oracle-ads 2.13.8__py3-none-any.whl → 2.13.9rc0__py3-none-any.whl

ads/jobs/templates/oci_metrics.py

@@ -1,181 +0,0 @@
-import datetime
-import oci
-import os
-import subprocess
-import time
-
-
-JOB_RUN_COMPARTMENT_OCID_ENV = "JOB_RUN_COMPARTMENT_OCID"
-METRIC_NAMESPACE = os.environ.get("OCI__METRICS_NAMESPACE")
-METRIC_COMPARTMENT = os.environ.get(JOB_RUN_COMPARTMENT_OCID_ENV)
-
-# When querying metrics, the smallest aggregation interval allowed is 1 minute.
-# See https://docs.oracle.com/iaas/Content/Monitoring/Reference/mql.htm#Interval
-METRIC_SUBMISSION_INTERVAL_SECONDS = 60
-
-
-class Metric:
-    """
-    Class containing metric details
-    """
-
-    def __init__(self, name: str, value: float, dimensions: dict = None):
-        """
-        Initializes a Metrics object
-
-        Parameters
-        ----------
-        name: str
-            The metric name
-        value: float
-            The metric value
-        dimensions: dict
-            Dictionary of dimensions to include on the metric.
-        """
-        self.name = name
-        self.value = value
-        self.dimensions = dimensions if dimensions else {}
-
-
-class GpuMetricsProvider:
-    """
-    Custom GPU utilization metrics provider
-    """
-
-    def get_metrics(self) -> list:
-        """
-        Get custom GPU metrics
-
-        Returns
-        -------
-        list
-            List of Metric objects.
-        """
-        gpu_metrics = []
-        try:
-            # Example output:
-            # 00000000:00:04.0, 42.27, 40, 20, 16384, 15287
-            # 00000000:00:05.0, 41.30, 42, 0, 16384, 479
-            nvidia_smi_output = subprocess.check_output(
-                [
-                    "nvidia-smi",
-                    "--query-gpu=pci.bus_id,power.draw,temperature.gpu,utilization.gpu,memory.total,memory.used",
-                    "--format=csv,noheader,nounits",
-                ]
-            ).decode("utf-8")
-
-            for line in nvidia_smi_output.split("\n"):
-                if len(line) > 0:
-                    values = line.split(", ")
-                    if len(values) >= 6:
-                        dimensions = {"pci_bus": values[0]}
-                        gpu_metrics.append(
-                            Metric("gpu.power_draw", float(values[1]), dimensions)
-                        )
-                        gpu_metrics.append(
-                            Metric("gpu.temperature", float(values[2]), dimensions)
-                        )
-                        gpu_metrics.append(
-                            Metric("gpu.gpu_utilization", float(values[3]), dimensions)
-                        )
-                        gpu_metrics.append(
-                            Metric(
-                                "gpu.memory_usage",
-                                round(float(values[5]) / float(values[4]) * 100, 2),
-                                dimensions,
-                            )
-                        )
-                    else:
-                        print(
-                            f"Unexpected nvidia-smi output format. Composing no metrics for output line: {line}"
-                        )
-        except Exception as e:
-            print(f"Unexpected error encountered querying GPU: {e}")
-
-        return gpu_metrics
-
-
-metric_providers = [
-    GpuMetricsProvider(),
-]
-
-
-def convert_to_metric_data_details(
-    metric: Metric, timestamp: datetime.datetime
-) -> oci.monitoring.models.MetricDataDetails:
-    """
-    Converts a Metric object into an oci.monitoring.models.MetricDataDetails object for submission to the Monitoring
-    Service. In addition to the dimensions defined on the input Metric, the job ocid and job run ocid are also added
-    as dimensions.
-
-    Parameters
-    ----------
-    metric: Metric
-        The Metric object to convert
-    timestamp: datetime.datetime
-        The timestamp to include on the metric datapoint
-
-    Returns
-    -------
-    oci.monitoring.models.MetricDataDetails
-        The oci.monitoring.models.MetricDataDetails object containing the metric details
-    """
-    dimensions = metric.dimensions
-    dimensions["job_run_ocid"] = os.environ.get("JOB_RUN_OCID")
-    dimensions["job_ocid"] = os.environ.get("JOB_OCID")
-    return oci.monitoring.models.MetricDataDetails(
-        namespace=METRIC_NAMESPACE,
-        compartment_id=METRIC_COMPARTMENT,
-        name=metric.name,
-        dimensions=dimensions,
-        datapoints=[
-            oci.monitoring.models.Datapoint(
-                timestamp=timestamp, value=metric.value, count=1
-            )
-        ],
-    )
-
-
-def submit_metrics(client: oci.monitoring.MonitoringClient) -> None:
-    """
-    Submit metrics to the Monitoring Service
-
-    Parameters
-    ----------
-    client: oci.monitoring.MonitoringClient
-        The OCI Monitoring Service client
-    """
-    metric_data_details = []
-    timestamp = datetime.datetime.now()
-    for provider in metric_providers:
-        for metric in provider.get_metrics():
-            metric_data_details.append(
-                convert_to_metric_data_details(metric, timestamp)
-            )
-    post_metric_details = oci.monitoring.models.PostMetricDataDetails(
-        metric_data=metric_data_details
-    )
-    client.post_metric_data(post_metric_data_details=post_metric_details)
-
-
-def collect_metrics():
-    if JOB_RUN_COMPARTMENT_OCID_ENV not in os.environ:
-        return
-    try:
-        signer = oci.auth.signers.get_resource_principals_signer()
-    except EnvironmentError:
-        return
-    client = oci.monitoring.MonitoringClient(
-        config={},
-        signer=signer,
-        # Metrics should be submitted with the "telemetry-ingestion" endpoint instead.
-        # See note here: https://docs.oracle.com/iaas/api/#/en/monitoring/20180401/MetricData/PostMetricData
-        service_endpoint=f"https://telemetry-ingestion.{signer.region}.oraclecloud.com",
-    )
-
-    while True:
-        try:
-            submit_metrics(client)
-        except:
-            pass
-        time.sleep(METRIC_SUBMISSION_INTERVAL_SECONDS)

ads/jobs/utils.py

@@ -1,104 +0,0 @@
-#!/usr/bin/env python
-# -*- coding: utf-8; -*-
-
-# Copyright (c) 2022, 2023 Oracle and/or its affiliates.
-# Licensed under the Universal Permissive License v 1.0 as shown at https://oss.oracle.com/licenses/upl/
-
-
-import os
-import json
-
-from oci.data_flow.models import Application
-
-from ads.opctl.config.utils import read_from_ini
-from ads.opctl.constants import (
-    ADS_DATAFLOW_CONFIG_FILE_NAME,
-    DEFAULT_ADS_CONFIG_FOLDER,
-)
-
-from ads.jobs import logger
-from ads.common.utils import oci_key_profile
-
-
-def get_dataflow_config(path=None, oci_profile=None):
-    if path:
-        dataflow_config_file_path = os.path.abspath(os.path.expanduser(path))
-    else:
-        dataflow_config_file_path = os.path.expanduser(
-            os.path.join(DEFAULT_ADS_CONFIG_FOLDER, ADS_DATAFLOW_CONFIG_FILE_NAME)
-        )
-    config = {}
-    if os.path.exists(dataflow_config_file_path):
-        parser = read_from_ini(dataflow_config_file_path)
-        if not oci_profile:
-            oci_profile = oci_key_profile()
-        if oci_profile in parser:
-            config = dict(parser[oci_profile])
-        if len(config) == 0:
-            logger.error(
-                f"Dataflow configuration with profile {oci_profile} not found."
-            )
-            raise ValueError(
-                f"Dataflow configuration with profile {oci_profile} not found."
-            )
-        return config
-    else:
-        logger.warning(f"{dataflow_config_file_path} not found. Follow this link https://accelerated-data-science.readthedocs.io/en/latest/user_guide/apachespark/dataflow.html to set up the config.")
-        return {}
-
-
-class DataFlowConfig(Application):
-    def __init__(self, path: str = None, oci_profile: str = None):
-        """Create a DataFlowConfig object. If a path to config file is given it is loaded from the path.
-
-        Parameters
-        ----------
-        path : str, optional
-            path to configuration file, by default None
-        oci_profile : str, optional
-            oci profile to use, by default None
-        """
-        self.config = get_dataflow_config(path, oci_profile)
-        self._script_bucket = None
-        self._archive_bucket = None
-        if len(self.config) > 0:
-            self._script_bucket = self.config.pop("script_bucket")
-            self._archive_bucket = self.config.pop("archive_bucket", None)
-        super().__init__(**self.config)
-
-    def __repr__(self):
-        config = json.loads(super().__repr__())
-        config["script_bucket"] = self.script_bucket
-        if self.archive_bucket:
-            config["archive_bucket"] = self.archive_bucket
-        return f"'{json.dumps({k: v for k, v in config.items() if v is not None})}'"
-
-    @property
-    def script_bucket(self):
-        """Bucket to save user script. Also accept a prefix in the format of oci://<bucket-name>@<namespace>/<prefix>.
-
-        Returns
-        -------
-        str
-            script bucket (path)
-        """
-        return self._script_bucket
-
-    @script_bucket.setter
-    def script_bucket(self, v: str):
-        self._script_bucket = v
-
-    @property
-    def archive_bucket(self):
-        """Bucket to save archive zip. Also accept a prefix in the format of oci://<bucket-name>@<namespace>/<prefix>.
-
-        Returns
-        -------
-        str :
-            archive bucket (path)
-        """
-        return self._archive_bucket
-
-    @archive_bucket.setter
-    def archive_bucket(self, v: str):
-        self._archive_bucket = v

ads/llm/__init__.py

@@ -1,28 +0,0 @@
-#!/usr/bin/env python
-
-# Copyright (c) 2025 Oracle and/or its affiliates.
-# Licensed under the Universal Permissive License v 1.0 as shown at https://oss.oracle.com/licenses/upl/
-
-try:
-    import langchain
-
-    from ads.llm.chat_template import ChatTemplates
-    from ads.llm.langchain.plugins.chat_models.oci_data_science import (
-        ChatOCIModelDeployment,
-        ChatOCIModelDeploymentTGI,
-        ChatOCIModelDeploymentVLLM,
-    )
-    from ads.llm.langchain.plugins.embeddings.oci_data_science_model_deployment_endpoint import (
-        OCIDataScienceEmbedding,
-    )
-    from ads.llm.langchain.plugins.llms.oci_data_science_model_deployment_endpoint import (
-        OCIModelDeploymentLLM,
-        OCIModelDeploymentTGI,
-        OCIModelDeploymentVLLM,
-    )
-except ImportError as ex:
-    if ex.name == "langchain":
-        raise ImportError(
-            f"{ex.msg}\nPlease install/update langchain with `pip install langchain -U`"
-        ) from ex
-    raise ex

ads/llm/autogen/__init__.py

@@ -1,2 +0,0 @@
-# Copyright (c) 2024, Oracle and/or its affiliates.  All rights reserved.
-# Licensed under the Universal Permissive License v 1.0 as shown at https://oss.oracle.com/licenses/upl/

ads/llm/autogen/constants.py

@@ -1,15 +0,0 @@
-# Copyright (c) 2024, Oracle and/or its affiliates.  All rights reserved.
-# Licensed under the Universal Permissive License v 1.0 as shown at https://oss.oracle.com/licenses/upl/
-
-
-class Events:
-    KEY = "event_name"
-
-    EXCEPTION = "exception"
-    LLM_CALL = "llm_call"
-    TOOL_CALL = "tool_call"
-    NEW_AGENT = "new_agent"
-    NEW_CLIENT = "new_client"
-    RECEIVED_MESSAGE = "received_message"
-    SESSION_START = "logging_session_start"
-    SESSION_STOP = "logging_session_stop"

ads/llm/autogen/reports/__init__.py

@@ -1,2 +0,0 @@
-# Copyright (c) 2024, Oracle and/or its affiliates.  All rights reserved.
-# Licensed under the Universal Permissive License v 1.0 as shown at https://oss.oracle.com/licenses/upl/

ads/llm/autogen/reports/base.py

@@ -1,67 +0,0 @@
-# Copyright (c) 2024, Oracle and/or its affiliates.  All rights reserved.
-# Licensed under the Universal Permissive License v 1.0 as shown at https://oss.oracle.com/licenses/upl/
-import json
-import logging
-import os
-
-from jinja2 import Environment, FileSystemLoader
-
-logger = logging.getLogger(__name__)
-
-
-class BaseReport:
-    """Base class containing utilities for generating reports."""
-
-    @staticmethod
-    def format_json_string(s) -> str:
-        """Formats the JSON string in markdown."""
-        return f"```json\n{json.dumps(json.loads(s), indent=2)}\n```"
-
-    @staticmethod
-    def _parse_date_time(datetime_string: str):
-        """Parses a datetime string in the logs into date and time.
-        Keeps only the seconds in the time.
-        """
-        date_str, time_str = datetime_string.split(" ", 1)
-        time_str = time_str.split(".", 1)[0]
-        return date_str, time_str
-
-    @staticmethod
-    def _preview_message(message: str, max_length=30) -> str:
-        """Shows the beginning part of a string message."""
-        # Return the entire string if it is less than the max_length
-        if len(message) <= max_length:
-            return message
-        # Go backward until we find the first whitespace
-        idx = 30
-        while not message[idx].isspace() and idx > 0:
-            idx -= 1
-        # If we found a whitespace
-        if idx > 0:
-            return message[:idx] + "..."
-        # If we didn't find a whitespace
-        return message[:30] + "..."
-
-    @classmethod
-    def _render_template(cls, template_path, **kwargs) -> str:
-        """Render Jinja template with kwargs."""
-        template_dir = os.path.join(os.path.dirname(__file__), "templates")
-        environment = Environment(
-            loader=FileSystemLoader(template_dir), autoescape=True
-        )
-        template = environment.get_template(template_path)
-        try:
-            html = template.render(**kwargs)
-        except Exception:
-            logger.error(
-                "Unable to render template %s with data:\n%s",
-                template_path,
-                str(kwargs),
-            )
-            return cls._render_template(
-                template_path=template_path,
-                sender=kwargs.get("sender", "N/A"),
-                content="TEMPLATE RENDER ERROR",
-                timestamp=kwargs.get("timestamp", ""),
-            )
-        return html

ads/llm/autogen/reports/data.py

@@ -1,103 +0,0 @@
-#!/usr/bin/env python
-# Copyright (c) 2024 Oracle and/or its affiliates.
-# Licensed under the Universal Permissive License v 1.0 as shown at https://oss.oracle.com/licenses/upl/
-"""Contains the data structure for logging and reporting."""
-import copy
-import json
-from dataclasses import asdict, dataclass, field
-from typing import Optional, Union
-
-from ads.llm.autogen.constants import Events
-
-
-@dataclass
-class LogData:
-    """Base class for the data field of LogRecord."""
-
-    def to_dict(self):
-        """Convert the log data to dictionary."""
-        return asdict(self)
-
-
-@dataclass
-class LogRecord:
-    """Represents a log record.
-
-    The `data` field is for pre-defined structured data, which should be an instance of LogData.
-    The `kwargs` field is for freeform key value pairs.
-    """
-
-    session_id: str
-    thread_id: int
-    timestamp: str
-    event_name: str
-    source_id: Optional[int] = None
-    source_name: Optional[str] = None
-    # Structured data for specific type of logs
-    data: Optional[LogData] = None
-    # Freeform data
-    kwargs: dict = field(default_factory=dict)
-
-    def to_dict(self):
-        """Convert the log record to dictionary."""
-        return asdict(self)
-
-    def to_string(self):
-        """Serialize the log record to JSON string."""
-        return json.dumps(self.to_dict(), default=str)
-
-    @classmethod
-    def from_dict(cls, data: dict) -> "LogRecord":
-        """Initializes a LogRecord object from dictionary."""
-        event_mapping = {
-            Events.NEW_AGENT: AgentData,
-            Events.TOOL_CALL: ToolCallData,
-            Events.LLM_CALL: LLMCompletionData,
-        }
-        if Events.KEY not in data:
-            raise KeyError("event_name not found in data.")
-
-        data = copy.deepcopy(data)
-
-        event_name = data["event_name"]
-        if event_name in event_mapping and data.get("data"):
-            data["data"] = event_mapping[event_name](**data.pop("data"))
-
-        return cls(**data)
-
-
-@dataclass
-class AgentData(LogData):
-    """Represents agent log Data."""
-
-    agent_name: str
-    agent_class: str
-    agent_module: Optional[str] = None
-    is_manager: Optional[bool] = None
-
-
-@dataclass
-class LLMCompletionData(LogData):
-    """Represents LLM completion log data."""
-
-    invocation_id: str
-    request: dict
-    response: dict
-    start_time: str
-    end_time: str
-    cost: Optional[float] = None
-    is_cached: Optional[bool] = None
-
-
-@dataclass
-class ToolCallData(LogData):
-    """Represents tool call log data."""
-
-    tool_name: str
-    start_time: str
-    end_time: str
-    agent_name: str
-    agent_class: str
-    agent_module: Optional[str] = None
-    input_args: dict = field(default_factory=dict)
-    returns: Optional[Union[str, list, dict, tuple]] = None