swarms 7.8.4__py3-none-any.whl → 7.8.8__py3-none-any.whl

swarms/utils/litellm_wrapper.py

@@ -1,6 +1,8 @@
+import traceback
 from typing import Optional
 import base64
 import requests
+from pathlib import Path
 
 import asyncio
 from typing import List
@@ -9,11 +11,7 @@ from loguru import logger
 import litellm
 from pydantic import BaseModel
 
-from litellm import completion, acompletion
-
-litellm.set_verbose = True
-litellm.ssl_verify = False
-# litellm._turn_on_debug()
+from litellm import completion, acompletion, supports_vision
 
 
 class LiteLLMException(Exception):
@@ -53,6 +51,35 @@ def get_audio_base64(audio_source: str) -> str:
     return encoded_string
 
 
+def get_image_base64(image_source: str) -> str:
+    """
+    Convert image from a given source to a base64 encoded string.
+    Handles URLs, local file paths, and data URIs.
+    """
+    # If already a data URI, return as is
+    if image_source.startswith("data:image"):
+        return image_source
+
+    # Handle URL
+    if image_source.startswith(("http://", "https://")):
+        response = requests.get(image_source)
+        response.raise_for_status()
+        image_data = response.content
+    # Handle local file
+    else:
+        with open(image_source, "rb") as file:
+            image_data = file.read()
+
+    # Get file extension for mime type
+    extension = Path(image_source).suffix.lower()
+    mime_type = (
+        f"image/{extension[1:]}" if extension else "image/jpeg"
+    )
+
+    encoded_string = base64.b64encode(image_data).decode("utf-8")
+    return f"data:{mime_type};base64,{encoded_string}"
+
+
 class LiteLLM:
     """
     This class represents a LiteLLM.
@@ -72,12 +99,15 @@ class LiteLLM:
         tool_choice: str = "auto",
         parallel_tool_calls: bool = False,
         audio: str = None,
-        retries: int = 3,
+        retries: int = 0,
         verbose: bool = False,
         caching: bool = False,
         mcp_call: bool = False,
         top_p: float = 1.0,
         functions: List[dict] = None,
+        return_all: bool = False,
+        base_url: str = None,
+        api_key: str = None,
         *args,
         **kwargs,
     ):
@@ -105,8 +135,11 @@ class LiteLLM:
         self.mcp_call = mcp_call
         self.top_p = top_p
         self.functions = functions
+        self.audio = audio
+        self.return_all = return_all
+        self.base_url = base_url
+        self.api_key = api_key
         self.modalities = []
-        self._cached_messages = {}  # Cache for prepared messages
         self.messages = []  # Initialize messages list
 
         # Configure litellm settings
@@ -135,7 +168,11 @@ class LiteLLM:
                 out = out.model_dump()
             return out
 
-    def _prepare_messages(self, task: str) -> list:
+    def _prepare_messages(
+        self,
+        task: str,
+        img: str = None,
+    ):
         """
         Prepare the messages for the given task.
 
@@ -145,91 +182,201 @@ class LiteLLM:
         Returns:
             list: A list of messages prepared for the task.
         """
-        # Check cache first
-        cache_key = f"{self.system_prompt}:{task}"
-        if cache_key in self._cached_messages:
-            return self._cached_messages[cache_key].copy()
+        self.check_if_model_supports_vision(img=img)
 
+        # Initialize messages
         messages = []
-        if self.system_prompt:
+
+        # Add system prompt if present
+        if self.system_prompt is not None:
             messages.append(
                 {"role": "system", "content": self.system_prompt}
             )
-        messages.append({"role": "user", "content": task})
 
-        # Cache the prepared messages
-        self._cached_messages[cache_key] = messages.copy()
+        # Handle vision case
+        if img is not None:
+            messages = self.vision_processing(
+                task=task, image=img, messages=messages
+            )
+        else:
+            messages.append({"role": "user", "content": task})
+
         return messages
 
-    def audio_processing(self, task: str, audio: str):
+    def anthropic_vision_processing(
+        self, task: str, image: str, messages: list
+    ) -> list:
         """
-        Process the audio for the given task.
-
-        Args:
-            task (str): The task to be processed.
-            audio (str): The path or identifier for the audio file.
+        Process vision input specifically for Anthropic models.
+        Handles Anthropic's specific image format requirements.
         """
-        self.modalities.append("audio")
-
-        encoded_string = get_audio_base64(audio)
+        # Get base64 encoded image
+        image_url = get_image_base64(image)
+
+        # Extract mime type from the data URI or use default
+        mime_type = "image/jpeg"  # default
+        if "data:" in image_url and ";base64," in image_url:
+            mime_type = image_url.split(";base64,")[0].split("data:")[
+                1
+            ]
+
+        # Ensure mime type is one of the supported formats
+        supported_formats = [
+            "image/jpeg",
+            "image/png",
+            "image/gif",
+            "image/webp",
+        ]
+        if mime_type not in supported_formats:
+            mime_type = (
+                "image/jpeg"  # fallback to jpeg if unsupported
+            )
 
-        # Append messages
-        self.messages.append(
+        # Construct Anthropic vision message
+        messages.append(
             {
                 "role": "user",
                 "content": [
                     {"type": "text", "text": task},
                     {
-                        "type": "input_audio",
-                        "input_audio": {
-                            "data": encoded_string,
-                            "format": "wav",
+                        "type": "image_url",
+                        "image_url": {
+                            "url": image_url,
+                            "format": mime_type,
                         },
                     },
                 ],
             }
         )
 
-    def vision_processing(self, task: str, image: str):
+        return messages
+
+    def openai_vision_processing(
+        self, task: str, image: str, messages: list
+    ) -> list:
+        """
+        Process vision input specifically for OpenAI models.
+        Handles OpenAI's specific image format requirements.
+        """
+        # Get base64 encoded image with proper format
+        image_url = get_image_base64(image)
+
+        # Prepare vision message
+        vision_message = {
+            "type": "image_url",
+            "image_url": {"url": image_url},
+        }
+
+        # Add format for specific models
+        extension = Path(image).suffix.lower()
+        mime_type = (
+            f"image/{extension[1:]}" if extension else "image/jpeg"
+        )
+        vision_message["image_url"]["format"] = mime_type
+
+        # Append vision message
+        messages.append(
+            {
+                "role": "user",
+                "content": [
+                    {"type": "text", "text": task},
+                    vision_message,
+                ],
+            }
+        )
+
+        return messages
+
+    def vision_processing(
+        self, task: str, image: str, messages: Optional[list] = None
+    ):
         """
         Process the image for the given task.
+        Handles different image formats and model requirements.
+        """
+        # # # Handle Anthropic models separately
+        # # if "anthropic" in self.model_name.lower() or "claude" in self.model_name.lower():
+        # #     messages = self.anthropic_vision_processing(task, image, messages)
+        # #     return messages
+
+        # # Get base64 encoded image with proper format
+        # image_url = get_image_base64(image)
+
+        # # Prepare vision message
+        # vision_message = {
+        #     "type": "image_url",
+        #     "image_url": {"url": image_url},
+        # }
+
+        # # Add format for specific models
+        # extension = Path(image).suffix.lower()
+        # mime_type = f"image/{extension[1:]}" if extension else "image/jpeg"
+        # vision_message["image_url"]["format"] = mime_type
+
+        # # Append vision message
+        # messages.append(
+        #     {
+        #         "role": "user",
+        #         "content": [
+        #             {"type": "text", "text": task},
+        #             vision_message,
+        #         ],
+        #     }
+        # )
+
+        # return messages
+        if (
+            "anthropic" in self.model_name.lower()
+            or "claude" in self.model_name.lower()
+        ):
+            messages = self.anthropic_vision_processing(
+                task, image, messages
+            )
+            return messages
+        else:
+            messages = self.openai_vision_processing(
+                task, image, messages
+            )
+            return messages
+
+    def audio_processing(self, task: str, audio: str):
         """
-        self.modalities.append("vision")
+        Process the audio for the given task.
 
-        # Append messages
+        Args:
+            task (str): The task to be processed.
+            audio (str): The path or identifier for the audio file.
+        """
+        encoded_string = get_audio_base64(audio)
+
+        # Append audio message
         self.messages.append(
             {
                 "role": "user",
                 "content": [
                     {"type": "text", "text": task},
                     {
-                        "type": "image_url",
-                        "image_url": {
-                            "url": image,
-                            # "detail": "high"
-                            # "format": "image",
+                        "type": "input_audio",
+                        "input_audio": {
+                            "data": encoded_string,
+                            "format": "wav",
                         },
                     },
                 ],
             }
         )
 
-    def handle_modalities(
-        self, task: str, audio: str = None, img: str = None
-    ):
+    def check_if_model_supports_vision(self, img: str = None):
         """
-        Handle the modalities for the given task.
+        Check if the model supports vision.
         """
-        self.messages = []  # Reset messages
-        self.modalities.append("text")
-
-        if audio is not None:
-            self.audio_processing(task=task, audio=audio)
-            self.modalities.append("audio")
-
         if img is not None:
-            self.vision_processing(task=task, image=img)
-            self.modalities.append("vision")
+            out = supports_vision(model=self.model_name)
+
+            if out is False:
+                raise ValueError(
+                    f"Model {self.model_name} does not support vision"
+                )
 
     def run(
         self,
@@ -256,13 +403,7 @@ class LiteLLM:
             Exception: If there is an error in processing the request.
         """
         try:
-            messages = self._prepare_messages(task)
-
-            if audio is not None or img is not None:
-                self.handle_modalities(
-                    task=task, audio=audio, img=img
-                )
-                messages = self.messages
+            messages = self._prepare_messages(task=task, img=img)
 
             # Base completion parameters
             completion_params = {
@@ -298,6 +439,9 @@ class LiteLLM:
                     {"functions": self.functions}
                 )
 
+            if self.base_url is not None:
+                completion_params["base_url"] = self.base_url
+
             # Add modalities if needed
             if self.modalities and len(self.modalities) >= 2:
                 completion_params["modalities"] = self.modalities
@@ -308,12 +452,16 @@ class LiteLLM:
             # Handle tool-based response
             if self.tools_list_dictionary is not None:
                 return self.output_for_tools(response)
+            elif self.return_all is True:
+                return response.model_dump()
             else:
                 # Return standard response content
                 return response.choices[0].message.content
 
         except LiteLLMException as error:
-            logger.error(f"Error in LiteLLM run: {str(error)}")
+            logger.error(
+                f"Error in LiteLLM run: {str(error)} Traceback: {traceback.format_exc()}"
+            )
             if "rate_limit" in str(error).lower():
                 logger.warning(
                     "Rate limit hit, retrying with exponential backoff..."

swarms/utils/output_types.py

@@ -0,0 +1,24 @@
+from typing import Literal
+
+HistoryOutputType = Literal[
+    "list",
+    "dict",
+    "dictionary",
+    "string",
+    "str",
+    "final",
+    "last",
+    "json",
+    "all",
+    "yaml",
+    "xml",
+    # "dict-final",
+    "dict-all-except-first",
+    "str-all-except-first",
+    "basemodel",
+    "dict-final",
+]
+
+OutputType = HistoryOutputType
+
+output_type: HistoryOutputType  # OutputType now includes 'xml'

swarms/utils/vllm_wrapper.py

@@ -133,12 +133,11 @@ class VLLMWrapper:
         Returns:
             List[str]: List of model responses.
         """
-        # Fetch 95% of the available CPU cores
-        num_cores = os.cpu_count()
-        num_workers = int(num_cores * 0.95)
-        with concurrent.futures.ThreadPoolExecutor(
-            max_workers=num_workers
-        ) as executor:
+        # Calculate the worker count based on 95% of available CPU cores
+        num_workers = max(1, int((os.cpu_count() or 1) * 0.95))
+        with concurrent.futures.ThreadPoolExecutor(
+            max_workers=num_workers
+        ) as executor:
             futures = [
                 executor.submit(self.run, task) for task in tasks
             ]

swarms/utils/xml_utils.py

@@ -3,7 +3,22 @@ from typing import Any
 
 
 def dict_to_xml(tag: str, d: dict) -> ET.Element:
-    """Convert a dictionary to an XML Element."""
+    """
+    Convert a dictionary to an XML Element.
+
+    Args:
+        tag (str): The tag name for the root element
+        d (dict): The dictionary to convert to XML
+
+    Returns:
+        ET.Element: An XML Element representing the dictionary structure
+
+    Example:
+        >>> data = {"person": {"name": "John", "age": 30}}
+        >>> elem = dict_to_xml("root", data)
+        >>> ET.tostring(elem, encoding="unicode")
+        '<root><person><name>John</name><age>30</age></person></root>'
+    """
     elem = ET.Element(tag)
     for key, val in d.items():
         child = ET.Element(str(key))
@@ -24,7 +39,27 @@ def dict_to_xml(tag: str, d: dict) -> ET.Element:
 
 
 def to_xml_string(data: Any, root_tag: str = "root") -> str:
-    """Convert a dict or list to an XML string."""
+    """
+    Convert a dict or list to an XML string.
+
+    Args:
+        data (Any): The data to convert to XML. Can be a dictionary, list, or other value
+        root_tag (str, optional): The tag name for the root element. Defaults to "root"
+
+    Returns:
+        str: An XML string representation of the input data
+
+    Example:
+        >>> data = {"person": {"name": "John", "age": 30}}
+        >>> xml_str = to_xml_string(data)
+        >>> print(xml_str)
+        <root><person><name>John</name><age>30</age></person></root>
+
+        >>> data = [1, 2, 3]
+        >>> xml_str = to_xml_string(data)
+        >>> print(xml_str)
+        <root><item>1</item><item>2</item><item>3</item></root>
+    """
     if isinstance(data, dict):
         elem = dict_to_xml(root_tag, data)
     elif isinstance(data, list):

{swarms-7.8.4.dist-info → swarms-7.8.8.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: swarms
-Version: 7.8.4
+Version: 7.8.8
 Summary: Swarms - TGSC
 License: MIT
 Keywords: artificial intelligence,deep learning,optimizers,Prompt Engineering,swarms,agents,llms,transformers,multi-agent,swarms of agents,Enterprise-Grade Agents,Production-Grade Agents,Agents,Multi-Grade-Agents,Swarms,Transformers,LLMs,Prompt Engineering,Agents,Generative Agents,Generative AI,Agent Marketplace,Agent Store,quant,finance,algorithmic trading,portfolio optimization,risk management,financial modeling,machine learning for finance,natural language processing for finance
@@ -1424,15 +1424,13 @@ print(out)
 -------
 
 ## SpreadSheetSwarm
+
 SpreadSheetSwarm manages thousands of agents concurrently for efficient task processing. It supports one-to-many task distribution, scalability, and autosaving results. Initialized with a name, description, agents, and settings, the run method executes tasks and returns a dictionary of agent outputs.
 
 [Learn more:](https://docs.swarms.world/en/latest/swarms/structs/spreadsheet_swarm/)
 
 ```python
-import os
 from swarms import Agent, SpreadSheetSwarm
-from swarm_models import OpenAIChat
-
 # Define custom system prompts for each social media platform
 TWITTER_AGENT_SYS_PROMPT = """
 You are a Twitter marketing expert specializing in real estate. Your task is to create engaging, concise tweets to promote properties, analyze trends to maximize engagement, and use appropriate hashtags and timing to reach potential buyers.
@@ -1459,7 +1457,7 @@ agents = [
     Agent(
         agent_name="Twitter-RealEstate-Agent",
         system_prompt=TWITTER_AGENT_SYS_PROMPT,
-        model_name="gpt-4o",
+        model_name="gpt-4o-mini",
         max_loops=1,
         dynamic_temperature_enabled=True,
         saved_state_path="twitter_realestate_agent.json",
@@ -1469,7 +1467,7 @@ agents = [
     Agent(
         agent_name="Instagram-RealEstate-Agent",
         system_prompt=INSTAGRAM_AGENT_SYS_PROMPT,
-        model_name="gpt-4o",
+        model_name="gpt-4o-mini",
         max_loops=1,
         dynamic_temperature_enabled=True,
         saved_state_path="instagram_realestate_agent.json",
@@ -1479,7 +1477,7 @@ agents = [
     Agent(
         agent_name="Facebook-RealEstate-Agent",
         system_prompt=FACEBOOK_AGENT_SYS_PROMPT,
-        model_name="gpt-4o",
+        model_name="gpt-4o-mini",
         max_loops=1,
         dynamic_temperature_enabled=True,
         saved_state_path="facebook_realestate_agent.json",
@@ -1489,7 +1487,7 @@ agents = [
     Agent(
         agent_name="LinkedIn-RealEstate-Agent",
         system_prompt=LINKEDIN_AGENT_SYS_PROMPT,
-        model_name="gpt-4o",
+        model_name="gpt-4o-mini",
         max_loops=1,
         dynamic_temperature_enabled=True,
         saved_state_path="linkedin_realestate_agent.json",
@@ -1499,7 +1497,7 @@ agents = [
     Agent(
         agent_name="Email-RealEstate-Agent",
         system_prompt=EMAIL_AGENT_SYS_PROMPT,
-        model_name="gpt-4o",
+        model_name="gpt-4o-mini",
         max_loops=1,
         dynamic_temperature_enabled=True,
         saved_state_path="email_realestate_agent.json",
@@ -1963,32 +1961,18 @@ A production-grade multi-agent system enabling sophisticated group conversations
 
 
 ```python
-
-import os
-from dotenv import load_dotenv
-from swarm_models import OpenAIChat
 from swarms import Agent, GroupChat, expertise_based
 
 
 if __name__ == "__main__":
 
-    load_dotenv()
-
-    # Get the OpenAI API key from the environment variable
-    api_key = os.getenv("OPENAI_API_KEY")
-
-    # Create an instance of the OpenAIChat class
-    model = OpenAIChat(
-        openai_api_key=api_key,
-        model_name="gpt-4o-mini",
-        temperature=0.1,
-    )
 
     # Example agents
     agent1 = Agent(
         agent_name="Financial-Analysis-Agent",
         system_prompt="You are a financial analyst specializing in investment strategies.",
-        llm=model,
+        model_name="gpt-4o-mini",
+        temperature=0.1,        
         max_loops=1,
         autosave=False,
         dashboard=False,
@@ -2004,7 +1988,8 @@ if __name__ == "__main__":
     agent2 = Agent(
         agent_name="Tax-Adviser-Agent",
         system_prompt="You are a tax adviser who provides clear and concise guidance on tax-related queries.",
-        llm=model,
+        model_name="gpt-4o-mini",
+        temperature=0.1,           
         max_loops=1,
         autosave=False,
         dashboard=False,
@@ -2029,7 +2014,8 @@ if __name__ == "__main__":
     history = chat.run(
         "How to optimize tax strategy for investments?"
     )
-    print(history.model_dump_json(indent=2))
+    print(history)
+
 
 ```
 
@@ -2087,35 +2073,18 @@ if __name__ == "__main__":
 ----------
 
 ## Onboarding Session
+
 Get onboarded now with the creator and lead maintainer of Swarms, Kye Gomez, who will show you how to get started with the installation, usage examples, and starting to build your custom use case! [CLICK HERE](https://cal.com/swarms/swarms-onboarding-session)
 
 
 ---
 
 ## Documentation
+
 Documentation is located here at: [docs.swarms.world](https://docs.swarms.world)
 
 -----
 
-## Folder Structure
-The swarms package has been meticulously crafted for extreme usability and understanding,the swarms package is split up into various modules such as `swarms.agents` that holds pre-built agents, `swarms.structs` that holds a vast array of structures like `Agent` and multi agent structures. The package is split into various modules, with the most important being `structs`, `tools`, and `agents`.
-
-```sh
-├── __init__.py
-├── agents/
-├── artifacts/
-├── client/
-├── cli/
-├── prompts/
-├── schemas/
-├── structs/
-├── telemetry/
-├── tools/
-└── utils/
-```
-
-----
-
 ## 🫶 Contributions:
 
 The easiest way to contribute is to pick any issue with the `good first issue` tag 💪. Read the Contributing guidelines [here](/CONTRIBUTING.md). Bug Report? [File here](https://github.com/swarms/gateway/issues) | Feature Request? [File here](https://github.com/swarms/gateway/issues)
@@ -2125,17 +2094,24 @@ Swarms is an open-source project, and contributions are VERY welcome. If you wan
 ----
 
 
-## Community
+### Connect With Us
+
+| Platform | Link | Description |
+|----------|------|-------------|
+| 📚 Documentation | [docs.swarms.world](https://docs.swarms.world) | Official documentation and guides |
+| 📝 Blog | [Medium](https://medium.com/@kyeg) | Latest updates and technical articles |
+| 💬 Discord | [Join Discord](https://discord.gg/jM3Z6M9uMq) | Live chat and community support |
+| 🐦 Twitter | [@kyegomez](https://twitter.com/kyegomez) | Latest news and announcements |
+| 👥 LinkedIn | [The Swarm Corporation](https://www.linkedin.com/company/the-swarm-corporation) | Professional network and updates |
+| 📺 YouTube | [Swarms Channel](https://www.youtube.com/channel/UC9yXyitkbU_WSy7bd_41SqQ) | Tutorials and demos |
+| 🎫 Events | [Sign up here](https://lu.ma/5p2jnc2v) | Join our community events |
+
+
+
+## Citation
 
-Join our growing community around the world, for real-time support, ideas, and discussions on Swarms 😊 
+If you use **swarms** in your research, please cite the project by referencing the metadata in [CITATION.cff](./CITATION.cff).
 
-- View our official [Documents](https://docs.swarms.world)
-- View our official [Blog](https://medium.com/@kyeg)
-- Chat live with us on [Discord](https://discord.gg/jM3Z6M9uMq)
-- Follow us on [Twitter](https://twitter.com/kyegomez)
-- Connect with us on [LinkedIn](https://www.linkedin.com/company/the-swarm-corporation)
-- Visit us on [YouTube](https://www.youtube.com/channel/UC9yXyitkbU_WSy7bd_41SqQ)
-- Sign up for our events [Sign up here](https://lu.ma/5p2jnc2v)
 
 # License