack)
+
+
+def list_alerts(run_id: str) -> list:
+    """List all alerts for a run."""
+    return _get_client().list_alerts(run_id)
+
+
+def delete_alert(alert_id: str) -> None:
+    """Delete an alert by ID."""
+    _get_client().delete_alert(alert_id)
+
+
+# ==================== Approvals ====================
+
+def list_pending_approvals() -> list:
+    """List all pending model promotion approval requests in the project."""
+    return _get_client().list_pending_approvals()
+
+
+def approve_promotion(request_id: str, comment: str = None) -> dict:
+    """Approve a pending model stage promotion request."""
+    return _get_client().approve_promotion(request_id, comment)
+
+
+def reject_promotion(request_id: str, comment: str = None) -> dict:
+    """Reject a pending model stage promotion request."""
+    return _get_client().reject_promotion(request_id, comment)
+
+
+# ==================== Schedules ====================
+
+def create_schedule(
+    name: str,
+    experiment_id: str,
+    cron_expr: str,
+    run_name: str = None,
+    run_config: dict = None,
+    webhook_url: str = None,
+) -> dict:
+    """
+    Create a recurring training schedule using a cron expression.
+
+    Args:
+        name: Schedule name.
+        experiment_id: Experiment to create runs in.
+        cron_expr: 5-field cron expression (e.g. "0 2 * * 1").
+        run_name: Base name for created runs.
+        run_config: Optional params to log on each scheduled run.
+        webhook_url: Optional URL to POST after each run fires.
+
+    Returns:
+        Schedule dict with id, next_fire_at, etc.
+    """
+    return _get_client().create_schedule(name, experiment_id, cron_expr, run_name, run_config, webhook_url)
+
+
+def list_schedules() -> list:
+    """List all training schedules in the project."""
+    return _get_client().list_schedules()
+
+
+def get_schedule(schedule_id: str) -> dict:
+    """Get a schedule by ID."""
+    return _get_client().get_schedule(schedule_id)
+
+
+def update_schedule(
+    schedule_id: str,
+    enabled: bool = None,
+    cron_expr: str = None,
+) -> dict:
+    """Update a schedule's enabled state or cron expression."""
+    return _get_client().update_schedule(schedule_id, enabled, cron_expr)
+
+
+def delete_schedule(schedule_id: str) -> None:
+    """Delete a schedule."""
+    _get_client().delete_schedule(schedule_id)

{podstack-1.3.14 → podstack-1.3.16}/podstack/registry/client.py

@@ -977,8 +977,10 @@ class RegistryClient:
         self,
         name: str,
         path: str = None,
+        source_path: str = None,
         df=None,
         context: str = "training",
+        split: str = None,
         digest: str = None,
         source_type: str = "local",
         tags: dict = None,
@@ -1021,6 +1023,13 @@ class RegistryClient:
         if not self._active_run:
             raise NoActiveRunError()
 
+        # Aliases
+        if source_path and not path:
+            path = source_path
+        if split:
+            _split_map = {"train": "training", "val": "validation", "valid": "validation", "test": "test"}
+            context = _split_map.get(split.lower(), split)
+
         schema: Dict[str, str] = {}
         profile: Dict[str, Any] = {}
 

{podstack-1.3.14 → podstack-1.3.16}/podstack.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: podstack
-Version: 1.3.14
+Version: 1.3.16
 Summary: Official Python SDK for Podstack GPU Notebook Platform
 Author-email: Podstack <support@podstack.ai>
 License-Expression: MIT

{podstack-1.3.14 → podstack-1.3.16}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "podstack"
-version = "1.3.14"
+version = "1.3.16"
 description = "Official Python SDK for Podstack GPU Notebook Platform"
 readme = "README.md"
 license = "MIT"