sunholo 0.104.5__tar.gz → 0.105.0__tar.gz

{sunholo-0.104.5 → sunholo-0.105.0}/PKG-INFO

@@ -1,9 +1,9 @@
 Metadata-Version: 2.1
 Name: sunholo
-Version: 0.104.5
+Version: 0.105.0
 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/v0.104.5.tar.gz
+Download-URL: https://github.com/sunholo-data/sunholo-py/archive/refs/tags/v0.105.0.tar.gz
 Author: Holosun ApS
 Author-email: multivac@sunholo.com
 License: Apache License, Version 2.0

{sunholo-0.104.5 → sunholo-0.105.0}/setup.py

@@ -1,6 +1,6 @@
 from setuptools import setup, find_packages
 
-version = '0.104.5'
+version = '0.105.0'
 
 setup(
     name='sunholo',

{sunholo-0.104.5 → sunholo-0.105.0}/sunholo/agents/flask/vac_routes.py

@@ -618,7 +618,7 @@ if __name__ == "__main__":
             user_id = user_id,
             session_id = session_id,
             tags = tags,
-            release = f"sunholo-v{package_version}"
+            release = package_version
         )
 
     def prep_vac(self, request, vector_name):
@@ -660,7 +660,12 @@ if __name__ == "__main__":
             raise ValueError(f"Unable to find vac_config for {vector_name} - {str(e)}")
 
         if trace:
-            trace.update(input=data, metadata=vac_config.configs_by_kind)
+            this_vac_config = vac_config.configs_by_kind.get("vacConfig")
+            metadata_config=None
+            if this_vac_config:
+                metadata_config = this_vac_config.get(vector_name)
+
+            trace.update(input=data, metadata=metadata_config)
 
         user_input = data.pop('user_input').strip()
         stream_wait_time = data.pop('stream_wait_time', 7)

{sunholo-0.104.5 → sunholo-0.105.0}/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
@@ -441,11 +441,11 @@ class GenAIFunctionProcessor:
 
             log.info(f"== Start input content for loop [{guardrail}]\n ## Content: {content_parse}")
             this_text = ""  # reset for this loop
-            response = []
+            response = None
 
             try:
                 token_queue.append("\n= Calling Agent =\n")
-                response = chat.send_message(content, stream=True, request_options=RequestOptions(
+                response:GenerateContentResponse = chat.send_message(content, stream=True, request_options=RequestOptions(
                                         retry=retry.Retry(
                                             initial=10, 
                                             multiplier=2, 
@@ -465,19 +465,20 @@ 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
 
             for chunk in response:
                 if not chunk:

{sunholo-0.104.5 → sunholo-0.105.0}/sunholo/streaming/streaming.py

@@ -87,25 +87,31 @@ def start_streaming_chat(question,
             break
     else:
         log.info(f"Stream has ended after {round(time.time() - first_start, 2)} seconds")
-        log.info("Sending final full message plus sources...")
+        
         
     
     # if  you need it to stop it elsewhere use 
     # stop_event.set()
     content_to_send = content_buffer.read()
     if content_to_send:
-        log.info(f"==\n{content_to_send}")
+        log.info(f"==Flush==\n{content_to_send}")
         yield content_to_send
         content_buffer.clear()
 
     # Stop the stream thread
     chat_thread.join()
 
-    # the json object with full response in 'answer' and the 'sources' array
-    final_result = result_queue.get()
+    if kwargs.get("stream_only"):
+        log.info("stream_only so finishing now")
+        final_yield = ""
+    else:
+        log.info("Sending final full message plus sources...")
+        # the json object with full response in 'answer' and the 'sources' array
+        final_result = result_queue.get()
+        final_yield = parse_output(final_result)
 
     # parses out source_documents if not present etc.
-    yield parse_output(final_result)
+    yield final_yield
 
 async def start_streaming_chat_async(question, vector_name, qna_func_async, chat_history=[], wait_time=2, timeout=120, **kwargs): 
     """

{sunholo-0.104.5 → sunholo-0.105.0}/sunholo.egg-info/PKG-INFO

@@ -1,9 +1,9 @@
 Metadata-Version: 2.1
 Name: sunholo
-Version: 0.104.5
+Version: 0.105.0
 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/v0.104.5.tar.gz
+Download-URL: https://github.com/sunholo-data/sunholo-py/archive/refs/tags/v0.105.0.tar.gz
 Author: Holosun ApS
 Author-email: multivac@sunholo.com
 License: Apache License, Version 2.0