lionagi 0.0.106__tar.gz → 0.0.107__tar.gz

{lionagi-0.0.106 → lionagi-0.0.107}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: lionagi
-Version: 0.0.106
+Version: 0.0.107
 Summary: Towards automated general intelligence.
 Author: HaiyangLi
 Author-email: Haiyang Li <ocean@lionagi.ai>
@@ -222,18 +222,13 @@ Requires-Dist: httpx==0.25.1
 
 ![PyPI - Version](https://img.shields.io/pypi/v/lionagi?labelColor=233476aa&color=231fc935) ![PyPI - Downloads](https://img.shields.io/pypi/dm/lionagi?labelColor=233476aa&color=231fc935) ![GitHub License](https://img.shields.io/github/license/lion-agi/lionagi?labelColor=233476aa&color=231fc935) 
 
-
-
-- PyPI: https://pypi.org/project/lionagi/
-- Documentation: https://lionagi.readthedocs.io/en/latest/  （still a lot TODO）
-- Website: TODO
-- Discord: [Join Our Discord](https://discord.gg/7RGWqpSxze)
+[PyPI](https://pypi.org/project/lionagi/) | [Documentation](https://lionagi.readthedocs.io/en/latest/) | [Website](https://www.lionagi.ai) | [Discord](https://discord.gg/7RGWqpSxze)
 
   
 # LionAGI
 **Towards Automated General Intelligence**
 
-LionAGI is a Python package that combines data manipulation with AI tools, aiming to simplify the integration of advanced machine learning tools, such as Large Language Models (i.e. OpenAI's GPT), with production level data centric projects. 
+LionAGI is a Python intelligent agent framework that combines data manipulation with AI tools, aiming to simplify the integration of advanced machine learning tools, such as Large Language Models (i.e. OpenAI's GPT), with production-level data-centric projects. 
 
 Install LionAGI with pip:
 
@@ -244,13 +239,13 @@ Download the `.env_template` file, input your OPENAI_API_KEY, save the file, ren
 
 ### Features
 
-- Robust performance. LionAGI is written in almost pure python. With minimum external dependency (aiohttp, httpx, python-dotenv, tiktoken)
+- Robust performance. LionAGI is written in almost pure python. With minimum external dependency (`aiohttp`, `httpx`, `python-dotenv`, `tiktoken`)
 - Efficient data operations for reading, chunking, binning, writing, storing and managing data.
-- Fast interaction with LLM services like OpenAI with configurable rate limiting concurrent API calls for maximum throughput. 
-- Create a production ready LLM application in hours. Intuitive workflow management to streamline and expedite the process from idea to market.
+- Fast interaction with LLM services like OpenAI with **configurable rate limiting concurrent API calls** for maximum throughput. 
+- Create a production ready LLM application **in hours**. Intuitive workflow management to streamline and expedite the process from idea to market.
 
 ---
-Currently, LionAGI only natively support OpenAI API calls, support for other LLM providers as well as open source models will be integrated in future releases. LionAGI is designed to be async only, please check python documentation [here](https://docs.python.org/3/library/asyncio.html)
+Currently, LionAGI only natively support OpenAI API calls, support for other LLM providers as well as open source models will be integrated in future releases. LionAGI is designed to be async only, please check python official documentation on how `async` work: [here](https://docs.python.org/3/library/asyncio.html)
 
 
 **Notice**: 

{lionagi-0.0.106 → lionagi-0.0.107}/README.md

@@ -1,17 +1,12 @@
 ![PyPI - Version](https://img.shields.io/pypi/v/lionagi?labelColor=233476aa&color=231fc935) ![PyPI - Downloads](https://img.shields.io/pypi/dm/lionagi?labelColor=233476aa&color=231fc935) ![GitHub License](https://img.shields.io/github/license/lion-agi/lionagi?labelColor=233476aa&color=231fc935) 
 
-
-
-- PyPI: https://pypi.org/project/lionagi/
-- Documentation: https://lionagi.readthedocs.io/en/latest/  （still a lot TODO）
-- Website: TODO
-- Discord: [Join Our Discord](https://discord.gg/7RGWqpSxze)
+[PyPI](https://pypi.org/project/lionagi/) | [Documentation](https://lionagi.readthedocs.io/en/latest/) | [Website](https://www.lionagi.ai) | [Discord](https://discord.gg/7RGWqpSxze)
 
   
 # LionAGI
 **Towards Automated General Intelligence**
 
-LionAGI is a Python package that combines data manipulation with AI tools, aiming to simplify the integration of advanced machine learning tools, such as Large Language Models (i.e. OpenAI's GPT), with production level data centric projects. 
+LionAGI is a Python intelligent agent framework that combines data manipulation with AI tools, aiming to simplify the integration of advanced machine learning tools, such as Large Language Models (i.e. OpenAI's GPT), with production-level data-centric projects. 
 
 Install LionAGI with pip:
 
@@ -22,13 +17,13 @@ Download the `.env_template` file, input your OPENAI_API_KEY, save the file, ren
 
 ### Features
 
-- Robust performance. LionAGI is written in almost pure python. With minimum external dependency (aiohttp, httpx, python-dotenv, tiktoken)
+- Robust performance. LionAGI is written in almost pure python. With minimum external dependency (`aiohttp`, `httpx`, `python-dotenv`, `tiktoken`)
 - Efficient data operations for reading, chunking, binning, writing, storing and managing data.
-- Fast interaction with LLM services like OpenAI with configurable rate limiting concurrent API calls for maximum throughput. 
-- Create a production ready LLM application in hours. Intuitive workflow management to streamline and expedite the process from idea to market.
+- Fast interaction with LLM services like OpenAI with **configurable rate limiting concurrent API calls** for maximum throughput. 
+- Create a production ready LLM application **in hours**. Intuitive workflow management to streamline and expedite the process from idea to market.
 
 ---
-Currently, LionAGI only natively support OpenAI API calls, support for other LLM providers as well as open source models will be integrated in future releases. LionAGI is designed to be async only, please check python documentation [here](https://docs.python.org/3/library/asyncio.html)
+Currently, LionAGI only natively support OpenAI API calls, support for other LLM providers as well as open source models will be integrated in future releases. LionAGI is designed to be async only, please check python official documentation on how `async` work: [here](https://docs.python.org/3/library/asyncio.html)
 
 
 **Notice**: 

{lionagi-0.0.106 → lionagi-0.0.107}/lionagi/api/oai_service.py

@@ -19,20 +19,25 @@ class OpenAIRateLimiter(RateLimiter):
     and replenishing these limits at regular intervals.
 
     Attributes:
-        max_requests_per_minute (int): Maximum number of requests allowed per minute.
-        max_tokens_per_minute (int): Maximum number of tokens allowed per minute.
+        max_requests_per_minute (int):
+            Maximum number of requests allowed per minute.
+        max_tokens_per_minute (int):
+            Maximum number of tokens allowed per minute.
 
     Methods:
-        rate_limit_replenisher: Coroutine to replenish rate limits over time.
-        calculate_num_token: Calculates the required tokens for a request.
+        rate_limit_replenisher:
+            Coroutine to replenish rate limits over time.
+        calculate_num_token:
+            Calculates the required tokens for a request.
     """
 
     def __init__(self, max_requests_per_minute: int, max_tokens_per_minute: int) -> None:
         """
         Initializes the rate limiter with specific limits for OpenAI API.
 
-        Args:
+        Parameters:
             max_requests_per_minute (int): The maximum number of requests allowed per minute.
+
             max_tokens_per_minute (int): The maximum number of tokens that can accumulate per minute.
         """
         super().__init__(max_requests_per_minute, max_tokens_per_minute)
@@ -66,8 +71,9 @@ class OpenAIRateLimiter(RateLimiter):
         This method should be implemented in a subclass to provide the specific calculation logic
         for the OpenAI API.
 
-        Args:
+        Parameters:
             payload (Dict[str, Any]): The payload of the request.
+
             api_endpoint (str): The specific API endpoint for the request.
 
         Returns:
@@ -160,12 +166,17 @@ class OpenAIService(BaseAPIService):
         """
         Initializes the OpenAI service with configuration for API interaction.
 
-        Args:
+        Parameters:
             api_key (str): The API key for authenticating with OpenAI.
+
             token_encoding_name (str): The name of the text encoding used by OpenAI.
+
             max_attempts (int): The maximum number of attempts for calling an API endpoint.
+
             status_tracker (Optional[StatusTracker]): Tracker for API call outcomes.
-            rate_limiter (Optional[OpenAIRateLimiter]): Rate limiter for OpenAI's limits.
+
+            ratelimiter (Optional[OpenAIRateLimiter]): Rate limiter for OpenAI's limits.
+
             queue (Optional[AsyncQueue]): Queue for managing asynchronous API calls.
 
         Example:
@@ -183,15 +194,16 @@ class OpenAIService(BaseAPIService):
         super().__init__(api_key, token_encoding_name, max_attempts, 
                          max_requests_per_minute, max_tokens_per_minute,
                          ratelimiter, status_tracker, queue)
-        
 
-    async def call_api(self, http_session, endpoint, payload: Dict[str, any] =None) -> Optional[Dict[str, any]]:
+    async def call_api(self, http_session, endpoint, payload: Dict[str, any] = None) -> Optional[Dict[str, any]]:
         """
         Call an OpenAI API endpoint with a specific payload and handle the response.
 
-        Args:
-            session: The session object for making HTTP requests.
-            request_url (str): The full URL of the OpenAI API endpoint to be called.
+        Parameters:
+            http_session: The session object for making HTTP requests.
+
+            endpoint (str): The full URL of the OpenAI API endpoint to be called.
+
             payload (Dict[str, any]): The payload to send with the API request.
 
         Returns:

{lionagi-0.0.106 → lionagi-0.0.107}/lionagi/session/conversation.py

@@ -1,5 +1,6 @@
 from .message import Message
 
+
 class Conversation:
     """
     A class representing a conversation between users and the assistant.
@@ -8,21 +9,22 @@ class Conversation:
     user instructions, and assistant responses.
 
     Attributes:
-        response_counts (int): The count of assistant responses in the conversation.
-        messages (list): A list to store messages in the conversation.
-        msg (Message): An instance of the Message class for creating messages.
-        responses (list): A list to store assistant responses in the conversation.
+        response_counts (int):
+            The count of assistant responses in the conversation.
+        messages (list):
+            A list to store messages in the conversation.
+        msg (Message):
+            An instance of the Message class for creating messages.
+        responses (list):
+            A list to store assistant responses in the conversation.
 
     Methods:
         initiate_conversation(system, instruction, context=None, name=None):
             Initiate a conversation with a system setting and user instruction.
-
         add_messages(system, instruction, context=None, response=None, tool=None, name=None):
             Add messages to the conversation, including system setting, user instruction, and assistant response.
-
         change_system(system):
             Change the system setting in the conversation.
-
         keep_last_n_exchanges(n: int):
             Keep the last n exchanges in the conversation.
     """
@@ -46,8 +48,11 @@ class Conversation:
 
         Parameters:
             system (str): The system setting for the conversation.
+
             instruction (str): The user instruction to initiate the conversation.
+
             context (dict): Additional context for the conversation. Default is None.
+
             name (str): The name associated with the user. Default is None.
         """
         self.messages, self.responses = [], []
@@ -61,10 +66,13 @@ class Conversation:
 
         Parameters:
             system (str): The system setting for the message. Default is None.
+
             instruction (str): The instruction content for the message. Default is None.
+
             context (dict): Additional context for the message. Default is None.
+
             response (dict): The response content for the message. Default is None.
-            tool (dict): The tool information for the message. Default is None.
+
             name (str): The name associated with the message. Default is None.
         """
         msg = self.msg(system=system, instruction=instruction, context=context, 
@@ -93,4 +101,5 @@ class Conversation:
         ]
         if len(response_indices) >= n:
             first_index_to_keep = response_indices[-n] + 1
-            self.messages = [self.system] + self.messages[first_index_to_keep:]
+            self.messages = self.messages[0] + self.messages[first_index_to_keep:]
+            

{lionagi-0.0.106 → lionagi-0.0.107}/lionagi/session/message.py

@@ -11,11 +11,16 @@ class Message:
     This class encapsulates messages from users, the assistant, systems, and external tools.
 
     Attributes:
-        role (str): The role of the message, indicating if it's from the user, assistant, system, or tool.
-        content: The content of the message, which can be an instruction, response, system setting, or tool information.
-        name (str): The name associated with the message, specifying the source (user, assistant, system, or tool).
-        metadata (dict): Additional metadata including id, timestamp, and name.
-        _logger (DataLogger): An instance of the DataLogger class for logging message details.
+        role (str):
+            The role of the message, indicating if it's from the user, assistant, system, or tool.
+        content:
+            The content of the message, which can be an instruction, response, system setting, or tool information.
+        name (str):
+            The name associated with the message, specifying the source (user, assistant, system, or tool).
+        metadata (dict):
+            Additional metadata including id, timestamp, and name.
+        _logger (DataLogger):
+            An instance of the DataLogger class for logging message details.
 
     Methods:
         create_message(system, instruction, context, response, tool, name):
@@ -47,10 +52,13 @@ class Message:
 
         Parameters:
             system (str): The system setting for the message. Default is None.
+
             instruction (str): The instruction content for the message. Default is None.
+
             context (dict): Additional context for the message. Default is None.
+
             response (dict): The response content for the message. Default is None.
-            tool (dict): The tool information for the message. Default is None.
+
             name (str): The name associated with the message. Default is None.
         """
         if sum(l_call([system, instruction, response], bool)) > 1:
@@ -63,10 +71,17 @@ class Message:
                     response = response["message"]
                     if str(response['content']) == "None":
                         try:
-                            if response['tool_calls'][0]['type'] == 'function':
-                                self.name = name or ("func_" + response['tool_calls'][0]['function']['name'])
-                                content = response['tool_calls'][0]['function']['arguments']
-                                self.content = {"function":self.name, "arguments": content}
+                            tool_count = 0
+                            func_list = []
+                            while tool_count < len(response['tool_calls']):
+                                if response['tool_calls'][tool_count]['type'] == 'function':
+                                    func_content = {"function": ("func_" + response['tool_calls'][tool_count]['function']['name']),
+                                                    "arguments": response['tool_calls'][tool_count]['function']['arguments']}
+                                    func_list.append(func_content)
+                                tool_count += 1
+
+                            self.name = name or "func_request"
+                            self.content = {'function_list': func_list}
                         except:
                             raise ValueError("Response message must be one of regular response or function calling")
                     else:
@@ -74,7 +89,7 @@ class Message:
                         self.name = name or "assistant"
                 except:
                     self.name = name or "func_call"
-                    self.content = {"function call result": response}
+                    self.content = response
                 
             elif instruction:
                 self.role = "user"
@@ -92,7 +107,7 @@ class Message:
         Convert the message to a JSON format.
 
         Returns:
-        - dict: The message in JSON format.
+            dict: The message in JSON format.
         """
         out = {
             "role": self.role,
@@ -114,11 +129,14 @@ class Message:
 
         Parameters:
             system (str): The system setting for the message. Default is None.
+
             instruction (str): The instruction content for the message. Default is None.
+
             context (dict): Additional context for the message. Default is None.
+
             response (dict): The response content for the message. Default is None.
+
             name (str): The name associated with the message. Default is None.
-            tool (dict): The tool information for the message. Default is None.
 
         Returns:
             dict: The message in JSON format.
@@ -133,10 +151,16 @@ class Message:
 
         Parameters:
             dir (str): The directory path for saving the CSV file. Default is None.
+
             filename (str): The filename for the CSV file. Default is None.
+
             verbose (bool): Whether to include verbose information in the CSV. Default is True.
+
             timestamp (bool): Whether to include timestamps in the CSV. Default is True.
+
             dir_exist_ok (bool): Whether to allow the directory to exist. Default is True.
+
             file_exist_ok (bool): Whether to allow the file to exist. Default is False.
         """
-        self._logger.to_csv(dir, filename, verbose, timestamp, dir_exist_ok, file_exist_ok)
+        self._logger.to_csv(dir, filename, verbose, timestamp, dir_exist_ok, file_exist_ok)
+