sunholo 0.55.13__tar.gz → 0.55.15__tar.gz

{sunholo-0.55.13/sunholo.egg-info → sunholo-0.55.15}/PKG-INFO

@@ -1,9 +1,9 @@
 Metadata-Version: 2.1
 Name: sunholo
-Version: 0.55.13
+Version: 0.55.15
 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/v0.55.13.tar.gz
+Download-URL: https://github.com/sunholo-data/sunholo-py/archive/refs/tags/v0.55.15.tar.gz
 Author: Holosun ApS
 Author-email: multivac@sunholo.com
 License: Apache License, Version 2.0

{sunholo-0.55.13 → sunholo-0.55.15}/setup.py

@@ -1,7 +1,7 @@
 from setuptools import setup, find_packages
 
 # Define your base version
-version = '0.55.13'
+version = '0.55.15'
 
 setup(
     name='sunholo',

{sunholo-0.55.13 → sunholo-0.55.15}/sunholo/agents/chat_history.py

@@ -25,7 +25,7 @@ def extract_chat_history(chat_history=None):
     # Output: [("Hello, AI!", "Hello, Human! How can I help you today?")]
     ```
     """
-    if chat_history is None:
+    if not chat_history:
         log.info("No chat history found")
         return []
 

{sunholo-0.55.13 → sunholo-0.55.15}/sunholo/agents/flask/qna_routes.py

@@ -52,15 +52,17 @@ def register_qna_routes(app, stream_interpreter, vac_interpreter):
             return jsonify(command_response)
 
         log.info(f'Streaming data with: {all_input}')
+        if span:
+            generation = span.generation(
+                name="start_streaming_chat",
+                metadata=vac_config,
+                input = all_input,
+                completion_start_time=datetime.datetime.now(),
+                model=vac_config.get("model") or vac_config.get("llm")
+            )
+
         def generate_response_content():
-            if span:
-                generation = span.generation(
-                    name="start_streaming_chat",
-                    metadata=vac_config,
-                    input = all_input,
-                    completion_start_time=datetime.datetime.now(),
-                    model=vac_config.get("model") or vac_config.get("llm")
-                )
+
             chunks = ""
             for chunk in start_streaming_chat(question=all_input["user_input"],
                                               vector_name=vector_name,
@@ -69,7 +71,7 @@ def register_qna_routes(app, stream_interpreter, vac_interpreter):
                                               wait_time=all_input["stream_wait_time"],
                                               timeout=all_input["stream_timeout"],
                                               #kwargs
-                                              message_author=all_input["message_author"]
+                                              **all_input["kwargs"]
                                               ):
                 if isinstance(chunk, dict) and 'answer' in chunk:
                     # When we encounter the dictionary, we yield it as a JSON string
@@ -92,14 +94,15 @@ def register_qna_routes(app, stream_interpreter, vac_interpreter):
                         
                     yield chunk
             
-            if trace:
-                generation.end(output=chunks)
-                span.end(output=chunks)
-                trace.update(output=chunks)
-
         # Here, the generator function will handle streaming the content to the client.
         response = Response(generate_response_content(), content_type='text/plain; charset=utf-8')
         response.headers['Transfer-Encoding'] = 'chunked'  
+
+        log.debug(f"streaming response: {response}")
+        if trace:
+            generation.end(output=response)
+            span.end(output=response)
+            trace.update(output=response)
         
         if langfuse:
             langfuse.flush()  
@@ -131,7 +134,7 @@ def register_qna_routes(app, stream_interpreter, vac_interpreter):
                 question=all_input["user_input"],
                 vector_name=vector_name,
                 chat_history=all_input["chat_history"],
-                message_author=all_input["message_author"]
+                **all_input["kwargs"]
             )
             if generation:
                 generation.end(output=bot_output)
@@ -198,18 +201,18 @@ def prep_vac(request, vector_name):
     if trace:
         trace.update(input=data, metadata=vac_config)
 
-    user_input = data['user_input'].strip()
-    message_author = data.get('message_author', None)
-    stream_wait_time = data.get('stream_wait_time', 7)
-    stream_timeout = data.get('stream_timeout', 120)
+    user_input = data.pop('user_input').strip()
+    stream_wait_time = data.pop('stream_wait_time', 7)
+    stream_timeout = data.pop('stream_timeout', 120)
+    chat_history = data.pop('chat_history', None)
+    paired_messages = extract_chat_history(chat_history)
 
-    paired_messages = extract_chat_history(data.get('chat_history', None))
     all_input = {'user_input': user_input, 
                  'vector_name': vector_name, 
                  'chat_history': paired_messages, 
-                 'message_author': message_author,
                  'stream_wait_time': stream_wait_time,
-                 'stream_timeout':stream_timeout}
+                 'stream_timeout':stream_timeout,
+                 'kwargs': data}
 
     if trace:
         span = trace.span(

{sunholo-0.55.13 → sunholo-0.55.15}/sunholo/streaming/content_buffer.py

@@ -130,6 +130,7 @@ class BufferStreamingStdOutCallbackHandler(StreamingStdOutCallbackHandler):
         patterns are detected.
         """
         log.debug(f"on_llm_new_token: {token}")
+        
         self.buffer += token
 
         if '```' in token:

{sunholo-0.55.13 → sunholo-0.55.15/sunholo.egg-info}/PKG-INFO

@@ -1,9 +1,9 @@
 Metadata-Version: 2.1
 Name: sunholo
-Version: 0.55.13
+Version: 0.55.15
 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/v0.55.13.tar.gz
+Download-URL: https://github.com/sunholo-data/sunholo-py/archive/refs/tags/v0.55.15.tar.gz
 Author: Holosun ApS
 Author-email: multivac@sunholo.com
 License: Apache License, Version 2.0