monocle-apptrace 0.1.1__py3-none-any.whl → 0.2.0__py3-none-any.whl

monocle_apptrace/utils.py

@@ -3,7 +3,10 @@ import json
 from importlib import import_module
 import os
 from opentelemetry.trace import Span
+from opentelemetry.context import attach, set_value, get_value
 from monocle_apptrace.constants import azure_service_map, aws_service_map
+from json.decoder import JSONDecodeError
+logger = logging.getLogger(__name__)
 
 embedding_model_context = {}
 
@@ -21,12 +24,14 @@ def dont_throw(func):
     """
     # Obtain a logger specific to the function's module
     logger = logging.getLogger(func.__module__)
+
     # pylint: disable=inconsistent-return-statements
     def wrapper(*args, **kwargs):
         try:
             return func(*args, **kwargs)
         except Exception as ex:
             logger.warning("Failed to execute %s, error: %s", func.__name__, str(ex))
+
     return wrapper
 
 def with_tracer_wrapper(func):
@@ -48,33 +53,79 @@ def resolve_from_alias(my_map, alias):
             return my_map[i]
     return None
 
-def load_wrapper_from_config(config_file_path: str, module_name: str = None):
-    wrapper_methods = []
-    with open(config_file_path, encoding='UTF-8') as config_file:
+def load_output_processor(wrapper_method, attributes_config_base_path):
+    """Load the output processor from a file if the file path is provided and valid."""
+    logger = logging.getLogger()
+    output_processor_file_path = wrapper_method["output_processor"][0]
+    logger.info(f'Output processor file path is: {output_processor_file_path}')
+
+    if isinstance(output_processor_file_path, str) and output_processor_file_path:  # Combined condition
+        if not attributes_config_base_path:
+            absolute_file_path = os.path.abspath(output_processor_file_path)
+        else:
+            absolute_file_path = os.path.join(attributes_config_base_path, output_processor_file_path)
+
+        logger.info(f'Absolute file path is: {absolute_file_path}')
+        try:
+            with open(absolute_file_path, encoding='UTF-8') as op_file:
+                wrapper_method["output_processor"] = json.load(op_file)
+                logger.info('Output processor loaded successfully.')
+        except FileNotFoundError:
+            logger.error(f"Error: File not found at {absolute_file_path}.")
+        except JSONDecodeError:
+            logger.error(f"Error: Invalid JSON content in the file {absolute_file_path}.")
+        except Exception as e:
+            logger.error(f"Error: An unexpected error occurred: {e}")
+    else:
+        logger.error("Invalid or missing output processor file path.")
+
+def get_wrapper_methods_config(
+        wrapper_methods_config_path: str,
+        attributes_config_base_path: str = None
+):
+    parent_dir = os.path.dirname(os.path.join(os.path.dirname(__file__), '..'))
+    wrapper_methods_config = load_wrapper_methods_config_from_file(
+        wrapper_methods_config_path=os.path.join(parent_dir, wrapper_methods_config_path))
+    process_wrapper_method_config(
+        wrapper_methods_config=wrapper_methods_config,
+        attributes_config_base_path=attributes_config_base_path)
+    return wrapper_methods_config
+
+def load_wrapper_methods_config_from_file(
+        wrapper_methods_config_path: str):
+    json_data = {}
+
+    with open(wrapper_methods_config_path, encoding='UTF-8') as config_file:
         json_data = json.load(config_file)
-        wrapper_methods = json_data["wrapper_methods"]
-        for wrapper_method in wrapper_methods:
+
+    return json_data["wrapper_methods"]
+
+def process_wrapper_method_config(
+        wrapper_methods_config: str,
+        attributes_config_base_path: str = ""):
+    for wrapper_method in wrapper_methods_config:
+        if "wrapper_package" in wrapper_method and "wrapper_method" in wrapper_method:
             wrapper_method["wrapper"] = get_wrapper_method(
                 wrapper_method["wrapper_package"], wrapper_method["wrapper_method"])
-            if "span_name_getter_method" in wrapper_method :
+            if "span_name_getter_method" in wrapper_method:
                 wrapper_method["span_name_getter"] = get_wrapper_method(
                     wrapper_method["span_name_getter_package"],
                     wrapper_method["span_name_getter_method"])
-        return wrapper_methods
+        if "output_processor" in wrapper_method and wrapper_method["output_processor"]:
+            load_output_processor(wrapper_method, attributes_config_base_path)
 
 def get_wrapper_method(package_name: str, method_name: str):
     wrapper_module = import_module("monocle_apptrace." + package_name)
     return getattr(wrapper_module, method_name)
 
 def update_span_with_infra_name(span: Span, span_key: str):
-    for key,val  in azure_service_map.items():
+    for key, val in azure_service_map.items():
         if key in os.environ:
             span.set_attribute(span_key, val)
-    for key,val  in aws_service_map.items():
+    for key, val in aws_service_map.items():
         if key in os.environ:
             span.set_attribute(span_key, val)
 
-
 def set_embedding_model(model_name: str):
     """
     Sets the embedding model in the global context.
@@ -83,7 +134,6 @@ def set_embedding_model(model_name: str):
     """
     embedding_model_context['embedding_model'] = model_name
 
-
 def get_embedding_model() -> str:
     """
     Retrieves the embedding model from the global context.
@@ -91,3 +141,32 @@ def get_embedding_model() -> str:
     @return: The name of the embedding model, or 'unknown' if not set
     """
     return embedding_model_context.get('embedding_model', 'unknown')
+
+def set_attribute(key: str, value: str):
+    """
+    Set a value in the global context for a given key.
+
+    Args:
+        key: The key for the context value to set.
+        value: The value to set for the given key.
+    """
+    attach(set_value(key, value))
+
+def get_attribute(key: str) -> str:
+    """
+    Retrieve a value from the global context for a given key.
+
+    Args:
+        key: The key for the context value to retrieve.
+
+    Returns:
+        The value associated with the given key.
+    """
+    return get_value(key)
+
+def get_workflow_name(span: Span) -> str:
+    try:
+        return get_value("workflow_name") or span.resource.attributes.get("service.name")
+    except Exception as e:
+        logger.exception(f"Error getting workflow name: {e}")
+        return None

monocle_apptrace/wrap_common.py

@@ -1,27 +1,28 @@
-#pylint: disable=protected-access
+# pylint: disable=protected-access
 import logging
 import os
+import inspect
 from urllib.parse import urlparse
-
 from opentelemetry.trace import Span, Tracer
-from monocle_apptrace.utils import resolve_from_alias, update_span_with_infra_name, with_tracer_wrapper, get_embedding_model
-
+from monocle_apptrace.utils import resolve_from_alias, update_span_with_infra_name, with_tracer_wrapper, get_embedding_model, get_attribute, get_workflow_name
+from monocle_apptrace.utils import set_attribute
+from opentelemetry.context import get_value, attach, set_value
 logger = logging.getLogger(__name__)
 WORKFLOW_TYPE_KEY = "workflow_type"
-CONTEXT_INPUT_KEY = "context_input"
-CONTEXT_OUTPUT_KEY = "context_output"
-PROMPT_INPUT_KEY = "input"
-PROMPT_OUTPUT_KEY = "output"
+DATA_INPUT_KEY = "data.input"
+DATA_OUTPUT_KEY = "data.output"
+PROMPT_INPUT_KEY = "data.input"
+PROMPT_OUTPUT_KEY = "data.output"
 QUERY = "question"
 RESPONSE = "response"
-TAGS = "tags"
 SESSION_PROPERTIES_KEY = "session"
 INFRA_SERVICE_KEY = "infra_service_name"
+
 TYPE = "type"
 PROVIDER = "provider_name"
 EMBEDDING_MODEL = "embedding_model"
 VECTOR_STORE = 'vector_store'
-
+META_DATA = 'metadata'
 
 WORKFLOW_TYPE_MAP = {
     "llama_index": "workflow.llamaindex",
@@ -29,24 +30,53 @@ WORKFLOW_TYPE_MAP = {
     "haystack": "workflow.haystack"
 }
 
+def get_embedding_model_for_vectorstore(instance):
+    # Handle Langchain or other frameworks where vectorstore exists
+    if hasattr(instance, 'vectorstore'):
+        vectorstore_dict = instance.vectorstore.__dict__
+
+        # Use inspect to check if the embedding function is from Sagemaker
+        if 'embedding_func' in vectorstore_dict:
+            embedding_func = vectorstore_dict['embedding_func']
+            class_name = embedding_func.__class__.__name__
+            file_location = inspect.getfile(embedding_func.__class__)
+
+            # Check if the class is SagemakerEndpointEmbeddings
+            if class_name == 'SagemakerEndpointEmbeddings' and 'langchain_community' in file_location:
+                # Set embedding_model as endpoint_name if it's Sagemaker
+                if hasattr(embedding_func, 'endpoint_name'):
+                    return embedding_func.endpoint_name
+
+        # Default to the regular embedding model if not Sagemaker
+        return instance.vectorstore.embeddings.model
+
+    # Handle llama_index where _embed_model is present
+    if hasattr(instance, '_embed_model') and hasattr(instance._embed_model, 'model_name'):
+        return instance._embed_model.model_name
+
+    # Fallback if no specific model is found
+    return "Unknown Embedding Model"
+
+
 framework_vector_store_mapping = {
     'langchain_core.retrievers': lambda instance: {
-        'provider': instance.tags[0],
-        'embedding_model': instance.tags[1],
+        'provider': type(instance.vectorstore).__name__,
+        'embedding_model': get_embedding_model_for_vectorstore(instance),
         'type': VECTOR_STORE,
     },
     'llama_index.core.indices.base_retriever': lambda instance: {
         'provider': type(instance._vector_store).__name__,
-        'embedding_model': instance._embed_model.model_name,
+        'embedding_model': get_embedding_model_for_vectorstore(instance),
         'type': VECTOR_STORE,
     },
-    'haystack.components.retrievers': lambda instance: {
+    'haystack.components.retrievers.in_memory': lambda instance: {
         'provider': instance.__dict__.get("document_store").__class__.__name__,
         'embedding_model': get_embedding_model(),
         'type': VECTOR_STORE,
     },
 }
 
+
 @with_tracer_wrapper
 def task_wrapper(tracer: Tracer, to_wrap, wrapped, instance, args, kwargs):
     """Instruments and calls every function defined in TO_WRAP."""
@@ -63,36 +93,83 @@ def task_wrapper(tracer: Tracer, to_wrap, wrapped, instance, args, kwargs):
         name = f"langchain.task.{instance.__class__.__name__}"
 
     with tracer.start_as_current_span(name) as span:
+        process_span(to_wrap, span, instance, args)
         pre_task_processing(to_wrap, instance, args, span)
         return_value = wrapped(*args, **kwargs)
         post_task_processing(to_wrap, span, return_value)
 
     return return_value
 
-def post_task_processing(to_wrap, span, return_value):
-    update_span_with_context_output(to_wrap=to_wrap, return_value=return_value, span=span)
 
+def process_span(to_wrap, span, instance, args):
+    # Check if the output_processor is a valid JSON (in Python, that means it's a dictionary)
+    span_index = 1
     if is_root_span(span):
-        workflow_name = span.resource.attributes.get("service.name")
-        span.set_attribute("workflow_name",workflow_name)
-        update_span_with_prompt_output(to_wrap=to_wrap, wrapped_args=return_value, span=span)
-        update_workflow_type(to_wrap, span)
+        workflow_name = get_workflow_name(span)
+        if workflow_name:
+            span.set_attribute(f"entity.{span_index}.name", workflow_name)
+        # workflow type
+        package_name = to_wrap.get('package')
+        for (package, workflow_type) in WORKFLOW_TYPE_MAP.items():
+            if (package_name is not None and package in package_name):
+                span.set_attribute(f"entity.{span_index}.type", workflow_type)
+        span_index += 1
+    if 'output_processor' in to_wrap:
+        output_processor=to_wrap['output_processor']
+        if isinstance(output_processor, dict) and len(output_processor) > 0:
+            if 'type' in output_processor:
+                span.set_attribute("span.type", output_processor['type'])
+            else:
+                logger.warning("type of span not found or incorrect written in entity json")
+            count = 0
+            if 'attributes' in output_processor:
+                count = len(output_processor["attributes"])
+                span.set_attribute("entity.count", count)
+                span_index = 1
+                for processors in output_processor["attributes"]:
+                    for processor in processors:
+                        attribute = processor.get('attribute')
+                        accessor = processor.get('accessor')
+
+                        if attribute and accessor:
+                            attribute_name = f"entity.{span_index}.{attribute}"
+                            try:
+                                result = eval(accessor)(instance, args)
+                                if result and isinstance(result, str):
+                                    span.set_attribute(attribute_name, result)
+                            except Exception as e:
+                                logger.error(f"Error processing accessor: {e}")
+                        else:
+                            logger.warning(f"{' and '.join([key for key in ['attribute', 'accessor'] if not processor.get(key)])} not found or incorrect in entity JSON")
+                    span_index += 1
+            else:
+                logger.warning("attributes not found or incorrect written in entity json")
+                span.set_attribute("span.count", count)
 
-def pre_task_processing(to_wrap, instance, args, span):
-    if is_root_span(span):
-        update_span_with_prompt_input(to_wrap=to_wrap, wrapped_args=args, span=span)
+        else:
+            logger.warning("empty or entities json is not in correct format")
 
-        update_span_with_infra_name(span, INFRA_SERVICE_KEY)
 
-    #capture the tags attribute of the instance if present, else ignore
+def post_task_processing(to_wrap, span, return_value):
     try:
-        update_tags(instance, span)
-        update_vectorstore_attributes(to_wrap, instance, span)
-    except AttributeError:
-        pass
-    update_span_with_context_input(to_wrap=to_wrap, wrapped_args=args, span=span)
+        update_span_with_context_output(to_wrap=to_wrap, return_value=return_value, span=span)
+
+        if is_root_span(span):
+            update_span_with_prompt_output(to_wrap=to_wrap, wrapped_args=return_value, span=span)
+    except:
+        logger.exception("exception in post_task_processing")
 
 
+def pre_task_processing(to_wrap, instance, args, span):
+    try:
+        if is_root_span(span):
+            update_span_with_prompt_input(to_wrap=to_wrap, wrapped_args=args, span=span)
+            update_span_with_infra_name(span, INFRA_SERVICE_KEY)
+
+        update_span_with_context_input(to_wrap=to_wrap, wrapped_args=args, span=span)
+    except:
+        logger.exception("exception in pre_task_processing")
+
 
 @with_tracer_wrapper
 async def atask_wrapper(tracer, to_wrap, wrapped, instance, args, kwargs):
@@ -109,12 +186,14 @@ async def atask_wrapper(tracer, to_wrap, wrapped, instance, args, kwargs):
     else:
         name = f"langchain.task.{instance.__class__.__name__}"
     with tracer.start_as_current_span(name) as span:
+        process_span(to_wrap, span, instance, args)
         pre_task_processing(to_wrap, instance, args, span)
         return_value = await wrapped(*args, **kwargs)
         post_task_processing(to_wrap, span, return_value)
 
     return return_value
 
+
 @with_tracer_wrapper
 async def allm_wrapper(tracer, to_wrap, wrapped, instance, args, kwargs):
     # Some Langchain objects are wrapped elsewhere, so we ignore them here
@@ -131,16 +210,24 @@ async def allm_wrapper(tracer, to_wrap, wrapped, instance, args, kwargs):
     else:
         name = f"langchain.task.{instance.__class__.__name__}"
     with tracer.start_as_current_span(name) as span:
-        update_llm_endpoint(curr_span= span, instance=instance)
+        if 'haystack.components.retrievers' in to_wrap['package'] and 'haystack.retriever' in span.name:
+            input_arg_text = get_attribute(DATA_INPUT_KEY)
+            span.add_event(DATA_INPUT_KEY, {QUERY: input_arg_text})
+        provider_name, inference_endpoint = get_provider_name(instance)
+        instance_args = {"provider_name": provider_name, "inference_endpoint": inference_endpoint}
+
+        process_span(to_wrap, span, instance, instance_args)
 
         return_value = await wrapped(*args, **kwargs)
-        update_span_from_llm_response(response = return_value, span = span)
+        if 'haystack.components.retrievers' in to_wrap['package'] and 'haystack.retriever' in span.name:
+            update_span_with_context_output(to_wrap=to_wrap, return_value=return_value, span=span)
+        update_span_from_llm_response(response=return_value, span=span, instance=instance)
 
     return return_value
 
+
 @with_tracer_wrapper
 def llm_wrapper(tracer: Tracer, to_wrap, wrapped, instance, args, kwargs):
-
     # Some Langchain objects are wrapped elsewhere, so we ignore them here
     if instance.__class__.__name__ in ("AgentExecutor"):
         return wrapped(*args, **kwargs)
@@ -154,87 +241,121 @@ def llm_wrapper(tracer: Tracer, to_wrap, wrapped, instance, args, kwargs):
         name = to_wrap.get("span_name")
     else:
         name = f"langchain.task.{instance.__class__.__name__}"
+
     with tracer.start_as_current_span(name) as span:
         if 'haystack.components.retrievers' in to_wrap['package'] and 'haystack.retriever' in span.name:
-            update_vectorstore_attributes(to_wrap, instance, span)
-        update_llm_endpoint(curr_span= span, instance=instance)
+            input_arg_text = get_attribute(DATA_INPUT_KEY)
+            span.add_event(DATA_INPUT_KEY, {QUERY: input_arg_text})
+        provider_name, inference_endpoint = get_provider_name(instance)
+        instance_args = {"provider_name": provider_name, "inference_endpoint": inference_endpoint}
+
+        process_span(to_wrap, span, instance, instance_args)
 
         return_value = wrapped(*args, **kwargs)
-        update_span_from_llm_response(response = return_value, span = span)
+        if 'haystack.components.retrievers' in to_wrap['package'] and 'haystack.retriever' in span.name:
+            update_span_with_context_output(to_wrap=to_wrap, return_value=return_value, span=span)
+        update_span_from_llm_response(response=return_value, span=span, instance=instance)
 
     return return_value
 
+
 def update_llm_endpoint(curr_span: Span, instance):
+    # Lambda to set attributes if values are not None
+    __set_span_attribute_if_not_none = lambda span, **kwargs: [
+        span.set_attribute(k, v) for k, v in kwargs.items() if v is not None
+    ]
+
     triton_llm_endpoint = os.environ.get("TRITON_LLM_ENDPOINT")
     if triton_llm_endpoint is not None and len(triton_llm_endpoint) > 0:
         curr_span.set_attribute("server_url", triton_llm_endpoint)
     else:
-        if 'temperature' in instance.__dict__:
-            temp_val = instance.__dict__.get("temperature")
-            curr_span.set_attribute("temperature", temp_val)
-            # handling for model name
-        model_name = resolve_from_alias(instance.__dict__ , ["model","model_name"])
-        curr_span.set_attribute("model_name", model_name)
-        set_provider_name(curr_span, instance)
-        # handling AzureOpenAI deployment
-        deployment_name = resolve_from_alias(instance.__dict__ , [ "engine", "azure_deployment",
-                                                                   "deployment_name", "deployment_id", "deployment"])
-        curr_span.set_attribute("az_openai_deployment", deployment_name)
-        # handling the inference endpoint
-        inference_ep = resolve_from_alias(instance.__dict__,["azure_endpoint","api_base"])
-        curr_span.set_attribute("inference_endpoint",inference_ep)
-
-def set_provider_name(curr_span, instance):
+        # Get temperature if present
+        temp_val = instance.__dict__.get("temperature")
+
+        # Resolve values for model name, deployment, and inference endpoint
+        model_name = resolve_from_alias(instance.__dict__, ["model", "model_name"])
+        deployment_name = resolve_from_alias(instance.__dict__,
+                                             ["engine", "azure_deployment", "deployment_name", "deployment_id",
+                                              "deployment"])
+        inference_ep = resolve_from_alias(instance.__dict__, ["azure_endpoint", "api_base"])
+
+        # Use the lambda to set attributes conditionally
+        __set_span_attribute_if_not_none(
+            curr_span,
+            temperature=temp_val,
+            model_name=model_name,
+            az_openai_deployment=deployment_name,
+            inference_endpoint=inference_ep
+        )
+
+
+def get_provider_name(instance):
     provider_url = ""
-
-    try :
-        if isinstance(instance.client._client.base_url.host, str) :
-            provider_url = instance. client._client.base_url.host
+    inference_endpoint = ""
+    try:
+        if isinstance(instance.client._client.base_url.host, str):
+            provider_url = instance.client._client.base_url.host
+        if isinstance(instance.client._client.base_url, str):
+            inference_endpoint = instance.client._client.base_url
+        else:
+            inference_endpoint = str(instance.client._client.base_url)
     except:
         pass
 
-    try :
+    try:
         if isinstance(instance.api_base, str):
             provider_url = instance.api_base
     except:
         pass
 
-    try :
+    try:
         if len(provider_url) > 0:
             parsed_provider_url = urlparse(provider_url)
-            curr_span.set_attribute("provider_name", parsed_provider_url.hostname or provider_url)
     except:
         pass
+    return parsed_provider_url.hostname or provider_url,inference_endpoint
+
 
 def is_root_span(curr_span: Span) -> bool:
     return curr_span.parent is None
 
+
 def get_input_from_args(chain_args):
     if len(chain_args) > 0 and isinstance(chain_args[0], str):
         return chain_args[0]
     return ""
 
-def update_span_from_llm_response(response, span: Span):
+
+def update_span_from_llm_response(response, span: Span, instance):
     # extract token uasge from langchain openai
     if (response is not None and hasattr(response, "response_metadata")):
         response_metadata = response.response_metadata
         token_usage = response_metadata.get("token_usage")
+        meta_dict = {}
         if token_usage is not None:
-            span.set_attribute("completion_tokens", token_usage.get("completion_tokens"))
-            span.set_attribute("prompt_tokens", token_usage.get("prompt_tokens"))
-            span.set_attribute("total_tokens", token_usage.get("total_tokens"))
+            temperature = instance.__dict__.get("temperature", None)
+            meta_dict.update({"temperature": temperature})
+            meta_dict.update({"completion_tokens": token_usage.get("completion_tokens")})
+            meta_dict.update({"prompt_tokens": token_usage.get("prompt_tokens")})
+            meta_dict.update({"total_tokens": token_usage.get("total_tokens")})
+            span.add_event(META_DATA, meta_dict)
     # extract token usage from llamaindex openai
-    if(response is not None and hasattr(response, "raw")):
+    if (response is not None and hasattr(response, "raw")):
         try:
+            meta_dict = {}
             if response.raw is not None:
-                token_usage = response.raw.get("usage") if isinstance(response.raw, dict) else getattr(response.raw, "usage", None)
+                token_usage = response.raw.get("usage") if isinstance(response.raw, dict) else getattr(response.raw,
+                                                                                                       "usage", None)
                 if token_usage is not None:
+                    temperature = instance.__dict__.get("temperature", None)
+                    meta_dict.update({"temperature": temperature})
                     if getattr(token_usage, "completion_tokens", None):
-                        span.set_attribute("completion_tokens", getattr(token_usage, "completion_tokens"))
+                        meta_dict.update({"completion_tokens": getattr(token_usage, "completion_tokens")})
                     if getattr(token_usage, "prompt_tokens", None):
-                        span.set_attribute("prompt_tokens", getattr(token_usage, "prompt_tokens"))
+                        meta_dict.update({"prompt_tokens": getattr(token_usage, "prompt_tokens")})
                     if getattr(token_usage, "total_tokens", None):
-                        span.set_attribute("total_tokens", getattr(token_usage, "total_tokens"))
+                        meta_dict.update({"total_tokens": getattr(token_usage, "total_tokens")})
+                    span.add_event(META_DATA, meta_dict)
         except AttributeError:
             token_usage = None
 
@@ -243,69 +364,54 @@ def update_workflow_type(to_wrap, span: Span):
     package_name = to_wrap.get('package')
 
     for (package, workflow_type) in WORKFLOW_TYPE_MAP.items():
-        if(package_name is not None and package in package_name):
+        if (package_name is not None and package in package_name):
             span.set_attribute(WORKFLOW_TYPE_KEY, workflow_type)
 
-def update_span_with_context_input(to_wrap, wrapped_args ,span: Span):
-    package_name: str = to_wrap.get('package')
-    if "langchain_core.retrievers" in package_name:
-        input_arg_text = wrapped_args[0]
-        span.add_event(CONTEXT_INPUT_KEY, {QUERY:input_arg_text})
-    if "llama_index.core.indices.base_retriever" in package_name:
-        input_arg_text = wrapped_args[0].query_str
-        span.add_event(CONTEXT_INPUT_KEY, {QUERY:input_arg_text})
 
-def update_span_with_context_output(to_wrap, return_value ,span: Span):
+def update_span_with_context_input(to_wrap, wrapped_args, span: Span):
     package_name: str = to_wrap.get('package')
-    if "llama_index.core.indices.base_retriever" in package_name:
-        output_arg_text = return_value[0].text
-        span.add_event(CONTEXT_OUTPUT_KEY, {RESPONSE:output_arg_text})
-
-def update_span_with_prompt_input(to_wrap, wrapped_args ,span: Span):
+    input_arg_text = ""
+    if "langchain_core.retrievers" in package_name and len(wrapped_args) > 0:
+        input_arg_text += wrapped_args[0]
+    if "llama_index.core.indices.base_retriever" in package_name and len(wrapped_args) > 0:
+        input_arg_text += wrapped_args[0].query_str
+    if "haystack.components.retrievers.in_memory" in package_name:
+        input_arg_text += get_attribute(DATA_INPUT_KEY)
+    if input_arg_text:
+        span.add_event(DATA_INPUT_KEY, {QUERY: input_arg_text})
+
+
+def update_span_with_context_output(to_wrap, return_value, span: Span):
+    package_name: str = to_wrap.get('package')
+    output_arg_text = ""
+    if "langchain_core.retrievers" in package_name:
+        output_arg_text += " ".join([doc.page_content for doc in return_value if hasattr(doc, 'page_content')])
+        if len(output_arg_text) > 100:
+            output_arg_text = output_arg_text[:100] + "..."
+    if "llama_index.core.indices.base_retriever" in package_name and len(return_value) > 0:
+        output_arg_text += return_value[0].text
+    if "haystack.components.retrievers.in_memory" in package_name:
+        output_arg_text += " ".join([doc.content for doc in return_value['documents']])
+        if len(output_arg_text) > 100:
+            output_arg_text = output_arg_text[:100] + "..."
+    if output_arg_text:
+        span.add_event(DATA_OUTPUT_KEY, {RESPONSE: output_arg_text})
+
+
+def update_span_with_prompt_input(to_wrap, wrapped_args, span: Span):
     input_arg_text = wrapped_args[0]
 
     if isinstance(input_arg_text, dict):
-        span.add_event(PROMPT_INPUT_KEY,input_arg_text)
+        span.add_event(PROMPT_INPUT_KEY, input_arg_text)
     else:
-        span.add_event(PROMPT_INPUT_KEY,{QUERY:input_arg_text})
+        span.add_event(PROMPT_INPUT_KEY, {QUERY: input_arg_text})
 
-def update_span_with_prompt_output(to_wrap, wrapped_args ,span: Span):
+
+def update_span_with_prompt_output(to_wrap, wrapped_args, span: Span):
     package_name: str = to_wrap.get('package')
     if isinstance(wrapped_args, str):
-        span.add_event(PROMPT_OUTPUT_KEY, {RESPONSE:wrapped_args})
+        span.add_event(PROMPT_OUTPUT_KEY, {RESPONSE: wrapped_args})
+    if isinstance(wrapped_args, dict):
+        span.add_event(PROMPT_OUTPUT_KEY, wrapped_args)
     if "llama_index.core.base.base_query_engine" in package_name:
-        span.add_event(PROMPT_OUTPUT_KEY, {RESPONSE:wrapped_args.response})
-
-def update_tags(instance, span):
-    try:
-        # copy tags as is from langchain
-        span.set_attribute(TAGS, getattr(instance, TAGS))
-    except:
-        pass
-    try:
-        # extract embed model and vector store names for llamaindex
-        model_name = instance.retriever._embed_model.model_name
-        vector_store_name = type(instance.retriever._vector_store).__name__
-        span.set_attribute(TAGS, [model_name, vector_store_name])
-    except:
-        pass
-
-
-def update_vectorstore_attributes(to_wrap, instance, span):
-    """
-       Updates the telemetry span attributes for vector store retrieval tasks.
-    """
-    try:
-        package = to_wrap.get('package')
-        if package in framework_vector_store_mapping:
-            attributes = framework_vector_store_mapping[package](instance)
-            span._attributes.update({
-                TYPE: attributes['type'],
-                PROVIDER: attributes['provider'],
-                EMBEDDING_MODEL: attributes['embedding_model']
-            })
-        else:
-            logger.warning(f"Package '{package}' not recognized for vector store telemetry.")
-
-    except Exception as e:
-        logger.error(f"Error updating span attributes: {e}")
+        span.add_event(PROMPT_OUTPUT_KEY, {RESPONSE: wrapped_args.response})