llm-dialog-manager 0.3.5__tar.gz → 0.4.2__tar.gz

{llm_dialog_manager-0.3.5 → llm_dialog_manager-0.4.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: llm_dialog_manager
-Version: 0.3.5
+Version: 0.4.2
 Summary: A Python package for managing LLM chat conversation history
 Author-email: xihajun <work@2333.fun>
 License: MIT
@@ -64,6 +64,7 @@ A Python package for managing AI chat conversation history with support for mult
 - Memory management options
 - Conversation search and indexing
 - Rich conversation display options
+- Vision & Json Output enabled [20240111]
 
 ## Installation
 

{llm_dialog_manager-0.3.5 → llm_dialog_manager-0.4.2}/README.md

@@ -17,6 +17,7 @@ A Python package for managing AI chat conversation history with support for mult
 - Memory management options
 - Conversation search and indexing
 - Rich conversation display options
+- Vision & Json Output enabled [20240111]
 
 ## Installation
 

{llm_dialog_manager-0.3.5 → llm_dialog_manager-0.4.2}/llm_dialog_manager/__init__.py

@@ -1,4 +1,4 @@
 from .chat_history import ChatHistory
 from .agent import Agent
 
-__version__ = "0.3.5" 
+__version__ = "0.4.2" 

{llm_dialog_manager-0.3.5 → llm_dialog_manager-0.4.2}/llm_dialog_manager/agent.py

@@ -38,15 +38,46 @@ def load_env_vars():
 
 load_env_vars()
 
+def format_messages_for_gemini(messages):
+    """
+    将标准化的消息格式转化为 Gemini 格式。
+    system 消息应该通过 GenerativeModel 的 system_instruction 参数传入,
+    不在这个函数处理。
+    """
+    gemini_messages = []
+
+    for msg in messages:
+        role = msg["role"]
+        content = msg["content"]
+        
+        # 跳过 system 消息,因为它会通过 system_instruction 设置
+        if role == "system":
+            continue
+            
+        # 处理 user/assistant 消息
+        # 如果 content 是单一对象,转换为列表
+        if not isinstance(content, list):
+            content = [content]
+        
+        gemini_messages.append({
+            "role": role,
+            "parts": content  # content 可以包含文本和 FileMedia
+        })
+
+    return gemini_messages
+
 def completion(model: str, messages: List[Dict[str, Union[str, List[Union[str, Image.Image, Dict]]]]], max_tokens: int = 1000, 
-              temperature: float = 0.5, api_key: Optional[str] = None, 
+              temperature: float = 0.5, top_p: float = 1.0, top_k: int = 40, api_key: Optional[str] = None, 
               base_url: Optional[str] = None, json_format: bool = False) -> str:
     """
     Generate a completion using the specified model and messages.
     """
     try:
         service = ""
-        if "claude" in model:
+        if "openai" in model:
+            service = "openai"
+            model
+        elif "claude" in model:
             service = "anthropic"
         elif "gemini" in model:
             service = "gemini"
@@ -64,7 +95,7 @@ def completion(model: str, messages: List[Dict[str, Union[str, List[Union[str, I
 
         def format_messages_for_api(model, messages):
             """Convert ChatHistory messages to the format required by the specific API."""
-            if "claude" in model:
+            if "claude" in model and "openai" not in model:
                 formatted = []
                 system_msg = ""
                 if messages and messages[0]["role"] == "system":
@@ -113,7 +144,7 @@ def completion(model: str, messages: List[Dict[str, Union[str, List[Union[str, I
                         formatted.append({"role": msg["role"], "content": combined_content})
                 return system_msg, formatted
 
-            elif "gemini" in model or "gpt" in model or "grok" in model:
+            elif ("gemini" in model or "gpt" in model or "grok" in model) and "openai" not in model:
                 formatted = []
                 for msg in messages:
                     content = msg["content"]
@@ -163,7 +194,7 @@ def completion(model: str, messages: List[Dict[str, Union[str, List[Union[str, I
 
         system_msg, formatted_messages = format_messages_for_api(model, messages.copy())
 
-        if "claude" in model:
+        if "claude" in model and "openai" not in model:
             # Check for Vertex configuration
             vertex_project_id = os.getenv('VERTEX_PROJECT_ID')
             vertex_region = os.getenv('VERTEX_REGION')
@@ -204,7 +235,7 @@ def completion(model: str, messages: List[Dict[str, Union[str, List[Union[str, I
 
             return response.completion
 
-        elif "gemini" in model:
+        elif "gemini" in model and "openai" not in model:
             try:
                 # First try OpenAI-style API
                 client = openai.OpenAI(
@@ -216,6 +247,9 @@ def completion(model: str, messages: List[Dict[str, Union[str, List[Union[str, I
 
                 response = client.chat.completions.create(
                     model=model,
+                    max_tokens=max_tokens,
+                    top_p=top_p,
+                    top_k=top_k,
                     messages=formatted_messages,
                     temperature=temperature,
                     response_format=response_format  # Added response_format
@@ -226,36 +260,34 @@ def completion(model: str, messages: List[Dict[str, Union[str, List[Union[str, I
                 # If OpenAI-style API fails, fall back to Google's genai library
                 logger.info("Falling back to Google's genai library")
                 genai.configure(api_key=api_key)
-
-                # Convert messages to Gemini format
-                gemini_messages = []
+                system_instruction = ""
                 for msg in messages:
                     if msg["role"] == "system":
-                        # Prepend system message to first user message if exists
-                        if gemini_messages:
-                            first_msg = gemini_messages[0]
-                            if "parts" in first_msg and len(first_msg["parts"]) > 0:
-                                first_msg["parts"][0] = f"{msg['content']}\n\n{first_msg['parts'][0]}"
-                    else:
-                        gemini_messages.append({"role": msg["role"], "parts": msg["content"]})
-
-                # Set response_mime_type based on json_format
+                        system_instruction = msg["content"]
+                        break
+                
+                # 将其他消息转换为 gemini 格式
+                gemini_messages = format_messages_for_gemini(messages)
                 mime_type = "application/json" if json_format else "text/plain"
+                generation_config = genai.types.GenerationConfig(
+                    temperature=temperature,
+                    top_p=top_p,
+                    top_k=top_k,
+                    max_output_tokens=max_tokens,
+                    response_mime_type=mime_type
+                )
 
-                # Create Gemini model and generate response
-                model_instance = genai.GenerativeModel(model_name=model)
-                response = model_instance.generate_content(
-                    gemini_messages,
-                    generation_config=genai.types.GenerationConfig(
-                        temperature=temperature,
-                        response_mime_type=mime_type,  # Modified based on json_format
-                        max_output_tokens=max_tokens
-                    )
+                model_instance = genai.GenerativeModel(
+                    model_name=model,
+                    system_instruction=system_instruction,  # system 消息通过这里传入
+                    generation_config=generation_config
                 )
 
+                response = model_instance.generate_content(gemini_messages, generation_config=generation_config)
+
                 return response.text
 
-        elif "grok" in model:
+        elif "grok" in model and "openai" not in model:
             # Randomly choose between OpenAI and Anthropic SDK
             use_anthropic = random.choice([True, False])
 
@@ -297,6 +329,8 @@ def completion(model: str, messages: List[Dict[str, Union[str, List[Union[str, I
                 return response.choices[0].message.content
 
         else:  # OpenAI models
+            if model.endswith("-openai"):
+                model = model[:-7]  # Remove last 7 characters ("-openai")
             client = openai.OpenAI(api_key=api_key, base_url=base_url)
             # Set response_format based on json_format
             response_format = {"type": "json_object"} if json_format else {"type": "plain_text"}
@@ -395,7 +429,7 @@ class Agent:
             # Start a new user message with the image
             self.history.add_message([image_block], "user")
 
-    def generate_response(self, max_tokens=3585, temperature=0.7, json_format: bool = False) -> str:
+    def generate_response(self, max_tokens=3585, temperature=0.7, top_p=1.0, top_k=40, json_format: bool = False) -> str:
         """Generate a response from the agent.
 
         Args:
@@ -410,12 +444,14 @@ class Agent:
             raise ValueError("No messages in history to generate response from")
         
         messages = self.history.messages
-
+        print(self.model_name)
         response_text = completion(
             model=self.model_name,
             messages=messages,
             max_tokens=max_tokens,
             temperature=temperature,
+            top_p=top_p,
+            top_k=top_k,
             api_key=self.api_key,
             json_format=json_format  # Pass json_format to completion
         )
@@ -486,13 +522,13 @@ class Agent:
 if __name__ == "__main__":
     # Example Usage
     # Create an Agent instance (Gemini model)
-    agent = Agent("gemini-1.5-flash", "you are an assistant", memory_enabled=True)
+    agent = Agent("gemini-1.5-flash", "you are Jack101", memory_enabled=True)
     
     # Add an image
     agent.add_image(image_path="/Users/junfan/Projects/Personal/oneapi/dialog_manager/example.png")
     
     # Add a user message
-    agent.add_message("user", "What's in this image?")
+    agent.add_message("user", "Who are you? What's in this image?")
     
     # Generate response with JSON format enabled
     try:

{llm_dialog_manager-0.3.5 → llm_dialog_manager-0.4.2}/llm_dialog_manager.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: llm_dialog_manager
-Version: 0.3.5
+Version: 0.4.2
 Summary: A Python package for managing LLM chat conversation history
 Author-email: xihajun <work@2333.fun>
 License: MIT
@@ -64,6 +64,7 @@ A Python package for managing AI chat conversation history with support for mult
 - Memory management options
 - Conversation search and indexing
 - Rich conversation display options
+- Vision & Json Output enabled [20240111]
 
 ## Installation
 

llm_dialog_manager-0.4.2/pyproject.toml

@@ -0,0 +1,32 @@
+[build-system]
+requires = [ "setuptools>=61.0", "wheel",]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "llm_dialog_manager"
+version = "0.4.2"
+description = "A Python package for managing LLM chat conversation history"
+readme = "README.md"
+classifiers = [ "Development Status :: 3 - Alpha", "Intended Audience :: Developers", "License :: OSI Approved :: MIT License", "Operating System :: OS Independent", "Programming Language :: Python :: 3.8", "Programming Language :: Python :: 3.9", "Programming Language :: Python :: 3.10", "Topic :: Software Development :: Libraries :: Python Modules", "Topic :: Scientific/Engineering :: Artificial Intelligence",]
+requires-python = ">=3.7"
+dependencies = [ "openai>=1.54.2", "anthropic>=0.39.0", "google-generativeai>=0.1.0", "python-dotenv>=1.0.0", "typing-extensions>=4.0.0", "uuid>=1.30",]
+[[project.authors]]
+name = "xihajun"
+email = "work@2333.fun"
+
+[project.license]
+text = "MIT"
+
+[project.optional-dependencies]
+dev = [ "pytest>=8.0.0", "pytest-asyncio>=0.21.1", "pytest-cov>=4.1.0", "black>=23.9.1", "isort>=5.12.0",]
+test = [ "pytest>=6.0", "pytest-asyncio>=0.14.0", "pytest-cov>=2.0",]
+lint = [ "black>=22.0", "isort>=5.0",]
+all = [ "pytest>=8.0.0", "pytest-asyncio>=0.21.1", "pytest-cov>=4.1.0", "black>=23.9.1", "isort>=5.12.0",]
+
+[project.urls]
+"Bug Tracker" = "https://github.com/xihajun/llm_dialog_manager/issues"
+Documentation = "https://github.com/xihajun/llm_dialog_manager#readme"
+"Source Code" = "https://github.com/xihajun/llm_dialog_manager"
+
+[tool.setuptools]
+packages = [ "llm_dialog_manager",]

llm_dialog_manager-0.3.5/pyproject.toml

@@ -1,64 +0,0 @@
-[build-system]
-requires = ["setuptools>=61.0", "wheel"]
-build-backend = "setuptools.build_meta"
-
-[project]
-name = "llm_dialog_manager"
-version = "0.3.5"
-description = "A Python package for managing LLM chat conversation history"
-readme = "README.md"
-authors = [{ name = "xihajun", email = "work@2333.fun" }]
-license = { text = "MIT" }
-classifiers = [
-    "Development Status :: 3 - Alpha",
-    "Intended Audience :: Developers",
-    "License :: OSI Approved :: MIT License",
-    "Operating System :: OS Independent",
-    "Programming Language :: Python :: 3.8",
-    "Programming Language :: Python :: 3.9",
-    "Programming Language :: Python :: 3.10",
-    "Topic :: Software Development :: Libraries :: Python Modules",
-    "Topic :: Scientific/Engineering :: Artificial Intelligence",
-]
-requires-python = ">=3.7"
-dependencies = [
-    "openai>=1.54.2",
-    "anthropic>=0.39.0",
-    "google-generativeai>=0.1.0",
-    "python-dotenv>=1.0.0",
-    "typing-extensions>=4.0.0",
-    "uuid>=1.30",
-]
-
-[project.optional-dependencies]
-dev = [
-    "pytest>=8.0.0",
-    "pytest-asyncio>=0.21.1",
-    "pytest-cov>=4.1.0",
-    "black>=23.9.1",
-    "isort>=5.12.0",
-]
-test = [
-    "pytest>=6.0",
-    "pytest-asyncio>=0.14.0",
-    "pytest-cov>=2.0",
-]
-lint = [
-    "black>=22.0",
-    "isort>=5.0",
-]
-all = [
-    "pytest>=8.0.0",
-    "pytest-asyncio>=0.21.1",
-    "pytest-cov>=4.1.0",
-    "black>=23.9.1",
-    "isort>=5.12.0",
-]
-
-[project.urls]
-"Bug Tracker" = "https://github.com/xihajun/llm_dialog_manager/issues"
-"Documentation" = "https://github.com/xihajun/llm_dialog_manager#readme"
-"Source Code" = "https://github.com/xihajun/llm_dialog_manager"
-
-[tool.setuptools]
-packages = ["llm_dialog_manager"]