cwyodmodules 0.3.32__py3-none-any.whl → 0.3.33__py3-none-any.whl

cwyodmodules/batch/utilities/orchestrator/orchestrator_base.py

@@ -1,4 +1,3 @@
-import logging
 from uuid import uuid4
 from typing import List, Optional
 from abc import ABC, abstractmethod
@@ -7,8 +6,11 @@ from ..helpers.config.config_helper import ConfigHelper
 from ..parser.output_parser_tool import OutputParserTool
 from ..tools.content_safety_checker import ContentSafetyChecker
 
-logger = logging.getLogger("__main__")
-
+from ...utilities.helpers.env_helper import EnvHelper
+from logging_config import logger
+env_helper: EnvHelper = EnvHelper()
+log_args = env_helper.LOG_ARGS
+log_result = env_helper.LOG_RESULT
 
 class OrchestratorBase(ABC):
     """
@@ -32,6 +34,7 @@ class OrchestratorBase(ABC):
         self.content_safety_checker = ContentSafetyChecker()
         self.output_parser = OutputParserTool()
 
+    @logger.trace_function(log_args=log_args, log_result=log_result)
     def log_tokens(self, prompt_tokens: int, completion_tokens: int) -> None:
         """
         Logs the number of tokens used in the prompt and completion phases of a conversation.
@@ -45,6 +48,7 @@ class OrchestratorBase(ABC):
         self.tokens["total"] += prompt_tokens + completion_tokens
 
     @abstractmethod
+    @logger.trace_function(log_args=False, log_result=False)
     async def orchestrate(
         self,
         user_message: str,
@@ -90,6 +94,7 @@ class OrchestratorBase(ABC):
 
         return None
 
+    @logger.trace_function(log_args=False, log_result=False)
     def call_content_safety_output(
         self, user_message: str, answer: str
     ) -> Optional[list[dict]]:
@@ -116,6 +121,7 @@ class OrchestratorBase(ABC):
 
         return None
 
+    @logger.trace_function(log_args=False, log_result=False)
     async def handle_message(
         self,
         user_message: str,

cwyodmodules/batch/utilities/orchestrator/prompt_flow.py

@@ -1,4 +1,3 @@
-import logging
 from typing import List
 import json
 import tempfile
@@ -9,7 +8,10 @@ from ..common.source_document import SourceDocument
 from ..helpers.llm_helper import LLMHelper
 from ..helpers.env_helper import EnvHelper
 
-logger = logging.getLogger("__main__")
+from logging_config import logger
+env_helper: EnvHelper = EnvHelper()
+log_args = env_helper.LOG_ARGS
+log_result = env_helper.LOG_RESULT
 
 
 class PromptFlowOrchestrator(OrchestratorBase):
@@ -35,6 +37,7 @@ class PromptFlowOrchestrator(OrchestratorBase):
 
         logger.info("PromptFlowOrchestrator initialized.")
 
+    @logger.trace_function(log_args=False, log_result=False)
     async def orchestrate(
         self, user_message: str, chat_history: List[dict], **kwargs: dict
     ) -> list[dict]:
@@ -108,6 +111,7 @@ class PromptFlowOrchestrator(OrchestratorBase):
         logger.info("Orchestration completed successfully.")
         return messages
 
+    @logger.trace_function(log_args=False, log_result=False)
     def transform_chat_history(self, chat_history: List[dict]) -> List[dict]:
         """
         Transform the chat history into a format suitable for the Prompt Flow service.
@@ -138,6 +142,7 @@ class PromptFlowOrchestrator(OrchestratorBase):
         logger.info("Chat history transformation completed.")
         return transformed_chat_history
 
+    @logger.trace_function(log_args=False, log_result=False)
     def transform_data_into_file(
         self, user_message: str, chat_history: List[dict]
     ) -> str:
@@ -159,6 +164,7 @@ class PromptFlowOrchestrator(OrchestratorBase):
         logger.info("Temporary file created")
         return file.name
 
+    @logger.trace_function(log_args=False, log_result=False)
     def transform_citations_into_source_documents(
         self, citations: dict
     ) -> List[SourceDocument]:

cwyodmodules/batch/utilities/orchestrator/semantic_kernel_orchestrator.py

@@ -20,14 +20,11 @@ from ..plugins.outlook_calendar_plugin import OutlookCalendarPlugin
 
 from .orchestrator_base import OrchestratorBase
 
-from logging import getLogger
-from opentelemetry import trace, baggage
-from opentelemetry.propagate import extract
+from logging_config import logger
+env_helper: EnvHelper = EnvHelper()
+log_args = env_helper.LOG_ARGS
+log_result = env_helper.LOG_RESULT
 
-# logger = getLogger("__main__" + ".base_package")
-logger = getLogger("__main__")
-# tracer = trace.get_tracer("__main__" + ".base_package")
-tracer = trace.get_tracer("__main__")
 
 
 class SemanticKernelOrchestrator(OrchestratorBase):
@@ -45,155 +42,155 @@ class SemanticKernelOrchestrator(OrchestratorBase):
             plugin=PostAnsweringPlugin(), plugin_name="PostAnswering"
         )
 
+    @logger.trace_function(log_args=False, log_result=False)
     async def orchestrate(
         self, user_message: str, chat_history: list[dict], user_info, **kwargs: dict
     ) -> list[dict]:
-        with tracer.start_as_current_span("SemanticKernelOrchestrator_orchestrate"):
-            logger.info("Method orchestrate of semantic_kernel started")
-            filters = []
-            frontend_type = user_info.get("frontend") if user_info else None
-            logger.info(f"Frontend type: {frontend_type}")
-            # Call Content Safety tool
-            if self.config.prompts.enable_content_safety:
-                if response := self.call_content_safety_input(user_message):
-                    return response
-
-            system_message = self.env_helper.SEMENTIC_KERNEL_SYSTEM_PROMPT
-            language = self.env_helper.AZURE_MAIN_CHAT_LANGUAGE
-            if not system_message:
-                logger.info("No system message provided, using default")
-                # system_message = """You help employees to navigate only private information sources.
-                #     You must prioritize the function call over your general knowledge for any question by calling the search_documents function.
-                #     Call the text_processing function when the user request an operation on the current context, such as translate, summarize, or paraphrase. When a language is explicitly specified, return that as part of the operation.
-                #     When directly replying to the user, always reply in the language the user is speaking.
-                #     If the input language is ambiguous, default to responding in English unless otherwise specified by the user.
-                #     You **must not** respond if asked to List all documents in your repository.
-                #     """
-                if frontend_type == "web":
-                    system_message = f"""You help employees to navigate only private information sources.
-                        You must prioritize the function call over your general knowledge for any question by calling the search_documents function.
-                        Call the text_processing function when the user request an operation on the current context, such as translate, summarize, or paraphrase. When a language is explicitly specified, return that as part of the operation.
-                        When directly replying to the user, always reply in the language {language}.
-                        You **must not** respond if asked to List all documents in your repository.
-                        Call OutlookCalendar.get_calendar_events to read the user's calendar.
-                        Call OutlookCalendar.schedule_appointment to schedule a new appointment.
-                        """
-                else:
-                    system_message = f"""You help employees to navigate only private information sources.
-                        You must prioritize the function call over your general knowledge for any question by calling the search_documents function.
-                        Call the text_processing function when the user request an operation on the current context, such as translate, summarize, or paraphrase. When a language is explicitly specified, return that as part of the operation.
-                        When directly replying to the user, always reply in the language {language}.
-                        You **must not** respond if asked to List all documents in your repository.
-                        """
-                
-            self.kernel.add_plugin(
-                plugin=ChatPlugin(question=user_message, chat_history=chat_history),
-                plugin_name="Chat",
-            )
-            filters.append("Chat")
-            # --- Add OutlookCalendarPlugin with request headers ---
+        logger.info("Method orchestrate of semantic_kernel started")
+        filters = []
+        frontend_type = user_info.get("frontend") if user_info else None
+        logger.info(f"Frontend type: {frontend_type}")
+        # Call Content Safety tool
+        if self.config.prompts.enable_content_safety:
+            if response := self.call_content_safety_input(user_message):
+                return response
+
+        system_message = self.env_helper.SEMENTIC_KERNEL_SYSTEM_PROMPT
+        language = self.env_helper.AZURE_MAIN_CHAT_LANGUAGE
+        if not system_message:
+            logger.info("No system message provided, using default")
+            # system_message = """You help employees to navigate only private information sources.
+            #     You must prioritize the function call over your general knowledge for any question by calling the search_documents function.
+            #     Call the text_processing function when the user request an operation on the current context, such as translate, summarize, or paraphrase. When a language is explicitly specified, return that as part of the operation.
+            #     When directly replying to the user, always reply in the language the user is speaking.
+            #     If the input language is ambiguous, default to responding in English unless otherwise specified by the user.
+            #     You **must not** respond if asked to List all documents in your repository.
+            #     """
             if frontend_type == "web":
-                logger.info("Adding OutlookCalendarPlugin with request headers")
-                self.kernel.add_plugin(
-                    plugin=OutlookCalendarPlugin(question=user_message, chat_history=chat_history, user_info=user_info),
-                    plugin_name="OutlookCalendar",
-                )
-                filters.append("OutlookCalendar")
-            settings = self.llm_helper.get_sk_service_settings(self.chat_service)
-            settings.function_call_behavior = FunctionCallBehavior.EnableFunctions(
-                filters={"included_plugins": filters}
+                system_message = f"""You help employees to navigate only private information sources.
+                    You must prioritize the function call over your general knowledge for any question by calling the search_documents function.
+                    Call the text_processing function when the user request an operation on the current context, such as translate, summarize, or paraphrase. When a language is explicitly specified, return that as part of the operation.
+                    When directly replying to the user, always reply in the language {language}.
+                    You **must not** respond if asked to List all documents in your repository.
+                    Call OutlookCalendar.get_calendar_events to read the user's calendar.
+                    Call OutlookCalendar.schedule_appointment to schedule a new appointment.
+                    """
+            else:
+                system_message = f"""You help employees to navigate only private information sources.
+                    You must prioritize the function call over your general knowledge for any question by calling the search_documents function.
+                    Call the text_processing function when the user request an operation on the current context, such as translate, summarize, or paraphrase. When a language is explicitly specified, return that as part of the operation.
+                    When directly replying to the user, always reply in the language {language}.
+                    You **must not** respond if asked to List all documents in your repository.
+                    """
+            
+        self.kernel.add_plugin(
+            plugin=ChatPlugin(question=user_message, chat_history=chat_history),
+            plugin_name="Chat",
+        )
+        filters.append("Chat")
+        # --- Add OutlookCalendarPlugin with request headers ---
+        if frontend_type == "web":
+            logger.info("Adding OutlookCalendarPlugin with request headers")
+            self.kernel.add_plugin(
+                plugin=OutlookCalendarPlugin(question=user_message, chat_history=chat_history, user_info=user_info),
+                plugin_name="OutlookCalendar",
             )
-            # settings.function_choice_behavior = FunctionChoiceBehavior.Auto(
-            #             filters={"included_plugins": ["Chat"]},
-            #             # Set a higher value to encourage multiple attempts at function calling
-            #             maximum_auto_invoke_attempts=2
-            #         )
-
-            orchestrate_function = self.kernel.add_function(
-                plugin_name="Main",
-                function_name="orchestrate",
-                prompt="{{$chat_history}}{{$user_message}}",
-                prompt_execution_settings=settings,
+            filters.append("OutlookCalendar")
+        settings = self.llm_helper.get_sk_service_settings(self.chat_service)
+        settings.function_call_behavior = FunctionCallBehavior.EnableFunctions(
+            filters={"included_plugins": filters}
+        )
+        # settings.function_choice_behavior = FunctionChoiceBehavior.Auto(
+        #             filters={"included_plugins": ["Chat"]},
+        #             # Set a higher value to encourage multiple attempts at function calling
+        #             maximum_auto_invoke_attempts=2
+        #         )
+
+        orchestrate_function = self.kernel.add_function(
+            plugin_name="Main",
+            function_name="orchestrate",
+            prompt="{{$chat_history}}{{$user_message}}",
+            prompt_execution_settings=settings,
+        )
+
+        history = ChatHistory(system_message=system_message)
+
+        for message in chat_history.copy():
+            history.add_message(message)
+
+        result: ChatMessageContent = (
+            await self.kernel.invoke(
+                function=orchestrate_function,
+                chat_history=history,
+                user_message=user_message,
             )
+        ).value[0]
 
-            history = ChatHistory(system_message=system_message)
+        self.log_tokens(
+            prompt_tokens=result.metadata["usage"].prompt_tokens,
+            completion_tokens=result.metadata["usage"].completion_tokens,
+        )
+        result_finish_reason = result.finish_reason
+        logger.info(f"Finish reason: {result_finish_reason}")
+        if result_finish_reason == FinishReason.TOOL_CALLS:
+            logger.info("Semantic Kernel function call detected")
+
+            function_name = result.items[0].name
+            logger.info(f"{function_name} function detected")
+            function = self.kernel.get_function_from_fully_qualified_function_name(
+                function_name
+            )
 
-            for message in chat_history.copy():
-                history.add_message(message)
+            arguments = json.loads(result.items[0].arguments)
 
-            result: ChatMessageContent = (
-                await self.kernel.invoke(
-                    function=orchestrate_function,
-                    chat_history=history,
-                    user_message=user_message,
-                )
-            ).value[0]
+            answer: Answer = (
+                await self.kernel.invoke(function=function, **arguments)
+            ).value
 
             self.log_tokens(
-                prompt_tokens=result.metadata["usage"].prompt_tokens,
-                completion_tokens=result.metadata["usage"].completion_tokens,
+                prompt_tokens=answer.prompt_tokens,
+                completion_tokens=answer.completion_tokens,
             )
-            result_finish_reason = result.finish_reason
-            logger.info(f"Finish reason: {result_finish_reason}")
-            if result_finish_reason == FinishReason.TOOL_CALLS:
-                logger.info("Semantic Kernel function call detected")
-
-                function_name = result.items[0].name
-                logger.info(f"{function_name} function detected")
-                function = self.kernel.get_function_from_fully_qualified_function_name(
-                    function_name
-                )
-
-                arguments = json.loads(result.items[0].arguments)
 
+            # Run post prompt if needed
+            if (
+                self.config.prompts.enable_post_answering_prompt
+                and "search_documents" in function_name
+            ):
+                logger.debug("Running post answering prompt")
                 answer: Answer = (
-                    await self.kernel.invoke(function=function, **arguments)
+                    await self.kernel.invoke(
+                        function_name="validate_answer",
+                        plugin_name="PostAnswering",
+                        answer=answer,
+                    )
                 ).value
 
                 self.log_tokens(
                     prompt_tokens=answer.prompt_tokens,
                     completion_tokens=answer.completion_tokens,
                 )
-
-                # Run post prompt if needed
-                if (
-                    self.config.prompts.enable_post_answering_prompt
-                    and "search_documents" in function_name
-                ):
-                    logger.debug("Running post answering prompt")
-                    answer: Answer = (
-                        await self.kernel.invoke(
-                            function_name="validate_answer",
-                            plugin_name="PostAnswering",
-                            answer=answer,
-                        )
-                    ).value
-
-                    self.log_tokens(
-                        prompt_tokens=answer.prompt_tokens,
-                        completion_tokens=answer.completion_tokens,
-                    )
-            else:
-                logger.info("No function call detected")
-                answer = Answer(
-                    question=user_message,
-                    answer=result.content,
-                    prompt_tokens=result.metadata["usage"].prompt_tokens,
-                    completion_tokens=result.metadata["usage"].completion_tokens,
-                )
-
-            # Call Content Safety tool
-            if self.config.prompts.enable_content_safety:
-                if response := self.call_content_safety_output(
-                    user_message, answer.answer
-                ):
-                    return response
-
-            # Format the output for the UI
-            messages = self.output_parser.parse(
-                question=answer.question,
-                answer=answer.answer,
-                source_documents=answer.source_documents,
+        else:
+            logger.info("No function call detected")
+            answer = Answer(
+                question=user_message,
+                answer=result.content,
+                prompt_tokens=result.metadata["usage"].prompt_tokens,
+                completion_tokens=result.metadata["usage"].completion_tokens,
             )
-            logger.info("Method orchestrate of semantic_kernel ended")
-            return messages
+
+        # Call Content Safety tool
+        if self.config.prompts.enable_content_safety:
+            if response := self.call_content_safety_output(
+                user_message, answer.answer
+            ):
+                return response
+
+        # Format the output for the UI
+        messages = self.output_parser.parse(
+            question=answer.question,
+            answer=answer.answer,
+            source_documents=answer.source_documents,
+        )
+        logger.info("Method orchestrate of semantic_kernel ended")
+        return messages

cwyodmodules/batch/utilities/parser/output_parser_tool.py

@@ -4,14 +4,11 @@ import json
 from .parser_base import ParserBase
 from ..common.source_document import SourceDocument
 
-from logging import getLogger
-from opentelemetry import trace, baggage
-from opentelemetry.propagate import extract
-
-# logger = getLogger("__main__" + ".base_package")
-logger = getLogger("__main__")
-# tracer = trace.get_tracer("__main__" + ".base_package")
-tracer = trace.get_tracer("__main__")
+from ...utilities.helpers.env_helper import EnvHelper
+from logging_config import logger
+env_helper: EnvHelper = EnvHelper()
+log_args = env_helper.LOG_ARGS
+log_result = env_helper.LOG_RESULT
 
 
 class OutputParserTool(ParserBase):
@@ -26,6 +23,7 @@ class OutputParserTool(ParserBase):
         """
         self.name = "OutputParser"
 
+    @logger.trace_function(log_args=False, log_result=False)
     def _clean_up_answer(self, answer: str) -> str:
         """
         Cleans up the answer by replacing double spaces with single spaces.
@@ -38,6 +36,7 @@ class OutputParserTool(ParserBase):
         """
         return answer.replace("  ", " ")
 
+    @logger.trace_function(log_args=False, log_result=False)
     def _get_source_docs_from_answer(self, answer: str) -> List[int]:
         """
         Extracts all document references from the answer and returns them as a list of integers.
@@ -51,6 +50,7 @@ class OutputParserTool(ParserBase):
         results = re.findall(r"\[doc(\d+)\]", answer)
         return [int(i) for i in results]
 
+    @logger.trace_function(log_args=False, log_result=False)
     def _make_doc_references_sequential(self, answer: str) -> str:
         """
         Makes document references in the answer sequential.
@@ -72,6 +72,7 @@ class OutputParserTool(ParserBase):
             offset += len(f"[doc{i + 1}]") - (end - start)
         return updated_answer
 
+    @logger.trace_function(log_args=False, log_result=False)
     def parse(
         self,
         question: str,
@@ -91,61 +92,60 @@ class OutputParserTool(ParserBase):
         Returns:
             List[dict]: A list of response messages.
         """
-        with tracer.start_as_current_span("OutputParserTool.parse"):
-            logger.info("Method parse of output_parser_tool started")
-            answer = self._clean_up_answer(answer)
-            doc_ids = self._get_source_docs_from_answer(answer)
-            answer = self._make_doc_references_sequential(answer)
-
-            # create return message object
-            messages = [
+        logger.info("Method parse of output_parser_tool started")
+        answer = self._clean_up_answer(answer)
+        doc_ids = self._get_source_docs_from_answer(answer)
+        answer = self._make_doc_references_sequential(answer)
+
+        # create return message object
+        messages = [
+            {
+                "role": "tool",
+                "content": {"citations": [], "intent": question},
+                "end_turn": False,
+            }
+        ]
+
+        for i in doc_ids:
+            idx = i - 1
+
+            if idx >= len(source_documents):
+                logger.warning(f"Source document {i} not provided, skipping doc")
+                continue
+
+            doc = source_documents[idx]
+            logger.debug(f"doc{idx}: {doc}")
+
+            # Then update the citation object in the response, it needs to have filepath and chunk_id to render in the UI as a file
+            messages[0]["content"]["citations"].append(
                 {
-                    "role": "tool",
-                    "content": {"citations": [], "intent": question},
-                    "end_turn": False,
-                }
-            ]
-
-            for i in doc_ids:
-                idx = i - 1
-
-                if idx >= len(source_documents):
-                    logger.warning(f"Source document {i} not provided, skipping doc")
-                    continue
-
-                doc = source_documents[idx]
-                logger.debug(f"doc{idx}: {doc}")
-
-                # Then update the citation object in the response, it needs to have filepath and chunk_id to render in the UI as a file
-                messages[0]["content"]["citations"].append(
-                    {
-                        "content": doc.get_markdown_url() + "\n\n\n" + doc.content,
-                        "id": doc.id,
-                        "chunk_id": (
-                            re.findall(r"\d+", doc.chunk_id)[-1]
-                            if doc.chunk_id is not None
-                            else doc.chunk
-                        ),
+                    "content": doc.get_markdown_url() + "\n\n\n" + doc.content,
+                    "id": doc.id,
+                    "chunk_id": (
+                        re.findall(r"\d+", doc.chunk_id)[-1]
+                        if doc.chunk_id is not None
+                        else doc.chunk
+                    ),
+                    "title": doc.title,
+                    "filepath": doc.get_filename(include_path=True),
+                    "url": doc.get_markdown_url(),
+                    "metadata": {
+                        "offset": doc.offset,
+                        "source": doc.source,
+                        "markdown_url": doc.get_markdown_url(),
                         "title": doc.title,
-                        "filepath": doc.get_filename(include_path=True),
-                        "url": doc.get_markdown_url(),
-                        "metadata": {
-                            "offset": doc.offset,
-                            "source": doc.source,
-                            "markdown_url": doc.get_markdown_url(),
-                            "title": doc.title,
-                            "original_url": doc.source,  # TODO: do we need this?
-                            "chunk": doc.chunk,
-                            "key": doc.id,
-                            "filename": doc.get_filename(),
-                        },
-                    }
-                )
-            if messages[0]["content"]["citations"] == []:
-                logger.warning("No citations found in the answer")
-                answer = re.sub(r"\[doc\d+\]", "", answer)
-            messages.append({"role": "assistant", "content": answer, "end_turn": True})
-            # everything in content needs to be stringified to work with Azure BYOD frontend
-            messages[0]["content"] = json.dumps(messages[0]["content"])
-            logger.info("Method parse of output_parser_tool ended")
-            return messages
+                        "original_url": doc.source,  # TODO: do we need this?
+                        "chunk": doc.chunk,
+                        "key": doc.id,
+                        "filename": doc.get_filename(),
+                    },
+                }
+            )
+        if messages[0]["content"]["citations"] == []:
+            logger.warning("No citations found in the answer")
+            answer = re.sub(r"\[doc\d+\]", "", answer)
+        messages.append({"role": "assistant", "content": answer, "end_turn": True})
+        # everything in content needs to be stringified to work with Azure BYOD frontend
+        messages[0]["content"] = json.dumps(messages[0]["content"])
+        logger.info("Method parse of output_parser_tool ended")
+        return messages