sunholo 0.139.1__tar.gz → 0.140.4__tar.gz

{sunholo-0.139.1/src/sunholo.egg-info → sunholo-0.140.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sunholo
-Version: 0.139.1
+Version: 0.140.4
 Summary: AI DevOps - a package to help deploy GenAI to the Cloud.
 Author-email: Holosun ApS <multivac@sunholo.com>
 License: Apache License, Version 2.0

{sunholo-0.139.1 → sunholo-0.140.4}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "sunholo"
-version = "0.139.1"
+version = "0.140.4"
 description = "AI DevOps - a package to help deploy GenAI to the Cloud."
 readme = "README.md"
 requires-python = ">=3.10"

{sunholo-0.139.1 → sunholo-0.140.4}/src/sunholo/agents/__init__.py

@@ -2,6 +2,6 @@ from .chat_history import extract_chat_history
 from .dispatch_to_qa import send_to_qa, send_to_qa_async
 from .pubsub import process_pubsub
 from .special_commands import handle_special_commands, app_to_store, handle_files
-from .flask import register_qna_routes, create_app, VACRoutes
+from .flask import create_app, VACRoutes
 from .fastapi import register_qna_fastapi_routes, create_fastapi_app
 from .swagger import config_to_swagger

{sunholo-0.139.1 → sunholo-0.140.4}/src/sunholo/agents/chat_history.py

@@ -1,6 +1,69 @@
 import json
 from ..custom_logging import log
 
+
+async def extract_chat_history_async(chat_history=None):
+    """
+    Extracts paired chat history between human and AI messages.
+    
+    For this lightweight processing, we use a simpler approach that minimizes overhead.
+    
+    Args:
+        chat_history (list): List of chat messages.
+    
+    Returns:
+        list: List of tuples with paired human and AI messages.
+    """
+    if not chat_history:
+        log.info("No chat history found")
+        return []
+
+    log.info(f"Extracting chat history: {chat_history}")
+    paired_messages = []
+    
+    # Handle special case of initial bot message
+    if chat_history and is_bot(chat_history[0]):
+        first_message = chat_history[0]
+        log.info(f"Extracting first_message: {first_message}")
+        blank_human_message = {"name": "Human", "content": "", "embeds": []}
+        
+        # Since create_message_element is so lightweight, we don't need async here
+        blank_element = create_message_element(blank_human_message)
+        bot_element = create_message_element(first_message)
+        
+        paired_messages.append((blank_element, bot_element))
+        chat_history = chat_history[1:]
+    
+    # Pre-process all messages in one batch (more efficient than one-by-one)
+    message_types = []
+    message_contents = []
+    
+    for message in chat_history:
+        is_human_msg = is_human(message)
+        is_bot_msg = is_bot(message)
+        
+        # Extract content for all messages at once
+        content = create_message_element(message)
+        
+        message_types.append((is_human_msg, is_bot_msg))
+        message_contents.append(content)
+    
+    # Pair messages efficiently
+    last_human_message = ""
+    for i, ((is_human_msg, is_bot_msg), content) in enumerate(zip(message_types, message_contents)):
+        if is_human_msg:
+            last_human_message = content
+            log.info(f"Extracted human message: {last_human_message}")
+        elif is_bot_msg:
+            ai_message = content
+            log.info(f"Extracted AI message: {ai_message}")
+            paired_messages.append((last_human_message, ai_message))
+            last_human_message = ""
+    
+    log.info(f"Paired messages: {paired_messages}")
+    return paired_messages
+
+
 def extract_chat_history(chat_history=None):
     """
     Extracts paired chat history between human and AI messages.

{sunholo-0.139.1 → sunholo-0.140.4}/src/sunholo/agents/flask/__init__.py

@@ -1,3 +1,2 @@
-from .qna_routes import register_qna_routes
 from .base import create_app
 from .vac_routes import VACRoutes

{sunholo-0.139.1 → sunholo-0.140.4}/src/sunholo/agents/flask/vac_routes.py

@@ -8,6 +8,7 @@ import inspect
 import asyncio
 
 from ...agents import extract_chat_history, handle_special_commands
+from ..chat_history import extract_chat_history_async
 from ...qna.parsers import parse_output
 from ...streaming import start_streaming_chat, start_streaming_chat_async
 from ...archive import archive_qa
@@ -57,11 +58,12 @@ if __name__ == "__main__":
 ```
     
     """
-    def __init__(self, app, stream_interpreter, vac_interpreter=None, additional_routes=None):
+    def __init__(self, app, stream_interpreter: callable, vac_interpreter:callable=None, additional_routes:dict=None, async_stream:bool=False):
         self.app = app
         self.stream_interpreter = stream_interpreter
         self.vac_interpreter = vac_interpreter or partial(self.vac_interpreter_default)
         self.additional_routes = additional_routes if additional_routes is not None else []
+        self.async_stream = async_stream
         self.register_routes()
         
 
@@ -94,12 +96,16 @@ if __name__ == "__main__":
         # Basic routes
         self.app.route("/", methods=['GET'])(self.home)
         self.app.route("/health", methods=['GET'])(self.health)
-
-        # Streaming VAC
-        self.app.route('/vac/streaming/<vector_name>', 
-                       methods=['POST'], 
-                       provide_automatic_options=False)(self.handle_stream_vac)
-
+        
+        if self.async_stream:  # Use async treatment
+            log.info("async_stream enabled")
+            self.app.route('/vac/streaming/<vector_name>', 
+                        methods=['POST'], 
+                        provide_automatic_options=False)(self.handle_stream_vac_async)
+        else:
+            self.app.route('/vac/streaming/<vector_name>', 
+                        methods=['POST'], 
+                        provide_automatic_options=False)(self.handle_stream_vac)
         # Static VAC
         self.app.route('/vac/<vector_name>', 
                        methods=['POST'], 
@@ -332,6 +338,51 @@ if __name__ == "__main__":
 
         return response
 
+    async def handle_stream_vac_async(self, vector_name):
+        observed_stream_interpreter = self.stream_interpreter
+        is_async = inspect.iscoroutinefunction(self.stream_interpreter)
+
+        if not is_async:
+            raise ValueError(f"Stream interpreter must be async: {observed_stream_interpreter}")
+
+        # Use the async version of prep_vac
+        prep = await self.prep_vac_async(request, vector_name)
+        log.info(f"Processing async prep: {prep}")
+        all_input = prep["all_input"]
+
+        log.info(f'Streaming async data with: {all_input}')
+
+        async def generate_response_content():
+            try:
+                # Direct async handling without the queue/thread approach
+                async_gen = start_streaming_chat_async(
+                    question=all_input["user_input"],
+                    vector_name=vector_name,
+                    qna_func_async=observed_stream_interpreter,
+                    chat_history=all_input["chat_history"],
+                    wait_time=all_input["stream_wait_time"],
+                    timeout=all_input["stream_timeout"],
+                    **all_input["kwargs"]
+                )
+                
+                log.info(f"{async_gen=}")
+                async for chunk in async_gen:
+                    if isinstance(chunk, dict) and 'answer' in chunk:
+                        await archive_qa(chunk, vector_name)
+                        yield json.dumps(chunk)
+                    else:
+                        yield chunk
+
+            except Exception as e:
+                yield f"Streaming async Error: {str(e)} {traceback.format_exc()}"
+
+        response = Response(generate_response_content(), content_type='text/plain; charset=utf-8')
+        response.headers['Transfer-Encoding'] = 'chunked'
+
+        log.debug(f"streaming async response: {response}")
+
+        return response
+
     @staticmethod
     async def _async_generator_to_stream(async_gen_func):
         """Helper function to stream the async generator's values to the client."""
@@ -699,6 +750,66 @@ if __name__ == "__main__":
             "vac_config": vac_config
         }
 
+    async def prep_vac_async(self, request, vector_name):
+        """Async version of prep_vac."""
+        # Parse request data
+        if request.content_type.startswith('application/json'):
+            data = request.get_json()
+        elif request.content_type.startswith('multipart/form-data'):
+            data = request.form.to_dict()
+            if 'file' in request.files:
+                file = request.files['file']
+                if file.filename != '':
+                    log.info(f"Found file: {file.filename} to upload to GCS")
+                    try:
+                        # Make file upload async if possible
+                        image_uri, mime_type = await self.handle_file_upload_async(file, vector_name)
+                        data["image_uri"] = image_uri
+                        data["mime"] = mime_type
+                    except Exception as e:
+                        log.error(traceback.format_exc())
+                        return jsonify({'error': str(e), 'traceback': traceback.format_exc()}), 500
+                else:
+                    log.error("No file selected")
+                    return jsonify({"error": "No file selected"}), 400
+        else:
+            return jsonify({"error": "Unsupported content type"}), 400
+
+        log.info(f"vac/{vector_name} got data: {data}")
+
+        # Run these operations concurrently
+        tasks = []
+        
+        # Extract other data while configs load
+        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)
+        vector_name_param = data.pop('vector_name', vector_name)
+        data.pop('trace_id', None)  # to ensure not in kwargs
+        
+        # Task 3: Process chat history
+        chat_history_task = asyncio.create_task(extract_chat_history_async(chat_history))
+        tasks.append(chat_history_task)
+        
+        # Await all tasks concurrently
+        results = await asyncio.gather(*tasks, return_exceptions=True)
+
+        paired_messages = results[0] if not isinstance(results[0], Exception) else []
+        
+        # Only create span after we have trace
+        all_input = {
+            'user_input': user_input, 
+            'vector_name': vector_name_param, 
+            'chat_history': paired_messages, 
+            'stream_wait_time': stream_wait_time,
+            'stream_timeout': stream_timeout,
+            'kwargs': data
+        }
+        
+        return {
+            "all_input": all_input
+        }
 
     def handle_file_upload(self, file, vector_name):
         try:

{sunholo-0.139.1 → sunholo-0.140.4/src/sunholo.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sunholo
-Version: 0.139.1
+Version: 0.140.4
 Summary: AI DevOps - a package to help deploy GenAI to the Cloud.
 Author-email: Holosun ApS <multivac@sunholo.com>
 License: Apache License, Version 2.0

{sunholo-0.139.1 → sunholo-0.140.4}/src/sunholo.egg-info/SOURCES.txt

@@ -24,7 +24,6 @@ src/sunholo/agents/fastapi/base.py
 src/sunholo/agents/fastapi/qna_routes.py
 src/sunholo/agents/flask/__init__.py
 src/sunholo/agents/flask/base.py
-src/sunholo/agents/flask/qna_routes.py
 src/sunholo/agents/flask/vac_routes.py
 src/sunholo/archive/__init__.py
 src/sunholo/archive/archive.py