[FEATURE] Sample-to-Dataset in the SDK

Co-authored-by: Gretel GitHub Bot <devops+github-bot@gretel.ai>
GitOrigin-RevId: 844656ed38e34c8ef58f630ec8f38c323ee6ff2e

src/gretel_client/navigator/__init__.py

@@ -1,2 +1,6 @@
+from gretel_client.navigator.data_designer.factory import DataDesignerFactory
 from gretel_client.navigator.data_designer.interface import DataDesigner
-from gretel_client.navigator.workflow import NavigatorWorkflow
+from gretel_client.navigator.data_designer.sample_to_dataset import (
+    DataDesignerFromSampleRecords,
+)
+from gretel_client.navigator.workflow import DataDesignerWorkflow

src/gretel_client/navigator/client/remote.py

@@ -219,7 +219,6 @@ def submit_batch_workflow(
         logger.info(f"▶️ Starting your workflow run to generate {num_records} records:")
         logger.info(f"  |-- project_name: {project.name}")
         logger.info(f"  |-- project_id: {project.project_guid}")
-        logger.info(f"  |-- workflow_id: {batch_response.workflow_id}")
         logger.info(f"  |-- workflow_run_id: {batch_response.workflow_run_id}")
         logger.info(f"🔗 -> {workflow_run_url}")
 

src/gretel_client/navigator/data_designer/data_column.py

@@ -117,6 +117,25 @@ def generate_context_column_string(self, exclude: Optional[set[str]] = None) ->
             + "\n"
         )
 
+    def get_system_prompt(
+        self, special_system_instructions: Optional[str] = None
+    ) -> str:
+        """Get the system prompt for the column generation task.
+
+        Args:
+            special_instructions: Special instructions to be added to the system prompt.
+
+        Returns:
+            System prompt string.
+        """
+        return system_prompt_dict[self.llm_type].format(
+            special_instructions=(
+                ""
+                if special_system_instructions is None
+                else f"\n{special_system_instructions}\n"
+            )
+        )
+
     def to_generation_task(
         self,
         special_system_instructions: Optional[str] = None,
@@ -137,13 +156,7 @@ def to_generation_task(
             response_column_name=self.name,
             workflow_label=f"generating {self.name}",
             llm_type=self.llm_type,
-            system_prompt=system_prompt_dict[self.llm_type].format(
-                special_instructions=(
-                    ""
-                    if special_system_instructions is None
-                    else f"\n{special_system_instructions}\n"
-                )
-            ),
+            system_prompt=self.get_system_prompt(special_system_instructions),
             client=client,
         )
 

src/gretel_client/navigator/data_designer/factory.py

@@ -0,0 +1,121 @@
+import logging
+
+from pathlib import Path
+from typing import Optional, Union
+
+import pandas as pd
+
+from gretel_client.navigator.data_designer.interface import DataDesigner
+from gretel_client.navigator.data_designer.sample_to_dataset import (
+    DataDesignerFromSampleRecords,
+)
+from gretel_client.navigator.log import get_logger
+from gretel_client.navigator.tasks.types import (
+    DEFAULT_MODEL_SUITE,
+    ModelSuite,
+    RecordsT,
+)
+
+logger = get_logger(__name__, level=logging.INFO)
+
+
+class DataDesignerFactory:
+    """Factory class for creating DataDesigner instances.
+
+    Each class method on this object provides a different way to instantiate
+    a DataDesigner object, depending on your use case and desired workflow.
+
+    Allowed session keyword arguments:
+        api_key (str): Your Gretel API key. If set to "prompt" and no API key
+            is found on the system, you will be prompted for the key.
+        endpoint (str): Specifies the Gretel API endpoint. This must be a fully
+            qualified URL. The default is "https://api.gretel.cloud".
+        default_runner (str): Specifies the runner mode. Must be one of "cloud",
+            "local", "manual", or "hybrid". The default is "cloud".
+        artifact_endpoint (str): Specifies the endpoint for project and model
+            artifacts. Defaults to "cloud" for running in Gretel Cloud. If
+            working in hybrid mode, set to the URL of your artifact storage bucket.
+        cache (str): Valid options are "yes" or "no". If set to "no", the session
+            configuration will not be written to disk. If set to "yes", the
+            session configuration will be written to disk only if one doesn't
+            already exist. The default is "no".
+        validate (bool): If `True`, will validate the login credentials at
+            instantiation. The default is `False`.
+        clear (bool): If `True`, existing Gretel credentials will be removed.
+            The default is `False.`
+    """
+
+    @classmethod
+    def from_blank_canvas(
+        cls, model_suite: ModelSuite = DEFAULT_MODEL_SUITE, **kwargs
+    ) -> DataDesigner:
+        """Instantiate an empty DataDesigner instance that can be built up programmatically.
+
+        This initialization method is equivalent to directly instantiating a DataDesigner object.
+
+        Args:
+            model_suite: The model suite to use for generating synthetic data. Defaults to the
+                apache-2.0 licensed model suite.
+            **kwargs: Additional keyword arguments to pass to the DataDesigner constructor.
+
+        Returns:
+            An instance of DataDesigner with a blank canvas.
+        """
+        logger.info("🎨 Creating DataDesigner instance from blank canvas")
+
+        return DataDesigner(model_suite=model_suite, **kwargs)
+
+    @classmethod
+    def from_config(cls, config: dict, **kwargs) -> DataDesigner:
+        """Instantiate a DataDesigner instance from a configuration dictionary.
+
+        This method allows you to specify your data design using a YAML configuration file,
+        which is then built into a DataDesigner instance the same way you would do so programmatically.
+
+        Args:
+            config: A YAML configuration file, dict, or string that fully specifies the data design.
+            **kwargs: Additional keyword arguments to pass to the DataDesigner constructor.
+
+        Returns:
+            An instance of DataDesigner configured with the data seeds and generated data columns
+            defined in the configuration dictionary.
+        """
+        logger.info("🎨 Creating DataDesigner instance from config")
+
+        return DataDesigner.from_config(config, **kwargs)
+
+    @classmethod
+    def from_sample_records(
+        cls,
+        sample_records: Union[str, Path, pd.DataFrame, RecordsT],
+        *,
+        subsample_size: Optional[int] = None,
+        model_suite: ModelSuite = DEFAULT_MODEL_SUITE,
+        **kwargs,
+    ) -> DataDesigner:
+        """Instantiate a DataDesigner instance from sample records.
+
+        Use this subclass of DataDesigner when you want to turn a few sample records
+        into a rich, diverse synthetic dataset (Sample-to-Dataset).
+
+        Args:
+            sample_records: Sample records from which categorical data seeds will be extracted
+                and optionally used to create generated data columns.
+            subsample_size: The number of records to use from the sample records. If None,
+                all records will be used. If the subsample size is larger than the sample records,
+                the full sample will be used.
+            model_suite: The model suite to use for generating synthetic data. Defaults to the
+                apache-2.0 licensed model suite.
+
+        Returns:
+            An instance of DataDesigner configured to extract data seeds from the sample records
+            and optionally create generated data columns for each field in the sample records.
+        """
+        logger.info("🎨 Creating DataDesigner instance from sample records")
+
+        return DataDesignerFromSampleRecords(
+            sample_records=sample_records,
+            subsample_size=subsample_size,
+            model_suite=model_suite,
+            **kwargs,
+        )