oracle-ads 2.12.9__py3-none-any.whl → 2.12.10rc0__py3-none-any.whl

ads/aqua/model/model.py

@@ -1,5 +1,5 @@
 #!/usr/bin/env python
-# Copyright (c) 2024 Oracle and/or its affiliates.
+# Copyright (c) 2024, 2025 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 pathlib
@@ -15,6 +15,7 @@ from oci.data_science.models import JobRun, Metadata, Model, UpdateModelDetails
 from ads.aqua import ODSC_MODEL_COMPARTMENT_OCID, logger
 from ads.aqua.app import AquaApp
 from ads.aqua.common.enums import (
+    CustomInferenceContainerTypeFamily,
     FineTuningContainerTypeFamily,
     InferenceContainerTypeFamily,
     Tags,
@@ -23,6 +24,7 @@ from ads.aqua.common.errors import AquaRuntimeError, AquaValueError
 from ads.aqua.common.utils import (
     LifecycleStatus,
     _build_resource_identifier,
+    cleanup_local_hf_model_artifact,
     copy_model_config,
     create_word_icon,
     generate_tei_cmd_var,
@@ -376,8 +378,10 @@ class AquaModelApp(AquaApp):
                 f"Failed to delete model:{model_id}. Only registered models or finetuned model can be deleted."
             )
 
-    @telemetry(entry_point="plugin=model&action=delete", name="aqua")
-    def edit_registered_model(self, id, inference_container, enable_finetuning, task):
+    @telemetry(entry_point="plugin=model&action=edit", name="aqua")
+    def edit_registered_model(
+        self, id, inference_container, inference_container_uri, enable_finetuning, task
+    ):
         """Edits the default config of unverified registered model.
 
         Parameters
@@ -386,6 +390,8 @@ class AquaModelApp(AquaApp):
             The model OCID.
         inference_container: str.
             The inference container family name
+        inference_container_uri: str
+            The inference container uri for embedding models
         enable_finetuning: str
             Flag to enable or disable finetuning over the model. Defaults to None
         task:
@@ -401,19 +407,44 @@ class AquaModelApp(AquaApp):
         if ds_model.freeform_tags.get(Tags.BASE_MODEL_CUSTOM, None):
             if ds_model.freeform_tags.get(Tags.AQUA_SERVICE_MODEL_TAG, None):
                 raise AquaRuntimeError(
-                    f"Failed to edit model:{id}. Only registered unverified models can be edited."
+                    "Only registered unverified models can be edited."
                 )
             else:
                 custom_metadata_list = ds_model.custom_metadata_list
                 freeform_tags = ds_model.freeform_tags
                 if inference_container:
-                    custom_metadata_list.add(
-                        key=ModelCustomMetadataFields.DEPLOYMENT_CONTAINER,
-                        value=inference_container,
-                        category=MetadataCustomCategory.OTHER,
-                        description="Deployment container mapping for SMC",
-                        replace=True,
-                    )
+                    if (
+                        inference_container in CustomInferenceContainerTypeFamily
+                        and inference_container_uri is None
+                    ):
+                        raise AquaRuntimeError(
+                            "Inference container URI must be provided."
+                        )
+                    else:
+                        custom_metadata_list.add(
+                            key=ModelCustomMetadataFields.DEPLOYMENT_CONTAINER,
+                            value=inference_container,
+                            category=MetadataCustomCategory.OTHER,
+                            description="Deployment container mapping for SMC",
+                            replace=True,
+                        )
+                if inference_container_uri:
+                    if (
+                        inference_container in CustomInferenceContainerTypeFamily
+                        or inference_container is None
+                    ):
+                        custom_metadata_list.add(
+                            key=ModelCustomMetadataFields.DEPLOYMENT_CONTAINER_URI,
+                            value=inference_container_uri,
+                            category=MetadataCustomCategory.OTHER,
+                            description=f"Inference container URI for {ds_model.display_name}",
+                            replace=True,
+                        )
+                    else:
+                        raise AquaRuntimeError(
+                            f"Inference container URI can be edited only with container values: {CustomInferenceContainerTypeFamily.values()}"
+                        )
+
                 if enable_finetuning is not None:
                     if enable_finetuning.lower() == "true":
                         custom_metadata_list.add(
@@ -448,9 +479,7 @@ class AquaModelApp(AquaApp):
                 )
                 AquaApp().update_model(id, update_model_details)
         else:
-            raise AquaRuntimeError(
-                f"Failed to edit model:{id}. Only registered unverified models can be edited."
-            )
+            raise AquaRuntimeError("Only registered unverified models can be edited.")
 
     def _fetch_metric_from_metadata(
         self,
@@ -869,8 +898,7 @@ class AquaModelApp(AquaApp):
             # only add cmd vars if inference container is not an SMC
             if (
                 inference_container not in smc_container_set
-                and inference_container
-                == InferenceContainerTypeFamily.AQUA_TEI_CONTAINER_FAMILY
+                and inference_container in CustomInferenceContainerTypeFamily.values()
             ):
                 cmd_vars = generate_tei_cmd_var(os_path)
                 metadata.add(
@@ -1322,20 +1350,20 @@ class AquaModelApp(AquaApp):
         Returns
         -------
         model_artifact_path (str): Location where the model artifacts are downloaded.
-
         """
         # Download the model from hub
-        if not local_dir:
-            local_dir = os.path.join(os.path.expanduser("~"), "cached-model")
-        local_dir = os.path.join(local_dir, model_name)
-        os.makedirs(local_dir, exist_ok=True)
-        snapshot_download(
+        if local_dir:
+            local_dir = os.path.join(local_dir, model_name)
+            os.makedirs(local_dir, exist_ok=True)
+
+        # if local_dir is not set, the return value points to the cached data folder
+        local_dir = snapshot_download(
             repo_id=model_name,
             local_dir=local_dir,
             allow_patterns=allow_patterns,
             ignore_patterns=ignore_patterns,
         )
-        # Upload to object storage and skip .cache/huggingface/ folder
+        # Upload to object storage
         model_artifact_path = upload_folder(
             os_path=os_path,
             local_dir=local_dir,
@@ -1365,6 +1393,8 @@ class AquaModelApp(AquaApp):
                 ignore_patterns (list): Model files matching any of the patterns are not downloaded.
                     Example: ["*.json"] will ignore all .json files. ["folder/*"] will ignore all files under `folder`.
                     Patterns are Standard Wildcards (globbing patterns) and rules can be found here: https://docs.python.org/3/library/fnmatch.html
+                cleanup_model_cache (bool): Deletes downloaded files from local machine after model is successfully
+                registered. Set to True by default.
 
         Returns:
             AquaModel:
@@ -1474,6 +1504,14 @@ class AquaModelApp(AquaApp):
             detail=validation_result.telemetry_model_name,
         )
 
+        if (
+            import_model_details.download_from_hf
+            and import_model_details.cleanup_model_cache
+        ):
+            cleanup_local_hf_model_artifact(
+                model_name=model_name, local_dir=import_model_details.local_dir
+            )
+
         return AquaModel(**aqua_model_attributes)
 
     def _if_show(self, model: DataScienceModel) -> bool:

ads/common/auth.py

@@ -424,7 +424,7 @@ def create_signer(
             "signer": signer,
             "client_kwargs": client_kwargs,
         }
-        logger.info(f"Using authentication signer type {type(signer)}.")
+        logger.debug(f"Using authentication signer type {type(signer)}.")
         return signer_dict
     else:
         signer_args = dict(
@@ -492,7 +492,7 @@ def default_signer(client_kwargs: Optional[Dict] = None) -> Dict:
                 **(client_kwargs or {}),
             },
         }
-        logger.info(f"Using authentication signer type {type(signer)}.")
+        logger.debug(f"Using authentication signer type {type(signer)}.")
         return signer_dict
     else:
         signer_args = dict(
@@ -621,7 +621,7 @@ class APIKey(AuthSignerGenerator):
             )
 
         oci.config.validate_config(configuration)
-        logger.info(f"Using 'api_key' authentication.")
+        logger.debug(f"Using 'api_key' authentication.")
         return {
             "config": configuration,
             "signer": oci.signer.Signer(
@@ -684,7 +684,7 @@ class ResourcePrincipal(AuthSignerGenerator):
             "signer": oci.auth.signers.get_resource_principals_signer(),
             "client_kwargs": self.client_kwargs,
         }
-        logger.info(f"Using 'resource_principal' authentication.")
+        logger.debug(f"Using 'resource_principal' authentication.")
         return signer_dict
 
     @staticmethod
@@ -747,7 +747,7 @@ class InstancePrincipal(AuthSignerGenerator):
             ),
             "client_kwargs": self.client_kwargs,
         }
-        logger.info(f"Using 'instance_principal' authentication.")
+        logger.debug(f"Using 'instance_principal' authentication.")
         return signer_dict
 
 
@@ -814,7 +814,7 @@ class SecurityToken(AuthSignerGenerator):
                 oci.config.from_file(self.oci_config_location, self.oci_key_profile)
             )
 
-        logger.info(f"Using 'security_token' authentication.")
+        logger.debug(f"Using 'security_token' authentication.")
 
         for parameter in self.SECURITY_TOKEN_REQUIRED:
             if parameter not in configuration:
@@ -883,7 +883,7 @@ class SecurityToken(AuthSignerGenerator):
                     )
 
         date_time = datetime.fromtimestamp(time_expired).strftime("%Y-%m-%d %H:%M:%S")
-        logger.info(f"Session is valid until {date_time}.")
+        logger.debug(f"Session is valid until {date_time}.")
 
     def _read_security_token_file(self, security_token_file: str) -> str:
         """Reads security token from file.
@@ -1020,10 +1020,10 @@ class OCIAuthContext:
         """
         if self.profile:
             ads.set_auth(auth=AuthType.API_KEY, profile=self.profile)
-            logger.info(f"OCI profile set to {self.profile}")
+            logger.debug(f"OCI profile set to {self.profile}")
         else:
             ads.set_auth(auth=AuthType.RESOURCE_PRINCIPAL)
-            logger.info(f"OCI auth set to resource principal")
+            logger.debug(f"OCI auth set to resource principal")
         return self
 
     def __exit__(self, exc_type, exc_val, exc_tb):

ads/llm/autogen/__init__.py

@@ -0,0 +1,2 @@
+# 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

@@ -0,0 +1,15 @@
+# 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

@@ -0,0 +1,2 @@
+# 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

@@ -0,0 +1,67 @@
+# 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

@@ -0,0 +1,103 @@
+#!/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