Merge pull request #436 from swarmauri/master

🚢 Shipping v0.4.4

.github/workflows/publish_dev.yml

@@ -14,7 +14,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        python-version: ["3.10"]
+        python-version: ["3.11"]
 
     env:
       # Model Provider Keys
@@ -29,6 +29,7 @@ jobs:
       DEEPSEEK_API_KEY: ${{ secrets.DEEPSEEK_API_KEY }}
       SHUTTLEAI_API_KEY: ${{ secrets.SHUTTLEAI_API_KEY }}
       DEEPINFRA_API_KEY: ${{ secrets.DEEPINFRA_API_KEY }}
+      OPENROUTER_API_KEY: ${{ secrets.OPENROUTER_API_KEY }}
 
       # Database Keys
       QDRANT_URL_KEY: ${{ secrets.QDRANT_URL_KEY }}

.github/workflows/publish_stable.yml

@@ -29,6 +29,7 @@ jobs:
       DEEPSEEK_API_KEY: ${{ secrets.DEEPSEEK_API_KEY }}
       SHUTTLEAI_API_KEY: ${{ secrets.SHUTTLEAI_API_KEY }}
       DEEPINFRA_API_KEY: ${{ secrets.DEEPINFRA_API_KEY }}
+      OPENROUTER_API_KEY: ${{ secrets.OPENROUTER_API_KEY }}
 
       # Database Keys
       QDRANT_URL_KEY: ${{ secrets.QDRANT_URL_KEY }}

.github/workflows/staging.yml

@@ -28,6 +28,7 @@ jobs:
       DEEPSEEK_API_KEY: ${{ secrets.DEEPSEEK_API_KEY }}
       SHUTTLEAI_API_KEY: ${{ secrets.SHUTTLEAI_API_KEY }}
       DEEPINFRA_API_KEY: ${{ secrets.DEEPINFRA_API_KEY }}
+      OPENROUTER_API_KEY: ${{ secrets.OPENROUTER_API_KEY }}
 
       # Database Keys
       QDRANT_URL_KEY: ${{ secrets.QDRANT_URL_KEY }}

setup.py

@@ -33,7 +33,7 @@
             'scikit-learn',
             'gensim',
             'textblob',
-            'spacy==3.7.4',
+            'spacy',
             'pygments',
             'gradio',
             'websockets',

swarmauri/__init__.py

@@ -1,4 +1,4 @@
-__version__ = "0.4.3"
+__version__ = "0.4.4"
 __long_desc__ = """
 # Swarmauri SDK
 

swarmauri/standard/agents/concrete/ToolAgent.py

@@ -11,12 +11,12 @@
 from swarmauri.standard.messages.concrete import HumanMessage, AgentMessage, FunctionMessage
 
 from swarmauri.core.typing import SubclassUnion
-from swarmauri.standard.toolkits.concrete.Toolkit import Toolkit
+from swarmauri.standard.toolkits.base.ToolkitBase import ToolkitBase
 from swarmauri.standard.conversations.base.ConversationBase import ConversationBase
 
 class ToolAgent(AgentToolMixin, AgentConversationMixin, AgentBase):
     llm: SubclassUnion[LLMBase]
-    toolkit: SubclassUnion[Toolkit]
+    toolkit: SubclassUnion[ToolkitBase]
     conversation: SubclassUnion[ConversationBase] # 🚧  Placeholder
     model_config = ConfigDict(extra='forbid', arbitrary_types_allowed=True)
     type: Literal['ToolAgent'] = 'ToolAgent'

swarmauri/standard/llms/concrete/DeepSeekModel.py

@@ -9,10 +9,13 @@
 
 
 class DeepSeekModel(LLMBase):
+    """
+    Provider resources: https://platform.deepseek.com/api-docs/quick_start/pricing
+    """
+
     api_key: str
     allowed_models: List[str] = [
-        "deepseek-chat",
-        "deepseek-coder",
+        "deepseek-chat"
     ]
     name: str = "deepseek-chat"
     type: Literal["DeepSeekModel"] = "DeepSeekModel"

swarmauri/standard/llms/concrete/GeminiProModel.py

@@ -9,81 +9,87 @@
 
 
 class GeminiProModel(LLMBase):
+    """
+    Provider resources: https://deepmind.google/technologies/gemini/pro/
+    """
+
     api_key: str
-    allowed_models: List[str] = ['gemini-1.5-pro-latest']
+    allowed_models: List[str] = ["gemini-1.5-pro-latest"]
     name: str = "gemini-1.5-pro-latest"
-    type: Literal['GeminiProModel'] = 'GeminiProModel'
-
-    def _format_messages(self, messages: List[SubclassUnion[MessageBase]]) -> List[Dict[str, str]]:
+    type: Literal["GeminiProModel"] = "GeminiProModel"
+
+    def _format_messages(
+        self, messages: List[SubclassUnion[MessageBase]]
+    ) -> List[Dict[str, str]]:
         # Remove system instruction from messages
-        message_properties = ['content', 'role']
-        sanitized_messages = [message.model_dump(include=message_properties) for message in messages 
-            if message.role != 'system']
+        message_properties = ["content", "role"]
+        sanitized_messages = [
+            message.model_dump(include=message_properties)
+            for message in messages
+            if message.role != "system"
+        ]
 
         for message in sanitized_messages:
-            if message['role'] == 'assistant':
-                message['role'] = 'model'
+            if message["role"] == "assistant":
+                message["role"] = "model"
 
             # update content naming
-            message['parts'] = message.pop('content')
+            message["parts"] = message.pop("content")
 
         return sanitized_messages
 
     def _get_system_context(self, messages: List[SubclassUnion[MessageBase]]) -> str:
         system_context = None
         for message in messages:
-            if message.role == 'system':
+            if message.role == "system":
                 system_context = message.content
         return system_context
-
-    def predict(self, 
-        conversation, 
-        temperature=0.7, 
-        max_tokens=256):
+
+    def predict(self, conversation, temperature=0.7, max_tokens=256):
         genai.configure(api_key=self.api_key)
         generation_config = {
             "temperature": temperature,
             "top_p": 0.95,
             "top_k": 0,
             "max_output_tokens": max_tokens,
-            }
+        }
 
         safety_settings = [
-          {
-            "category": "HARM_CATEGORY_HARASSMENT",
-            "threshold": "BLOCK_MEDIUM_AND_ABOVE"
-          },
-          {
-            "category": "HARM_CATEGORY_HATE_SPEECH",
-            "threshold": "BLOCK_MEDIUM_AND_ABOVE"
-          },
-          {
-            "category": "HARM_CATEGORY_SEXUALLY_EXPLICIT",
-            "threshold": "BLOCK_MEDIUM_AND_ABOVE"
-          },
-          {
-            "category": "HARM_CATEGORY_DANGEROUS_CONTENT",
-            "threshold": "BLOCK_MEDIUM_AND_ABOVE"
-          },
+            {
+                "category": "HARM_CATEGORY_HARASSMENT",
+                "threshold": "BLOCK_MEDIUM_AND_ABOVE",
+            },
+            {
+                "category": "HARM_CATEGORY_HATE_SPEECH",
+                "threshold": "BLOCK_MEDIUM_AND_ABOVE",
+            },
+            {
+                "category": "HARM_CATEGORY_SEXUALLY_EXPLICIT",
+                "threshold": "BLOCK_MEDIUM_AND_ABOVE",
+            },
+            {
+                "category": "HARM_CATEGORY_DANGEROUS_CONTENT",
+                "threshold": "BLOCK_MEDIUM_AND_ABOVE",
+            },
         ]
 
-
         system_context = self._get_system_context(conversation.history)
         formatted_messages = self._format_messages(conversation.history)
 
-
         next_message = formatted_messages.pop()
 
-        client = genai.GenerativeModel(model_name=self.name,
+        client = genai.GenerativeModel(
+            model_name=self.name,
             safety_settings=safety_settings,
             generation_config=generation_config,
-            system_instruction=system_context)
+            system_instruction=system_context,
+        )
 
         convo = client.start_chat(
             history=formatted_messages,
-            )
+        )
 
-        convo.send_message(next_message['parts'])
+        convo.send_message(next_message["parts"])
 
         message_content = convo.last.text
         conversation.add_message(AgentMessage(content=message_content))

swarmauri/standard/llms/concrete/GroqModel.py

@@ -7,33 +7,49 @@
 from swarmauri.standard.messages.concrete.AgentMessage import AgentMessage
 from swarmauri.standard.llms.base.LLMBase import LLMBase
 
+
 class GroqModel(LLMBase):
+    """Provider resources: https://console.groq.com/docs/models"""
+
     api_key: str
-    allowed_models: List[str] = ['llama3-8b-8192', 
-    'llama3-70b-8192', 
-    'mixtral-8x7b-32768', 
-    'gemma-7b-it']
+    allowed_models: List[str] = [
+        "llama3-8b-8192",
+        "llama3-70b-8192",
+        "mixtral-8x7b-32768",
+        "gemma-7b-it",
+        "gemma-2-9b-it",
+        "llama-3.1-70b-versatile",
+        "llama-3.1-8b-instant",
+        "llava-v1.5-7b-4096-preview",
+    ]
     name: str = "gemma-7b-it"
-    type: Literal['GroqModel'] = 'GroqModel'
+    type: Literal["GroqModel"] = "GroqModel"
 
-    def _format_messages(self, messages: List[SubclassUnion[MessageBase]]) -> List[Dict[str, str]]:
-        message_properties = ['content', 'role', 'name']
-        formatted_messages = [message.model_dump(include=message_properties, exclude_none=True) for message in messages]
+    def _format_messages(
+        self, messages: List[SubclassUnion[MessageBase]]
+    ) -> List[Dict[str, str]]:
+        message_properties = ["content", "role", "name"]
+        formatted_messages = [
+            message.model_dump(include=message_properties, exclude_none=True)
+            for message in messages
+        ]
         return formatted_messages
 
-    def predict(self, 
-        conversation, 
-        temperature: float = 0.7, 
-        max_tokens: int = 256, 
-        top_p: float = 1.0, 
-        enable_json: bool = False, 
-        stop: Optional[List[str]] = None) -> str:
+    def predict(
+        self,
+        conversation,
+        temperature: float = 0.7,
+        max_tokens: int = 256,
+        top_p: float = 1.0,
+        enable_json: bool = False,
+        stop: Optional[List[str]] = None,
+    ) -> str:
 
         formatted_messages = self._format_messages(conversation.history)
 
         client = Groq(api_key=self.api_key)
         stop = stop or []
-        
+
         response_format = {"type": "json_object"} if enable_json else None
         response = client.chat.completions.create(
             model=self.name,
@@ -44,10 +60,10 @@ def predict(self,
             top_p=top_p,
             frequency_penalty=0,
             presence_penalty=0,
-            stop=stop
+            stop=stop,
         )
-        
+
         result = json.loads(response.json())
-        message_content = result['choices'][0]['message']['content']
+        message_content = result["choices"][0]["message"]["content"]
         conversation.add_message(AgentMessage(content=message_content))
-        return conversation
+        return conversation

swarmauri/standard/llms/concrete/MistralModel.py

@@ -1,66 +1,76 @@
 import json
 from typing import List, Literal, Dict
-from mistralai import Mistral 
+from mistralai import Mistral
 from swarmauri.core.typing import SubclassUnion
 
 from swarmauri.standard.messages.base.MessageBase import MessageBase
 from swarmauri.standard.messages.concrete.AgentMessage import AgentMessage
 from swarmauri.standard.llms.base.LLMBase import LLMBase
 
+
 class MistralModel(LLMBase):
+    """Provider resources: https://docs.mistral.ai/getting-started/models/"""
+
     api_key: str
-    allowed_models: List[str] = ['open-mistral-7b', 
-    'open-mixtral-8x7b', 
-    'open-mixtral-8x22b', 
-    'mistral-small-latest',
-    'mistral-medium-latest',
-    'mistral-large-latest',
-    'codestral', 
-    'open-mistral-nemo', 
-    'codestral-latest', 
-    'open-codestral-mamba', 
+    allowed_models: List[str] = [
+        "open-mistral-7b",
+        "open-mixtral-8x7b",
+        "open-mixtral-8x22b",
+        "mistral-small-latest",
+        "mistral-medium-latest",
+        "mistral-large-latest",
+        "open-mistral-nemo",
+        "codestral-latest",
+        "open-codestral-mamba",
     ]
     name: str = "open-mixtral-8x7b"
-    type: Literal['MistralModel'] = 'MistralModel'
+    type: Literal["MistralModel"] = "MistralModel"
 
-    def _format_messages(self, messages: List[SubclassUnion[MessageBase]]) -> List[Dict[str, str]]:
-        message_properties = ['content', 'role']
-        formatted_messages = [message.model_dump(include=message_properties, exclude_none=True) for message in messages]
+    def _format_messages(
+        self, messages: List[SubclassUnion[MessageBase]]
+    ) -> List[Dict[str, str]]:
+        message_properties = ["content", "role"]
+        formatted_messages = [
+            message.model_dump(include=message_properties, exclude_none=True)
+            for message in messages
+        ]
         return formatted_messages
 
-    def predict(self, 
-        conversation, 
-        temperature: int = 0.7, 
-        max_tokens: int = 256, 
+    def predict(
+        self,
+        conversation,
+        temperature: int = 0.7,
+        max_tokens: int = 256,
         top_p: int = 1,
-        enable_json: bool=False, 
-        safe_prompt: bool=False):
-
+        enable_json: bool = False,
+        safe_prompt: bool = False,
+    ):
+
         formatted_messages = self._format_messages(conversation.history)
 
-        client =  Mistral(api_key=self.api_key)        
+        client = Mistral(api_key=self.api_key)
         if enable_json:
             response = client.chat.complete(
                 model=self.name,
                 messages=formatted_messages,
                 temperature=temperature,
-                response_format={ "type": "json_object" },
+                response_format={"type": "json_object"},
                 max_tokens=max_tokens,
                 top_p=top_p,
-                safe_prompt=safe_prompt
+                safe_prompt=safe_prompt,
             )
         else:
             response = client.chat.complete(
                 model=self.name,
                 messages=formatted_messages,
                 temperature=temperature,
                 max_tokens=max_tokens,
-                top_p=top_p,                
-                safe_prompt=safe_prompt
+                top_p=top_p,
+                safe_prompt=safe_prompt,
             )
-        
+
         result = json.loads(response.json())
-        message_content = result['choices'][0]['message']['content']
+        message_content = result["choices"][0]["message"]["content"]
         conversation.add_message(AgentMessage(content=message_content))
 
-        return conversation
+        return conversation