sutro 0.1.37__py3-none-any.whl → 0.1.38__py3-none-any.whl

sutro/templates/classification.py

@@ -0,0 +1,117 @@
+import json
+from typing import Union, List
+import polars as pl
+import pandas as pd
+from pydantic import BaseModel
+
+from ..common import ModelOptions
+from ..interfaces import BaseSutroClient
+
+
+class ClassificationTemplates(BaseSutroClient):
+    def classify(
+        self,
+        data: Union[List, pd.DataFrame, pl.DataFrame, str],
+        classes: Union[dict[str, str], list[str]],
+        model: ModelOptions = "gemma-3-12b-it",
+        job_priority: int = 0,
+        name: Union[str, List[str]] = None,
+        description: Union[str, List[str]] = None,
+        output_column: str = "inference_result",
+        column: Union[str, List[str]] = None,
+        truncate_rows: bool = True,
+        include_scratchpad: bool = False,
+    ):
+        """
+        A simple template style function to perform classification on the provided data with Sutro. The intention is that the implemented code should be very easy to extend further, while showing a basic structure for large-scale classification with Sutro.
+
+        It uses structured outputs with a scratchpad field, enabling the model to reason step-by-step before providing the final classification.
+        The method supports various input formats including lists, DataFrames (Polars or Pandas), file paths, and datasets.
+        The method will wait for the classification job to complete before returning the results.
+
+        Args:
+            data (Union[List, pd.DataFrame, pl.DataFrame, str]): The data to classify. Each row should contain some text to classifiy that fits into one of the passed in labels.
+            classes (Union[dict[str, str], list[str]]): The classification classes. Can be either:
+                - A list of class names, ie ["Positive", "Negative", "Neutral"]
+                - A dict mapping class labels to descriptions, ie {"Positive": "Expresses satisfaction...", ...}
+                Providing descriptions can improve classification accuracy, especially for ambiguous or domain-specific categories.
+            model (ModelOptions, optional): The LLM to use. Defaults to "gemma-3-12b-it"; a model chosen for its balance of performance and efficiency, that also retains competency across a broad number of different domains.
+            job_priority (int, optional): The priority of the job. Defaults to 0.
+            name (Union[str, List[str]], optional): A job name for experiment/metadata tracking purposes. Defaults to None.
+            description (Union[str, List[str]], optional): A job description for experiment/metadata tracking purposes. Defaults to None.
+            output_column (str, optional): The column name to store the classification results in if the input is a DataFrame. Defaults to "inference_result".
+            column (Union[str, List[str]], optional): The column name to use for classification. Required if data is a DataFrame, file path, or dataset. If a list is supplied, it will concatenate the columns of the list into a single column, accepting separator strings.
+            truncate_rows (bool, optional): If True, any rows that have a token count exceeding the context window length of the selected model will be truncated to the max length that will fit within the context window. Defaults to True.
+            include_scratchpad (bool, optional): If True, includes the model's thinking scratchpad in the output. If False, only returns the final classification. Defaults to False.
+
+        Returns:
+            The completed classification results for the provided data. If include_scratchpad is True, returns both scratchpad and classification fields in JSON object. If False, returns only the classification as a string.
+
+        """
+        if isinstance(classes, dict):
+            formatted_classes = "\n".join(
+                [f"- {name}: {desc}" for name, desc in classes.items()]
+            )
+        else:
+            formatted_classes = "\n".join([f"- {c}" for c in classes])
+
+        system_prompt = f"""You are an expert classifier. Your task is to accurately categorize the input into one of the provided classes.
+
+## Classes
+
+{formatted_classes}
+
+## Instructions
+
+1. **Analyze the input carefully**: Read and understand the full context - identify key elements, themes, and characteristics
+
+2. **Consider each class**: For each possible class, evaluate how similar the input is to its typical characteristics
+
+3. **Provide your reasoning in the scratchpad**: Think through which class fits best and why
+
+4. **Provide output**: Give your final classification
+
+If needed, use the scratchpad field to work through steps 1-3, then provide your final answer in the classification field.
+
+## Guidelines
+
+- Select exactly ONE class, even if multiple seem applicable (choose the best match)
+- If the input is ambiguous, choose the closest fit and explain your reasoning
+- Base your decision on the actual content, not assumptions or implications
+- Similar inputs should receive the same classification
+
+Respond using the structured format with scratchpad and classification fields."""
+
+        class ClassificationOutput(BaseModel):
+            # Since we're using structured outputs, we want to give the model some
+            # space to reason and think as needed
+            scratchpad: str
+            classification: str
+
+        job_id = self.infer(
+            data,
+            model,
+            name,
+            description,
+            system_prompt=system_prompt,
+            output_schema=ClassificationOutput,
+            column=column,
+            output_column=output_column,
+            job_priority=job_priority,
+            truncate_rows=truncate_rows,
+            stay_attached=False,
+        )
+
+        results = self.await_job_completion(job_id)
+
+        # Filter out scratchpad if not wanted
+        if not include_scratchpad:
+            results = results.with_columns(
+                pl.col(output_column)
+                .map_elements(
+                    lambda x: json.loads(x)["classification"], return_dtype=pl.Utf8
+                )
+                .alias(output_column)
+            )
+
+        return results

sutro/templates/embed.py

@@ -0,0 +1,53 @@
+from typing import Union, List
+import polars as pl
+import pandas as pd
+from ..common import EmbeddingModelOptions
+from ..interfaces import BaseSutroClient
+
+
+class EmbeddingTemplates(BaseSutroClient):
+    def embed(
+        self,
+        data: Union[List, pd.DataFrame, pl.DataFrame, str],
+        model: EmbeddingModelOptions = "qwen-3-embedding-0.6b",
+        job_priority: int = 0,
+        name: Union[str, List[str]] = None,
+        description: Union[str, List[str]] = None,
+        output_column: str = "inference_result",
+        column: Union[str, List[str]] = None,
+        truncate_rows: bool = True,
+    ):
+        """
+        A simple template style function to generate embeddings for the provided data, with Sutro. The intention is that the implemented code should be very easy to extend further, while showing a basic structure for large scale embedding generation with Sutro.
+
+        This method allows you to generate vector embeddings for the provided data using Sutro.
+        It supports various options for inputting data, such as lists, DataFrames (Polars or Pandas), file paths and datasets.
+        The method will wait for the embedding job to complete before returning the results.
+
+        Args:
+            data (Union[List, pd.DataFrame, pl.DataFrame, str]): The data to generate embeddings for.
+            model (ModelOptions, optional): The embedding model to use. Defaults to "qwen-3-embedding-0.6b"; a model we chose as its small & fast, yet performs well on a variety of tasks.
+            job_priority (int, optional): The priority of the job. Defaults to 0.
+            name (Union[str, List[str]], optional): A job name for experiment/metadata tracking purposes. Defaults to None.
+            description (Union[str, List[str]], optional): A job description for experiment/metadata tracking purposes. Defaults to None.
+            output_column (str, optional): The column name to store the embedding results in if the input is a DataFrame. Defaults to "inference_result".
+            column (Union[str, List[str]], optional): The column name to use for embedding generation. Required if data is a DataFrame, file path, or dataset. If a list is supplied, it will concatenate the columns of the list into a single column, accepting separator strings.
+            truncate_rows (bool, optional): If True, any rows that have a token count exceeding the context window length of the selected model will be truncated to the max length that will fit within the context window. Defaults to True.
+
+        Returns:
+            The completed embedding results for the provided data.
+
+        """
+        job_id = self.infer(
+            data,
+            model,
+            name,
+            description,
+            column,
+            output_column,
+            job_priority,
+            truncate_rows=truncate_rows,
+            stay_attached=False,
+        )
+
+        return self.await_job_completion(job_id)

sutro/validation.py

@@ -0,0 +1,60 @@
+import importlib.metadata
+import json
+import os
+
+import requests
+
+from sutro.common import to_colored_text
+
+
+def check_version(package_name: str):
+    try:
+        # Local version
+        local_version = importlib.metadata.version(package_name)
+    except importlib.metadata.PackageNotFoundError:
+        print(f"{package_name} is not installed.")
+        return
+
+    try:
+        # Latest release from PyPI
+        resp = requests.get(f"https://pypi.org/pypi/{package_name}/json", timeout=2)
+        resp.raise_for_status()
+        latest_version = resp.json()["info"]["version"]
+
+        if local_version != latest_version:
+            msg = (
+                f"⚠️  You are using {package_name} {local_version}, "
+                f"but the latest release is {latest_version}. "
+                f"Run `[uv] pip install -U {package_name}` to upgrade."
+            )
+            print(to_colored_text(msg, state="callout"))
+    except Exception:
+        # Fail silently or log, you don’t want this blocking usage
+        pass
+
+
+def check_for_api_key():
+    """
+    Check for an API key in the user's home directory.
+
+    This method looks for a configuration file named 'config.json' in the
+    '.sutro' directory within the user's home directory.
+    If the file exists, it attempts to read the API key from it.
+
+    Returns:
+        str or None: The API key if found in the configuration file, or None if not found.
+
+    Note:
+        The expected structure of the config.json file is:
+        {
+            "api_key": "your_api_key_here"
+        }
+    """
+    CONFIG_DIR = os.path.expanduser("~/.sutro")
+    CONFIG_FILE = os.path.join(CONFIG_DIR, "config.json")
+    if os.path.exists(CONFIG_FILE):
+        with open(CONFIG_FILE, "r") as f:
+            config = json.load(f)
+        return config.get("api_key")
+    else:
+        return None

{sutro-0.1.37.dist-info → sutro-0.1.38.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sutro
-Version: 0.1.37
+Version: 0.1.38
 Summary: Sutro Python SDK
 License-Expression: Apache-2.0
 Requires-Dist: numpy>=2.1.1,<3.0.0

sutro-0.1.38.dist-info/RECORD

@@ -0,0 +1,12 @@
+sutro/__init__.py,sha256=yUiVwcZ8QamSqDdRHgzoANyTZ-x3cPzlt2Fs5OllR_w,402
+sutro/cli.py,sha256=0NuqRInXA-_7TRw-T0OxP8otmUHUarMtY7kuLbWDous,13751
+sutro/common.py,sha256=FuTYTzy82Ul56r9SVH0XMOqcBBspDAFvrtHM93ZbT_8,6945
+sutro/interfaces.py,sha256=j8k4iEHjHu6HDEb9XqiuJrIRbXhLZi9WsiVmxC97R8s,2972
+sutro/sdk.py,sha256=qhpmQNHZDaeGffPyCopmlc6YQuA1_hLFmuHeQXlNbSM,56107
+sutro/templates/classification.py,sha256=iNFiyuR8bZc9Xe-NdimklpQUveqg3p_eJOlEAHaj7Is,6080
+sutro/templates/embed.py,sha256=csvLA0hw5Qaro_yZvALRRp9_SbfWABFN0iQXrf8E8_I,2941
+sutro/validation.py,sha256=FlFH5e5PAPIPpCrzU7mwfZKDDvrmkHt2yYsFm0Ahfmg,1849
+sutro-0.1.38.dist-info/WHEEL,sha256=X16MKk8bp2DRsAuyteHJ-9qOjzmnY0x1aj0P1ftqqWA,78
+sutro-0.1.38.dist-info/entry_points.txt,sha256=s-dtPZ0AScjvR8S_ykhzXxtVcUjrRlxVxyJymI81A3E,41
+sutro-0.1.38.dist-info/METADATA,sha256=VeFTzSqKUiPkx8Ey-g1DkZ-_EkrLt6BwVVQa7-XC-sw,6259
+sutro-0.1.38.dist-info/RECORD,,

sutro-0.1.37.dist-info/RECORD

@@ -1,7 +0,0 @@
-sutro/__init__.py,sha256=yUiVwcZ8QamSqDdRHgzoANyTZ-x3cPzlt2Fs5OllR_w,402
-sutro/cli.py,sha256=_FU8PwP4dMzXXg5ldxCXP3kaZvQtOKdA8Kzjc34xmQ0,13727
-sutro/sdk.py,sha256=dysuW6jwtuMjVTdDH1zCoycWLvjzBZa_Mi6dSM_zWpY,63799
-sutro-0.1.37.dist-info/WHEEL,sha256=X16MKk8bp2DRsAuyteHJ-9qOjzmnY0x1aj0P1ftqqWA,78
-sutro-0.1.37.dist-info/entry_points.txt,sha256=s-dtPZ0AScjvR8S_ykhzXxtVcUjrRlxVxyJymI81A3E,41
-sutro-0.1.37.dist-info/METADATA,sha256=pOSPs0yhCpKEhHZJIPIaL-wxSXYoUVBTLQNqN7WjO3E,6259
-sutro-0.1.37.dist-info/RECORD,,