sunholo 0.104.6__py3-none-any.whl → 0.105.1__py3-none-any.whl

sunholo/genai/process_funcs_cls.py

@@ -12,7 +12,7 @@ from collections import deque
 try:
     import google.generativeai as genai
     import proto
-    from google.generativeai.types import RequestOptions
+    from google.generativeai.types import RequestOptions, GenerateContentResponse
     from google.api_core import retry
     from google.generativeai import ChatSession
     from google.api_core.exceptions import RetryError
@@ -60,7 +60,7 @@ class GenAIFunctionProcessor:
     ```
     """
 
-    def __init__(self, config: ConfigManager=None, model_name=None):
+    def __init__(self, config: ConfigManager=None, model_name=None, trace=None, parent_observation_id=None):
         """
         Initializes the GenAIFunctionProcessor with the given configuration.
 
@@ -84,6 +84,9 @@ class GenAIFunctionProcessor:
         elif model_name:
             self.model_name = model_name
         
+        self.trace = trace
+        self.parent_observation_id = parent_observation_id
+        
         self.last_api_requests_and_responses = []
         self._validate_functions()
 
@@ -430,6 +433,12 @@ class GenAIFunctionProcessor:
         # Initialize token queue to ensure sequential processing
         token_queue = deque()
 
+        span = self.trace.span(
+            name=f"GenAIFunctionProcesser_{self.__class__.__name__}",
+            parent_observation_id=self.parent_observation_id,
+            input = {'content': content},
+        ) if self.trace else None
+
         while guardrail < guardrail_max:
 
             token_queue.append(f"\n----Loop [{guardrail}] Start------\nFunctions: {list(self.funcs.keys())}\n")
@@ -441,11 +450,23 @@ class GenAIFunctionProcessor:
 
             log.info(f"== Start input content for loop [{guardrail}]\n ## Content: {content_parse}")
             this_text = ""  # reset for this loop
-            response = []
+            response = None
+            loop_span = span.span(
+                name=f"loop_{guardrail}",
+                model=self.model_name,
+                input = {'content': content},
+            ) if span else None
 
             try:
                 token_queue.append("\n= Calling Agent =\n")
-                response = chat.send_message(content, stream=True, request_options=RequestOptions(
+
+                gen = loop_span.generation(
+                    name=f"loop_{guardrail}",
+                    model=self.model_name,
+                    input = {'content': content},
+                ) if loop_span else None
+
+                response: GenerateContentResponse = chat.send_message(content, stream=True, request_options=RequestOptions(
                                         retry=retry.Retry(
                                             initial=10, 
                                             multiplier=2, 
@@ -465,19 +486,23 @@ class GenAIFunctionProcessor:
                 token_queue.append(msg)
                 break
 
-            loop_metadata = response.usage_metadata
-            if loop_metadata:
-                usage_metadata = {
-                    "prompt_token_count": usage_metadata["prompt_token_count"] + (loop_metadata.prompt_token_count or 0),
-                    "candidates_token_count": usage_metadata["candidates_token_count"] + (loop_metadata.candidates_token_count or 0),
-                    "total_token_count": usage_metadata["total_token_count"] + (loop_metadata.total_token_count or 0),
-                }
-                token_queue.append((
-                    "\n-- Agent response -- " 
-                    f"Loop tokens: [{loop_metadata.prompt_token_count}]/[{usage_metadata['prompt_token_count']}] "
-                    f"Session tokens: [{loop_metadata.total_token_count}]/[{usage_metadata['total_token_count']}] \n"
-                ))
-            loop_metadata = None
+            if response:
+                loop_metadata = response.usage_metadata
+                if loop_metadata:
+                    usage_metadata = {
+                        "prompt_token_count": usage_metadata["prompt_token_count"] + (loop_metadata.prompt_token_count or 0),
+                        "candidates_token_count": usage_metadata["candidates_token_count"] + (loop_metadata.candidates_token_count or 0),
+                        "total_token_count": usage_metadata["total_token_count"] + (loop_metadata.total_token_count or 0),
+                    }
+                    token_queue.append((
+                        "\n-- Agent response -- " 
+                        f"Loop tokens: [{loop_metadata.prompt_token_count}]/[{usage_metadata['prompt_token_count']}] "
+                        f"Session tokens: [{loop_metadata.total_token_count}]/[{usage_metadata['total_token_count']}] \n"
+                    ))
+                loop_metadata = None
+                gen.end(output=dict(response)) if gen else None
+            else:
+                gen.end(output="No response received") if gen else None
 
             for chunk in response:
                 if not chunk:
@@ -494,7 +519,7 @@ class GenAIFunctionProcessor:
                     
                 except ValueError as err:
                     token_queue.append(f"{str(err)} for {chunk=}")
-            
+            fn_span = loop_span.span(name="function_execution", input=response) if loop_span else None
             try:
                 executed_responses = self.process_funcs(response) 
             except Exception as err:
@@ -503,9 +528,11 @@ class GenAIFunctionProcessor:
                 token_queue.append(f"{str(err)} for {response=}")
 
             log.info(f"[{guardrail}] {executed_responses=}")
+            fn_span.end(output=executed_responses) if fn_span else None
 
             if executed_responses:  
                 token_queue.append("\n-- Agent Actions:\n")
+                fn_exec = loop_span.span(name="function_actions", input=executed_responses) if loop_span else None
                 for executed_response in executed_responses:
                     token = ""
                     fn = executed_response.function_response.name
@@ -521,6 +548,7 @@ class GenAIFunctionProcessor:
                         callback.on_llm_new_token(token=token)
 
                     log.info(f"{fn_log} created a result={type(fn_result)=}")
+                    fn_exec_one = fn_exec.span(name=fn_log, input=fn_result) if fn_exec else None
                     
                     fn_result_json = None
                     # Convert MapComposite to a standard Python dictionary
@@ -573,11 +601,15 @@ class GenAIFunctionProcessor:
                     
                     this_text += token
                     token_queue.append(token)
+                    fn_exec_one.end(output=token) if fn_exec_one else None
+                    
             else:
                 token = "\nNo function executions were found\n"
                 token_queue.append(token)
                 this_text += token
 
+            fn_exec.end(output=this_text) if fn_exec else None
+
             if this_text:
                 content.append(f"Agent: {this_text}")    
                 # if text includes gs:// try to download it
@@ -594,6 +626,7 @@ class GenAIFunctionProcessor:
                 content.append(f"Agent: No response was found for loop [{guardrail}]")
 
             token_queue.append(f"\n----Loop [{guardrail}] End------\n{usage_metadata}\n----------------------")
+            loop_span.end(output=content, metadata=loop_metadata) if loop_span else None
 
             go_on_check = self.check_function_result("decide_to_go_on", {"go_on": False})
             if go_on_check:
@@ -616,6 +649,7 @@ class GenAIFunctionProcessor:
         usage_metadata["functions_called"] = functions_called
 
         big_text = "\n".join(big_result[-loop_return:])
+        span.end(output=big_text, metadata=usage_metadata) if span else None
 
         return big_text, usage_metadata
 

{sunholo-0.104.6.dist-info → sunholo-0.105.1.dist-info}/METADATA

@@ -1,9 +1,9 @@
 Metadata-Version: 2.1
 Name: sunholo
-Version: 0.104.6
+Version: 0.105.1
 Summary: Large Language Model DevOps - a package to help deploy LLMs to the Cloud.
 Home-page: https://github.com/sunholo-data/sunholo-py
-Download-URL: https://github.com/sunholo-data/sunholo-py/archive/refs/tags/v0.104.6.tar.gz
+Download-URL: https://github.com/sunholo-data/sunholo-py/archive/refs/tags/v0.105.1.tar.gz
 Author: Holosun ApS
 Author-email: multivac@sunholo.com
 License: Apache License, Version 2.0

{sunholo-0.104.6.dist-info → sunholo-0.105.1.dist-info}/RECORD

@@ -88,7 +88,7 @@ sunholo/gcs/metadata.py,sha256=oQLcXi4brsZ74aegWyC1JZmhlaEV270HS5_UWtAYYWE,898
 sunholo/genai/__init__.py,sha256=dBl6IA3-Fx6-Vx81r0XqxHlUq6WeW1iDX188dpChu8s,115
 sunholo/genai/images.py,sha256=EyjsDqt6XQw99pZUQamomCpMOoIah9bp3XY94WPU7Ms,1678
 sunholo/genai/init.py,sha256=yG8E67TduFCTQPELo83OJuWfjwTnGZsyACospahyEaY,687
-sunholo/genai/process_funcs_cls.py,sha256=6NZiYd9VWEzRN_-25PvOpBnQcfuJfWMuqi2hgAYm9ls,29530
+sunholo/genai/process_funcs_cls.py,sha256=c3pNWj8jWahMEwIbejm9euBr31WW0z50H5CPr9FW7cY,31310
 sunholo/genai/safety.py,sha256=mkFDO_BeEgiKjQd9o2I4UxB6XI7a9U-oOFjZ8LGRUC4,1238
 sunholo/invoke/__init__.py,sha256=o1RhwBGOtVK0MIdD55fAIMCkJsxTksi8GD5uoqVKI-8,184
 sunholo/invoke/async_class.py,sha256=G8vD2H94fpBc37mSJSQODEKJ67P2mPQEHabtDaLOvxE,8033
@@ -147,9 +147,9 @@ sunholo/vertex/init.py,sha256=1OQwcPBKZYBTDPdyU7IM4X4OmiXLdsNV30C-fee2scQ,2875
 sunholo/vertex/memory_tools.py,sha256=tBZxqVZ4InTmdBvLlOYwoSEWu4-kGquc-gxDwZCC4FA,7667
 sunholo/vertex/safety.py,sha256=S9PgQT1O_BQAkcqauWncRJaydiP8Q_Jzmu9gxYfy1VA,2482
 sunholo/vertex/type_dict_to_json.py,sha256=uTzL4o9tJRao4u-gJOFcACgWGkBOtqACmb6ihvCErL8,4694
-sunholo-0.104.6.dist-info/LICENSE.txt,sha256=SdE3QjnD3GEmqqg9EX3TM9f7WmtOzqS1KJve8rhbYmU,11345
-sunholo-0.104.6.dist-info/METADATA,sha256=kOR4eTd1sEjgHhwLIy6EK1KihYG3aMGKyG3c2N029K4,8312
-sunholo-0.104.6.dist-info/WHEEL,sha256=GV9aMThwP_4oNCtvEC2ec3qUYutgWeAzklro_0m4WJQ,91
-sunholo-0.104.6.dist-info/entry_points.txt,sha256=bZuN5AIHingMPt4Ro1b_T-FnQvZ3teBes-3OyO0asl4,49
-sunholo-0.104.6.dist-info/top_level.txt,sha256=wt5tadn5--5JrZsjJz2LceoUvcrIvxjHJe-RxuudxAk,8
-sunholo-0.104.6.dist-info/RECORD,,
+sunholo-0.105.1.dist-info/LICENSE.txt,sha256=SdE3QjnD3GEmqqg9EX3TM9f7WmtOzqS1KJve8rhbYmU,11345
+sunholo-0.105.1.dist-info/METADATA,sha256=XxAhxZcKvX8_GXcUpvePkoRnEIuCCcwATuCSk_gn3BA,8312
+sunholo-0.105.1.dist-info/WHEEL,sha256=OVMc5UfuAQiSplgO0_WdW7vXVGAt9Hdd6qtN4HotdyA,91
+sunholo-0.105.1.dist-info/entry_points.txt,sha256=bZuN5AIHingMPt4Ro1b_T-FnQvZ3teBes-3OyO0asl4,49
+sunholo-0.105.1.dist-info/top_level.txt,sha256=wt5tadn5--5JrZsjJz2LceoUvcrIvxjHJe-RxuudxAk,8
+sunholo-0.105.1.dist-info/RECORD,,

{sunholo-0.104.6.dist-info → sunholo-0.105.1.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: setuptools (75.1.0)
+Generator: setuptools (75.2.0)
 Root-Is-Purelib: true
 Tag: py3-none-any