sunholo 0.140.4__tar.gz → 0.140.5__tar.gz

{sunholo-0.140.4/src/sunholo.egg-info → sunholo-0.140.5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sunholo
-Version: 0.140.4
+Version: 0.140.5
 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-0.140.4 → sunholo-0.140.5}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "sunholo"
-version = "0.140.4"
+version = "0.140.5"
 description = "AI DevOps - a package to help deploy GenAI to the Cloud."
 readme = "README.md"
 requires-python = ">=3.10"

sunholo-0.140.5/src/sunholo/agents/chat_history.py

@@ -0,0 +1,523 @@
+import json
+from ..custom_logging import log
+import time
+import hashlib
+from functools import lru_cache
+from typing import List, Tuple, Optional
+
+
+class ChatHistoryCache:
+    """
+    Incremental cache for chat history processing.
+    
+    Caches processed message pairs and only processes new messages
+    when the chat history is extended.
+    """
+    
+    def __init__(self, max_cache_size: int = 1000):
+        self.cache = {}
+        self.max_cache_size = max_cache_size
+    
+    def _get_cache_key(self, chat_history: List[dict]) -> str:
+        """Generate a cache key based on the chat history content."""
+        # Use the hash of the serialized chat history for the key
+        # Only hash the first few and last few messages to balance performance vs accuracy
+        if len(chat_history) <= 10:
+            content = str(chat_history)
+        else:
+            # Hash first 5 and last 5 messages + length
+            content = str(chat_history[:5] + chat_history[-5:] + [len(chat_history)])
+        
+        return hashlib.md5(content.encode()).hexdigest()
+    
+    def _find_cached_prefix(self, current_history: List[dict]) -> Tuple[Optional[List[Tuple]], int]:
+        """
+        Find the longest cached prefix of the current chat history.
+        
+        Returns:
+            Tuple of (cached_pairs, cache_length) or (None, 0) if no cache found
+        """
+        current_length = len(current_history)
+        
+        # Check for cached versions of prefixes, starting from longest
+        for cache_length in range(current_length - 1, 0, -1):
+            prefix = current_history[:cache_length]
+            cache_key = self._get_cache_key(prefix)
+            
+            if cache_key in self.cache:
+                cached_data = self.cache[cache_key]
+                cached_pairs = cached_data['pairs']
+                
+                # Verify the cache is still valid by checking a few messages
+                if self._verify_cache_validity(prefix, cached_data['original_history']):
+                    return cached_pairs, cache_length
+                else:
+                    # Cache is stale, remove it
+                    del self.cache[cache_key]
+        
+        return None, 0
+    
+    def _verify_cache_validity(self, current_prefix: List[dict], cached_prefix: List[dict]) -> bool:
+        """Quick verification that cached data is still valid."""
+        if len(current_prefix) != len(cached_prefix):
+            return False
+        
+        # Check first and last few messages for equality
+        check_indices = [0, -1] if len(current_prefix) >= 2 else [0]
+        
+        for i in check_indices:
+            if current_prefix[i] != cached_prefix[i]:
+                return False
+        
+        return True
+    
+    def extract_chat_history_incremental(self, chat_history: List[dict]) -> List[Tuple]:
+        """
+        Extract chat history with incremental caching.
+        
+        Args:
+            chat_history: List of chat message dictionaries
+            
+        Returns:
+            List of (human_message, ai_message) tuples
+        """
+        if not chat_history:
+            return []
+        
+        # Try to find cached prefix
+        cached_pairs, cache_length = self._find_cached_prefix(chat_history)
+        
+        if cached_pairs is not None:
+            log.debug(f"Found cached pairs for {cache_length} messages, processing {len(chat_history) - cache_length} new messages")
+            
+            # Process only the new messages
+            new_messages = chat_history[cache_length:]
+            new_pairs = self._process_new_messages(new_messages, cached_pairs)
+            
+            # Combine cached and new pairs
+            all_pairs = cached_pairs + new_pairs
+        else:
+            log.debug(f"No cache found, processing all {len(chat_history)} messages")
+            # Process all messages from scratch
+            all_pairs = self._extract_chat_history_full(chat_history)
+        
+        # Cache the result
+        self._update_cache(chat_history, all_pairs)
+        
+        return all_pairs
+    
+    def _process_new_messages(self, new_messages: List[dict], cached_pairs: List[Tuple]) -> List[Tuple]:
+        """
+        Process only the new messages, considering the state from cached pairs.
+        
+        Args:
+            new_messages: New messages to process
+            cached_pairs: Previously processed message pairs
+            
+        Returns:
+            List of new message pairs
+        """
+        if not new_messages:
+            return []
+        
+        new_pairs = []
+        
+        # Determine if we're waiting for a bot response based on cached pairs
+        waiting_for_bot = True
+        if cached_pairs:
+            last_pair = cached_pairs[-1]
+            # If last pair has both human and AI message, we're ready for a new human message
+            waiting_for_bot = not (last_pair[0] and last_pair[1])
+        
+        # If we ended with an unpaired human message, get it
+        last_human_message = ""
+        if cached_pairs and waiting_for_bot:
+            last_human_message = cached_pairs[-1][0]
+        
+        # Process new messages
+        for message in new_messages:
+            try:
+                is_human_msg = is_human(message)
+                content = create_message_element(message)
+                
+                if is_human_msg:
+                    last_human_message = content
+                    waiting_for_bot = True
+                else:  # Bot message
+                    if waiting_for_bot and last_human_message:
+                        new_pairs.append((last_human_message, content))
+                        last_human_message = ""
+                        waiting_for_bot = False
+                    # If not waiting for bot or no human message, this is an orphaned bot message
+                    
+            except (KeyError, TypeError) as e:
+                log.warning(f"Error processing new message: {e}")
+                continue
+        
+        return new_pairs
+    
+    def _extract_chat_history_full(self, chat_history: List[dict]) -> List[Tuple]:
+        """Full extraction when no cache is available."""
+        # Use the optimized version from before
+        paired_messages = []
+        
+        # Handle initial bot message
+        start_idx = 0
+        if chat_history and is_bot(chat_history[0]):
+            try:
+                first_message = chat_history[0]
+                blank_element = ""
+                bot_element = create_message_element(first_message)
+                paired_messages.append((blank_element, bot_element))
+                start_idx = 1
+            except (KeyError, TypeError):
+                pass
+        
+        # Process remaining messages
+        last_human_message = ""
+        for i in range(start_idx, len(chat_history)):
+            message = chat_history[i]
+            
+            try:
+                is_human_msg = is_human(message)
+                content = create_message_element(message)
+                
+                if is_human_msg:
+                    last_human_message = content
+                else:  # Bot message
+                    if last_human_message:
+                        paired_messages.append((last_human_message, content))
+                        last_human_message = ""
+                        
+            except (KeyError, TypeError) as e:
+                log.warning(f"Error processing message {i}: {e}")
+                continue
+        
+        return paired_messages
+    
+    def _update_cache(self, chat_history: List[dict], pairs: List[Tuple]):
+        """Update cache with new result."""
+        # Only cache if the history is of reasonable size
+        if len(chat_history) < 2:
+            return
+        
+        cache_key = self._get_cache_key(chat_history)
+        
+        # Implement simple LRU by removing oldest entries
+        if len(self.cache) >= self.max_cache_size:
+            # Remove 20% of oldest entries
+            remove_count = self.max_cache_size // 5
+            oldest_keys = list(self.cache.keys())[:remove_count]
+            for key in oldest_keys:
+                del self.cache[key]
+        
+        self.cache[cache_key] = {
+            'pairs': pairs,
+            'original_history': chat_history.copy(),  # Store copy for validation
+            'timestamp': time.time()
+        }
+        
+        log.debug(f"Cached {len(pairs)} pairs for history of length {len(chat_history)}")
+    
+    def clear_cache(self):
+        """Clear the entire cache."""
+        self.cache.clear()
+        log.info("Chat history cache cleared")
+
+
+# Global cache instance
+_chat_history_cache = ChatHistoryCache()
+
+
+def extract_chat_history_with_cache(chat_history: List[dict] = None) -> List[Tuple]:
+    """
+    Main function to replace the original extract_chat_history.
+    
+    Uses incremental caching for better performance with growing chat histories.
+    """
+    if not chat_history:
+        log.debug("No chat history found")
+        return []
+    
+    return _chat_history_cache.extract_chat_history_incremental(chat_history)
+
+
+# Async version that wraps the cached version
+async def extract_chat_history_async_cached(chat_history: List[dict] = None) -> List[Tuple]:
+    """
+    Async version that uses the cache and runs in a thread pool if needed.
+    """
+    import asyncio
+    
+    if not chat_history:
+        return []
+    
+    # For very large histories, run in thread pool to avoid blocking
+    if len(chat_history) > 1000:
+        loop = asyncio.get_event_loop()
+        return await loop.run_in_executor(
+            None, 
+            extract_chat_history_with_cache, 
+            chat_history
+        )
+    else:
+        # For smaller histories, just run directly
+        return extract_chat_history_with_cache(chat_history)
+
+
+# Utility function to warm up the cache
+def warm_up_cache(chat_histories: List[List[dict]]):
+    """
+    Pre-populate cache with common chat histories.
+    
+    Args:
+        chat_histories: List of chat history lists to cache
+    """
+    for history in chat_histories:
+        extract_chat_history_with_cache(history)
+    
+    log.info(f"Warmed up cache with {len(chat_histories)} chat histories")
+
+
+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.
+
+    This function takes a chat history and returns a list of pairs of messages,
+    where each pair consists of a human message followed by the corresponding AI response.
+
+    Args:
+        chat_history (list): List of chat messages.
+
+    Returns:
+        list: List of tuples with paired human and AI messages.
+
+    Example:
+    ```python
+    chat_history = [
+        {"name": "Human", "text": "Hello, AI!"},
+        {"name": "AI", "text": "Hello, Human! How can I help you today?"}
+    ]
+    paired_messages = extract_chat_history(chat_history)
+    print(paired_messages)
+    # Output: [("Hello, AI!", "Hello, Human! How can I help you today?")]
+    ```
+    """
+    if not chat_history:
+        log.info("No chat history found")
+        return []
+
+    log.info(f"Extracting chat history: {chat_history}")
+    paired_messages = []
+
+    first_message = chat_history[0]
+    log.info(f"Extracting first_message: {first_message}")
+    if is_bot(first_message):
+        blank_human_message = {"name": "Human", "content": "", "embeds": []}
+        paired_messages.append((create_message_element(blank_human_message), 
+                                create_message_element(first_message)))
+        chat_history = chat_history[1:]
+
+    last_human_message = ""
+    for message in chat_history:
+        log.info(f"Extracing message: {message}")
+        if is_human(message):
+            last_human_message = create_message_element(message)
+            log.info(f"Extracted human message: {last_human_message}")
+        elif is_bot(message):
+            ai_message = create_message_element(message)
+            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 embeds_to_json(message: dict):
+    """
+    Converts the 'embeds' field in a message to a JSON string.
+
+    Args:
+        message (dict): The message containing the 'embeds' field.
+
+    Returns:
+        str: JSON string representation of the 'embeds' field or an empty string if no embeds are found.
+
+    Example:
+    ```python
+    message = {"embeds": [{"type": "image", "url": "https://example.com/image.png"}]}
+    json_string = embeds_to_json(message)
+    print(json_string)
+    # Output: '[{"type": "image", "url": "https://example.com/image.png"}]'
+    ```
+    """
+    if 'embeds' in message and len(message['embeds']) > 0:
+        return json.dumps(message.get("embeds"))
+    else:
+        return ""
+
+def create_message_element(message: dict):
+    """
+    Extracts the main content of a message.
+
+    Args:
+        message (dict): The message to extract content from.
+
+    Returns:
+        str: The text or content of the message.
+
+    Raises:
+        KeyError: If neither 'content' nor 'text' fields are found.
+
+    Example:
+    ```python
+    message = {"text": "Hello, AI!"}
+    content = create_message_element(message)
+    print(content)
+    # Output: 'Hello, AI!'
+    ```
+    """
+    if 'text' in message:  # This is a Slack or Google Chat message
+        log.info(f"Found text element - {message['text']}")
+        return message['text']
+    elif 'content' in message: # Discord or OpenAI history message
+        log.info(f"Found content element - {message['content']}")
+        return message['content']
+    else:  
+        raise KeyError(f"Could not extract 'content' or 'text' element from message: {message}, {type(message)}")
+
+def is_human(message: dict):
+    """
+    Checks if a message was sent by a human.
+
+    Args:
+        message (dict): The message to check.
+
+    Returns:
+        bool: True if the message was sent by a human, otherwise False.
+
+    Example:
+    ```python
+    message = {"name": "Human"}
+    print(is_human(message))
+    # Output: True
+    ```
+    """
+    if 'name' in message:
+        return message["name"] == "Human"
+    elif 'sender' in message:  # Google Chat
+        return message['sender']['type'] == 'HUMAN'
+    elif 'role' in message:
+        return message['role'] == 'user'
+    else:
+        # Slack: Check for the 'user' field and absence of 'bot_id' field
+        return 'user' in message and 'bot_id' not in message
+
+def is_bot(message: dict):
+    """
+    Checks if a message was sent by a bot.
+
+    Args:
+        message (dict): The message to check.
+
+    Returns:
+        bool: True if the message was sent by a bot, otherwise False.
+
+    Example:
+    ```python
+    message = {"name": "AI"}
+    print(is_bot(message))
+    # Output: True
+    ```
+    """
+    return not is_human(message)
+
+def is_ai(message: dict):
+    """
+    Checks if a message was specifically sent by an AI.
+
+    Args:
+        message (dict): The message to check.
+
+    Returns:
+        bool: True if the message was sent by an AI, otherwise False.
+
+    Example:
+    ```python
+    message = {"name": "AI"}
+    print(is_ai(message))
+    # Output: True
+    ```
+    """
+    if 'name' in message:
+        return message["name"] == "AI"
+    elif 'sender' in message:  # Google Chat
+        return message['sender']['type'] == 'BOT'
+    elif 'role' in message:
+        return message['role'] == 'assistant'
+    else:
+        return 'bot_id' in message  # Slack
+

{sunholo-0.140.4 → sunholo-0.140.5}/src/sunholo/agents/flask/vac_routes.py

@@ -7,8 +7,8 @@ from functools import partial
 import inspect
 import asyncio
 
-from ...agents import extract_chat_history, handle_special_commands
-from ..chat_history import extract_chat_history_async
+from ...agents import handle_special_commands
+from ..chat_history import extract_chat_history_with_cache, extract_chat_history_async_cached
 from ...qna.parsers import parse_output
 from ...streaming import start_streaming_chat, start_streaming_chat_async
 from ...archive import archive_qa
@@ -58,12 +58,18 @@ if __name__ == "__main__":
 ```
     
     """
-    def __init__(self, app, stream_interpreter: callable, vac_interpreter:callable=None, additional_routes:dict=None, async_stream:bool=False):
+    def __init__(self, app, 
+                 stream_interpreter: callable, 
+                 vac_interpreter:callable=None, 
+                 additional_routes:dict=None, 
+                 async_stream:bool=False,
+                 add_langfuse_eval:bool=True):
         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.add_langfuse_eval = add_langfuse_eval
         self.register_routes()
         
 
@@ -550,7 +556,8 @@ if __name__ == "__main__":
         else:
             log.info(f"User message: {user_message}")
     
-        paired_messages = extract_chat_history(chat_history)
+        paired_messages = extract_chat_history_with_cache(chat_history)
+
         command_response = handle_special_commands(user_message, vector_name, paired_messages)
 
         if command_response is not None:
@@ -694,10 +701,10 @@ if __name__ == "__main__":
 
         trace = None
         span = None
-
-        trace_id = data.get('trace_id')
-        trace = self.create_langfuse_trace(request, vector_name, trace_id)
-        log.info(f"Using existing langfuse trace: {trace_id}")
+        if self.add_langfuse_eval:
+            trace_id = data.get('trace_id')
+            trace = self.create_langfuse_trace(request, vector_name, trace_id)
+            log.info(f"Using existing langfuse trace: {trace_id}")
         
         #config, _ = load_config("config/llm_config.yaml")
         try:
@@ -721,7 +728,7 @@ if __name__ == "__main__":
         vector_name = data.pop('vector_name', vector_name)
         data.pop('trace_id', None) # to ensure not in kwargs
 
-        paired_messages = extract_chat_history(chat_history)
+        paired_messages = extract_chat_history_with_cache(chat_history)
 
         all_input = {'user_input': user_input, 
                      'vector_name': vector_name, 
@@ -737,15 +744,10 @@ if __name__ == "__main__":
                 metadata=vac_config.configs_by_kind,
                 input = all_input
             )
-        command_response = handle_special_commands(user_input, vector_name, paired_messages)
-        if command_response is not None:
-            if trace:
-                trace.update(output=jsonify(command_response))
     
         return {
             "trace": trace,
             "span": span,
-            "command_response": command_response,
             "all_input": all_input,
             "vac_config": vac_config
         }
@@ -789,7 +791,7 @@ if __name__ == "__main__":
         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))
+        chat_history_task = asyncio.create_task(extract_chat_history_async_cached(chat_history))
         tasks.append(chat_history_task)
         
         # Await all tasks concurrently

{sunholo-0.140.4 → sunholo-0.140.5/src/sunholo.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sunholo
-Version: 0.140.4
+Version: 0.140.5
 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