quantalogic 0.2.8__tar.gz → 0.2.10__tar.gz

{quantalogic-0.2.8 → quantalogic-0.2.10}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: quantalogic
-Version: 0.2.8
+Version: 0.2.10
 Summary: QuantaLogic ReAct Agents
 Author: Raphaël MANSUY
 Author-email: raphael.mansuy@gmail.com
@@ -123,8 +123,10 @@ Usage: quantalogic [OPTIONS] COMMAND [ARGS]...
 
 Options:
   --version                       Show version information.
-  --model-name TEXT               Specify the model to use (litellm format,
+  --model-name TEXT               Specify the text model to use (litellm format,
                                   e.g. "openrouter/deepseek-chat").
+  --vision-model-name TEXT        Specify the vision model to use (litellm format,
+                                  e.g. "openrouter/A/gpt-4o-mini").
   --log [info|debug|warning]      Set logging level (info/debug/warning).
   --verbose                       Enable verbose output.
   --mode [code|basic|interpreter|full|code-basic]
@@ -385,7 +387,7 @@ By integrating these tools into its architecture, QuantaLogic allows agents to p
 | Script Execution       | Python Tool, Node.js Tool, Elixir Tool                                                            |
 | File Operations        | Read File Tool, Write File Tool, Edit Whole Content Tool, Replace In File Tool                   |
 | Code Analysis          | Search Definition Names Tool, Ripgrep Tool                                                        |
-| Content Generation      | LLM Tool                                                                                          |
+| Content Generation      | LLM Tool, LLMVisionTool                                                                           |
 | Utility and Management  | Download HTTP File Tool, List Directory Tool, Markitdown Tool, Unified Diff Tool                 |
 
 ---
@@ -681,7 +683,30 @@ print("Ripgrep Results:", output)
 
 ---
 
-### 14. LLM Tool
+#### 14. LLMVisionTool
+
+The **LLMVisionTool** enables processing of visual inputs using vision-language models.
+
+##### Parameters
+
+| Parameter      | Type    | Description                                                            | Example                                    |
+|----------------|---------|------------------------------------------------------------------------|--------------------------------------------|
+| `image_path`   | string  | Path to the image file to process                                      | `./path/to/image.png`                      |
+| `prompt`       | string  | The question or instruction for the vision model                       | `Describe the contents of this image`      |
+| `temperature`  | float   | Sampling temperature between 0.0 and 1.0                              | `0.7`                                      |
+
+##### Example Usage
+```python
+vision_tool = LLMVisionTool()
+response = vision_tool.execute(
+    image_path="./path/to/image.png",
+    prompt="Describe the contents of this image",
+    temperature=0.7
+)
+print("Vision Model Response:", response)
+```
+
+#### 15. LLM Tool
 
 The **LLM Tool** generates answers using a specified language model.
 
@@ -706,7 +731,7 @@ print("LLM Response:", response)
 
 ---
 
-### 15. Download HTTP File Tool
+### 16. Download HTTP File Tool
 
 The **Download HTTP File Tool** downloads a file from a specified HTTP URL.
 
@@ -726,7 +751,7 @@ print(result)
 
 ---
 
-### 16. List Directory Tool
+### 17. List Directory Tool
 
 The **List Directory Tool** lists files in a specified directory.
 
@@ -745,7 +770,7 @@ print("Directory Files:", result)
 
 ---
 
-### 17. Markitdown Tool
+### 18. Markitdown Tool
 
 The **Markitdown Tool** processes markdown files, possibly for conversion or rendering.
 
@@ -762,26 +787,6 @@ result = markitdown_tool.execute(markdown_path="./path/to/file.md")
 print("Processed Markdown Output:", result)
 ```
 
----
-
-### 18. Unified Diff Tool
-
-The **Unified Diff Tool** generates a unified diff between two texts or files.
-
-#### Parameters
-
-| Parameter    | Type   | Description                                    | Example                |
-|--------------|--------|------------------------------------------------|------------------------|
-| `original`   | string | The original content or file path.            | `old_text.txt`         |
-| `updated`    | string | The updated content or file path.             | `new_text.txt`         |
-
-#### Example Usage
-```python
-diff_tool = UnifiedDiffTool()
-result = diff_tool.execute(original="old_text.txt", updated="new_text.txt")
-print("Unified Diff Output:", result)
-```
-
 
 #### Creating Custom Tools
 

{quantalogic-0.2.8 → quantalogic-0.2.10}/README.md

@@ -86,8 +86,10 @@ Usage: quantalogic [OPTIONS] COMMAND [ARGS]...
 
 Options:
   --version                       Show version information.
-  --model-name TEXT               Specify the model to use (litellm format,
+  --model-name TEXT               Specify the text model to use (litellm format,
                                   e.g. "openrouter/deepseek-chat").
+  --vision-model-name TEXT        Specify the vision model to use (litellm format,
+                                  e.g. "openrouter/A/gpt-4o-mini").
   --log [info|debug|warning]      Set logging level (info/debug/warning).
   --verbose                       Enable verbose output.
   --mode [code|basic|interpreter|full|code-basic]
@@ -348,7 +350,7 @@ By integrating these tools into its architecture, QuantaLogic allows agents to p
 | Script Execution       | Python Tool, Node.js Tool, Elixir Tool                                                            |
 | File Operations        | Read File Tool, Write File Tool, Edit Whole Content Tool, Replace In File Tool                   |
 | Code Analysis          | Search Definition Names Tool, Ripgrep Tool                                                        |
-| Content Generation      | LLM Tool                                                                                          |
+| Content Generation      | LLM Tool, LLMVisionTool                                                                           |
 | Utility and Management  | Download HTTP File Tool, List Directory Tool, Markitdown Tool, Unified Diff Tool                 |
 
 ---
@@ -644,7 +646,30 @@ print("Ripgrep Results:", output)
 
 ---
 
-### 14. LLM Tool
+#### 14. LLMVisionTool
+
+The **LLMVisionTool** enables processing of visual inputs using vision-language models.
+
+##### Parameters
+
+| Parameter      | Type    | Description                                                            | Example                                    |
+|----------------|---------|------------------------------------------------------------------------|--------------------------------------------|
+| `image_path`   | string  | Path to the image file to process                                      | `./path/to/image.png`                      |
+| `prompt`       | string  | The question or instruction for the vision model                       | `Describe the contents of this image`      |
+| `temperature`  | float   | Sampling temperature between 0.0 and 1.0                              | `0.7`                                      |
+
+##### Example Usage
+```python
+vision_tool = LLMVisionTool()
+response = vision_tool.execute(
+    image_path="./path/to/image.png",
+    prompt="Describe the contents of this image",
+    temperature=0.7
+)
+print("Vision Model Response:", response)
+```
+
+#### 15. LLM Tool
 
 The **LLM Tool** generates answers using a specified language model.
 
@@ -669,7 +694,7 @@ print("LLM Response:", response)
 
 ---
 
-### 15. Download HTTP File Tool
+### 16. Download HTTP File Tool
 
 The **Download HTTP File Tool** downloads a file from a specified HTTP URL.
 
@@ -689,7 +714,7 @@ print(result)
 
 ---
 
-### 16. List Directory Tool
+### 17. List Directory Tool
 
 The **List Directory Tool** lists files in a specified directory.
 
@@ -708,7 +733,7 @@ print("Directory Files:", result)
 
 ---
 
-### 17. Markitdown Tool
+### 18. Markitdown Tool
 
 The **Markitdown Tool** processes markdown files, possibly for conversion or rendering.
 
@@ -725,26 +750,6 @@ result = markitdown_tool.execute(markdown_path="./path/to/file.md")
 print("Processed Markdown Output:", result)
 ```
 
----
-
-### 18. Unified Diff Tool
-
-The **Unified Diff Tool** generates a unified diff between two texts or files.
-
-#### Parameters
-
-| Parameter    | Type   | Description                                    | Example                |
-|--------------|--------|------------------------------------------------|------------------------|
-| `original`   | string | The original content or file path.            | `old_text.txt`         |
-| `updated`    | string | The updated content or file path.             | `new_text.txt`         |
-
-#### Example Usage
-```python
-diff_tool = UnifiedDiffTool()
-result = diff_tool.execute(original="old_text.txt", updated="new_text.txt")
-print("Unified Diff Output:", result)
-```
-
 
 #### Creating Custom Tools
 

{quantalogic-0.2.8 → quantalogic-0.2.10}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "quantalogic"
-version = "0.2.8"
+version = "0.2.10"
 description = "QuantaLogic ReAct Agents"
 authors = ["Raphaël MANSUY <raphael.mansuy@gmail.com>"]
 readme = "README.md"

{quantalogic-0.2.8 → quantalogic-0.2.10}/quantalogic/agent.py

@@ -64,6 +64,7 @@ class Agent(BaseModel):
     event_emitter: EventEmitter = EventEmitter()
     config: AgentConfig
     task_to_solve: str
+    task_to_solve_summary: str = ""
     ask_for_user_validation: Callable[[str], bool] = console_ask_for_user_validation
     last_tool_call: dict[str, Any] = {}  # Stores the last tool call information
     total_tokens: int = 0  # Total tokens in the conversation
@@ -137,6 +138,9 @@ class Agent(BaseModel):
         logger.debug(f"Solving task... {task}")
         self._reset_session(task_to_solve=task, max_iterations=max_iterations)
 
+        # Generate task summary
+        self.task_to_solve_summary = self._generate_task_summary(task)
+
         # Add system prompt to memory
         self.memory.add(Message(role="system", content=self.config.system_prompt))
 
@@ -166,7 +170,7 @@ class Agent(BaseModel):
                 self._update_total_tokens(message_history=self.memory.memory, prompt=current_prompt)
 
                 # Emit event: Task Think Start after updating total tokens
-                self._emit_event("task_think_start")
+                self._emit_event("task_think_start", {"prompt": current_prompt})
 
                 self._compact_memory_if_needed(current_prompt)
 
@@ -443,6 +447,10 @@ class Agent(BaseModel):
             "You must analyze this answer and evaluate what to do next to solve the task.\n"
             "If the step failed, take a step back and rethink your approach.\n"
             "\n"
+            "--- Task to solve summary ---\n"
+            "\n"
+            f"{self.task_to_solve_summary}"
+            "\n"
             "--- Format ---\n"
             "\n"
             "You MUST respond with exactly two XML blocks formatted in markdown:\n"
@@ -546,6 +554,7 @@ class Agent(BaseModel):
             "\n### Tools:\n"
             "-----------------------------\n"
             f"{self._get_tools_names_prompt()}\n"
+            "\n"
             "### Variables:\n"
             "-----------------------------\n"
             f"{self._get_variable_prompt()}\n"
@@ -574,7 +583,7 @@ class Agent(BaseModel):
             "\n"
             "Available variables:\n"
             "\n"
-            f"{', '.join(self.variable_store.keys())}\n"
+            f"{', '.join(self.variable_store.keys())}\n" if len(self.variable_store.keys()) > 0 else "None\n"
         )
         return prompt_use_variables
 
@@ -619,6 +628,28 @@ class Agent(BaseModel):
         self.memory.memory = memory_copy
         return summary.response
 
+    def _generate_task_summary(self, content: str) -> str:
+        """Generate a concise summary of the given content using the generative model.
+        
+        Args:
+            content (str): The content to summarize
+            
+        Returns:
+            str: Generated summary
+        """
+        try:
+            prompt = (
+                "Rewrite this task in a precise, dense, and concise manner:\n"
+                f"{content}\n"
+                "Summary should be 2-3 sentences maximum. No extra comments should be added.\n"
+            )
+            result = self.model.generate(prompt=prompt)
+            logger.debug(f"Generated summary: {result.response}")
+            return result.response
+        except Exception as e:
+            logger.error(f"Error generating summary: {str(e)}")
+            return f"Summary generation failed: {str(e)}"
+
     def _update_session_memory(self, user_content: str, assistant_content: str) -> None:
         """
         Log session messages to memory and emit events.
@@ -639,3 +670,5 @@ class Agent(BaseModel):
             "session_add_message",
             {"role": "assistant", "content": assistant_content},
         )
+        
+

quantalogic-0.2.10/quantalogic/agent_config.py

@@ -0,0 +1,169 @@
+"""Module for agent configuration and creation."""
+
+# Standard library imports
+
+# Local application imports
+from quantalogic.agent import Agent
+from quantalogic.coding_agent import create_coding_agent
+from quantalogic.tools import (
+    AgentTool,
+    DownloadHttpFileTool,
+    EditWholeContentTool,
+    ExecuteBashCommandTool,
+    InputQuestionTool,
+    ListDirectoryTool,
+    LLMTool,
+    LLMVisionTool,
+    MarkitdownTool,
+    NodeJsTool,
+    PythonTool,
+    ReadFileBlockTool,
+    ReadFileTool,
+    ReplaceInFileTool,
+    RipgrepTool,
+    SearchDefinitionNames,
+    TaskCompleteTool,
+    WriteFileTool,
+)
+
+MODEL_NAME = "deepseek/deepseek-chat"
+
+
+def create_agent(model_name: str, vision_model_name: str | None) -> Agent:
+    """Create an agent with the specified model and tools.
+
+    Args:
+        model_name (str): Name of the model to use
+        vision_model_name (str | None): Name of the vision model to use
+
+    Returns:
+        Agent: An agent with the specified model and tools
+    """
+    tools = [
+        TaskCompleteTool(),
+        ReadFileTool(),
+        ReadFileBlockTool(),
+        WriteFileTool(),
+        EditWholeContentTool(),
+        InputQuestionTool(),
+        ListDirectoryTool(),
+        ExecuteBashCommandTool(),
+        ReplaceInFileTool(),
+        RipgrepTool(),
+        SearchDefinitionNames(),
+        MarkitdownTool(),
+        LLMTool(model_name=model_name),
+        DownloadHttpFileTool(),
+    ]
+
+    if vision_model_name:
+        tools.append(LLMVisionTool(model_name=vision_model_name))
+
+    return Agent(
+        model_name=model_name,
+        tools=tools,
+    )
+
+
+def create_interpreter_agent(model_name: str, vision_model_name: str | None) -> Agent:
+    """Create an interpreter agent with the specified model and tools.
+
+    Args:
+        model_name (str): Name of the model to use
+        vision_model_name (str | None): Name of the vision model to use
+
+    Returns:
+        Agent: An interpreter agent with the specified model and tools
+    """
+    tools = [
+        TaskCompleteTool(),
+        ReadFileTool(),
+        ReadFileBlockTool(),
+        WriteFileTool(),
+        EditWholeContentTool(),
+        InputQuestionTool(),
+        ListDirectoryTool(),
+        ExecuteBashCommandTool(),
+        ReplaceInFileTool(),
+        RipgrepTool(),
+        PythonTool(),
+        NodeJsTool(),
+        SearchDefinitionNames(),
+        MarkitdownTool(),
+        LLMTool(model_name=model_name),
+        DownloadHttpFileTool(),
+    ]
+    return Agent(model_name=model_name, tools=tools)
+
+
+def create_full_agent(model_name: str, vision_model_name: str | None) -> Agent:
+    """Create an agent with the specified model and many tools.
+
+    Args:
+        model_name (str): Name of the model to use
+        vision_model_name (str | None): Name of the vision model to use
+
+    Returns:
+        Agent: An agent with the specified model and tools
+
+    """
+    tools=[
+        TaskCompleteTool(),
+        ReadFileTool(),
+        ReadFileBlockTool(),
+        WriteFileTool(),
+        EditWholeContentTool(),
+        InputQuestionTool(),
+        ListDirectoryTool(),
+        ExecuteBashCommandTool(),
+        ReplaceInFileTool(),
+        RipgrepTool(),
+        PythonTool(),
+        NodeJsTool(),
+        SearchDefinitionNames(),
+        MarkitdownTool(),
+        LLMTool(model_name=model_name),
+        DownloadHttpFileTool(),
+    ]
+
+    if vision_model_name:
+        tools.append(LLMVisionTool(model_name=vision_model_name))
+
+    return Agent(
+        model_name=model_name,
+        tools=tools,
+    )
+
+
+def create_orchestrator_agent(model_name: str, vision_model_name: str | None = None) -> Agent:
+    """Create an agent with the specified model and tools.
+
+    Args:
+        model_name (str): Name of the model to use
+        vision_model_name (str | None): Name of the vision model to use
+
+    Returns:
+        Agent: An agent with the specified model and tools
+    """
+    # Rebuild AgentTool to resolve forward references
+    AgentTool.model_rebuild()
+
+    coding_agent_instance = create_coding_agent(model_name)
+
+    tools = [
+        TaskCompleteTool(),
+        ListDirectoryTool(),
+        ReadFileBlockTool(),
+        RipgrepTool(),
+        SearchDefinitionNames(),
+        LLMTool(model_name=MODEL_NAME),
+        AgentTool(agent=coding_agent_instance, agent_role="software expert", name="coder_agent_tool"),
+    ]
+
+    if vision_model_name:
+        tools.append(LLMVisionTool(model_name=vision_model_name))
+
+    return Agent(
+        model_name=model_name,
+        tools=tools,
+    )

{quantalogic-0.2.8 → quantalogic-0.2.10}/quantalogic/coding_agent.py

@@ -5,6 +5,7 @@ from quantalogic.tools import (
     InputQuestionTool,
     ListDirectoryTool,
     LLMTool,
+    LLMVisionTool,
     ReadFileBlockTool,
     ReadFileTool,
     ReplaceInFileTool,
@@ -17,11 +18,12 @@ from quantalogic.utils import get_coding_environment
 from quantalogic.utils.get_quantalogic_rules_content import get_quantalogic_rules_file_content
 
 
-def create_coding_agent(model_name: str, basic: bool = False) -> Agent:
+def create_coding_agent(model_name: str,vision_model_name: str | None = None, basic: bool = False) -> Agent:
     """Creates and configures a coding agent with a comprehensive set of tools.
 
     Args:
         model_name (str): Name of the language model to use for the agent's core capabilities
+        vision_model_name (str | None): Name of the vision model to use for the agent's core capabilities
         basic (bool, optional): If True, the agent will be configured with a basic set of tools.
 
     Returns:
@@ -59,6 +61,9 @@ def create_coding_agent(model_name: str, basic: bool = False) -> Agent:
         InputQuestionTool(),
     ]
 
+    if vision_model_name:
+        tools.append(LLMVisionTool(model_name=vision_model_name))
+
     if not basic:
         tools.append(
             LLMTool(

{quantalogic-0.2.8 → quantalogic-0.2.10}/quantalogic/generative_model.py

@@ -5,16 +5,17 @@ from litellm import completion, exceptions, get_max_tokens, get_model_info, toke
 from loguru import logger
 from pydantic import BaseModel, Field, field_validator
 
-MIN_RETRIES = 3
+MIN_RETRIES = 1
 
 
 class Message(BaseModel):
     """Represents a message in a conversation with a specific role and content."""
 
     role: str = Field(..., min_length=1)
-    content: str = Field(..., min_length=1)
+    content: str | dict = Field(..., min_length=1)
+    image_url: str | None = Field(default=None, pattern=r"^https?://")
 
-    @field_validator("role", "content")
+    @field_validator("role")
     @classmethod
     def validate_not_empty(cls, v: str) -> str:
         """Validate that the field is not empty or whitespace-only."""
@@ -22,6 +23,26 @@ class Message(BaseModel):
             raise ValueError("Field cannot be empty or whitespace-only")
         return v
 
+    @field_validator("content")
+    @classmethod
+    def validate_content(cls, v: str | dict) -> str | dict:
+        """Validate content based on its type."""
+        if isinstance(v, str):
+            if not v or not v.strip():
+                raise ValueError("Text content cannot be empty or whitespace-only")
+        elif isinstance(v, dict):
+            if not v.get("text") or not v.get("image_url"):
+                raise ValueError("Multimodal content must have both text and image_url")
+        return v
+
+    @field_validator("image_url")
+    @classmethod
+    def validate_image_url(cls, v: str | None) -> str | None:
+        """Validate image URL format if present."""
+        if v and not v.startswith(("http://", "https://")):
+            raise ValueError("Image URL must start with http:// or https://")
+        return v
+
 
 class TokenUsage(BaseModel):
     """Represents token usage statistics for a language model."""
@@ -59,6 +80,7 @@ class GenerativeModel:
             temperature: Sampling temperature between 0 and 1.
                 Defaults to 0.7.
         """
+        logger.debug(f"Initializing GenerativeModel with model={model}, temperature={temperature}")
         self.model = model
         self.temperature = temperature
 
@@ -85,15 +107,16 @@ class GenerativeModel:
     )
 
     # Retry on specific retriable exceptions
-    def generate_with_history(self, messages_history: list[Message], prompt: str) -> ResponseStats:
-        """Generate a response with conversation history.
+    def generate_with_history(self, messages_history: list[Message], prompt: str, image_url: str | None = None) -> ResponseStats:
+        """Generate a response with conversation history and optional image.
 
-        Generates a response based on previous conversation messages
-        and a new user prompt.
+        Generates a response based on previous conversation messages,
+        a new user prompt, and an optional image URL.
 
         Args:
             messages_history: Previous conversation messages.
             prompt: Current user prompt.
+            image_url: Optional image URL for visual queries.
 
         Returns:
             Detailed response statistics.
@@ -105,7 +128,22 @@ class GenerativeModel:
             Exception: For other unexpected errors.
         """
         messages = [{"role": msg.role, "content": str(msg.content)} for msg in messages_history]
-        messages.append({"role": "user", "content": str(prompt)})
+        
+        if image_url:
+            messages.append({
+                "role": "user",
+                "content": [
+                    {"type": "text", "text": str(prompt)},
+                    {
+                        "type": "image_url",
+                        "image_url": {
+                            "url": image_url
+                        }
+                    }
+                ]
+            })
+        else:
+            messages.append({"role": "user", "content": str(prompt)})
 
         try:
             logger.debug(f"Generating response for prompt: {prompt}")
@@ -140,9 +178,12 @@ class GenerativeModel:
             }
 
             logger.error("LLM Generation Error: {}", error_details)
+            logger.debug(f"Error details: {error_details}")
+            logger.debug(f"Model: {self.model}, Temperature: {self.temperature}")
 
             # Handle authentication and permission errors
             if isinstance(e, self.AUTH_EXCEPTIONS):
+                logger.debug("Authentication error occurred")
                 raise openai.AuthenticationError(
                     f"Authentication failed with provider {error_details['provider']}"
                 ) from e
@@ -162,7 +203,7 @@ class GenerativeModel:
             # Wrap unknown errors in APIError
             raise openai.APIError(f"Unexpected error during generation: {str(e)}") from e
 
-    def generate(self, prompt: str) -> ResponseStats:
+    def generate(self, prompt: str, image_url: str | None = None) -> ResponseStats:
         """Generate a response without conversation history.
 
         Generates a response for a single user prompt without
@@ -170,11 +211,12 @@ class GenerativeModel:
 
         Args:
             prompt: User prompt.
+            image_url: Optional image URL for visual queries.
 
         Returns:
             Detailed response statistics.
         """
-        return self.generate_with_history([], prompt)
+        return self.generate_with_history([], prompt, image_url)
 
     def get_max_tokens(self) -> int:
         """Get the maximum number of tokens that can be generated by the model."""
@@ -182,8 +224,11 @@ class GenerativeModel:
 
     def token_counter(self, messages: list[Message]) -> int:
         """Count the number of tokens in a list of messages."""
+        logger.debug(f"Counting tokens for {len(messages)} messages using model {self.model}")
         litellm_messages = [{"role": msg.role, "content": str(msg.content)} for msg in messages]
-        return token_counter(model=self.model, messages=litellm_messages)
+        token_count = token_counter(model=self.model, messages=litellm_messages)
+        logger.debug(f"Token count: {token_count}")
+        return token_count
 
     def token_counter_with_history(self, messages_history: list[Message], prompt: str) -> int:
         """Count the number of tokens in a list of messages and a prompt."""
@@ -193,12 +238,18 @@ class GenerativeModel:
 
     def get_model_info(self) -> dict | None:
         """Get information about the model."""
+        logger.debug(f"Retrieving model info for {self.model}")
         model_info = get_model_info(self.model)
 
         if not model_info:
-            # Search without prefix "openrouter/"
+            logger.debug("Model info not found, trying without openrouter/ prefix")
             model_info = get_model_info(self.model.replace("openrouter/", ""))
 
+        if model_info:
+            logger.debug(f"Model info retrieved: {model_info.keys()}")
+        else:
+            logger.debug("No model info available")
+            
         return model_info
 
     def get_model_max_input_tokens(self) -> int: