v0.2.4, added the CodeToolWorkflow and associated examples/documentat…

…ion.

docs/workflows/code_tool_workflow.md

@@ -0,0 +1,151 @@
+# ReXia.AI CodeToolWorkflow Class
+
+## Overview
+
+The `CodeToolWorkflow` class is a specialized workflow implementation in the ReXia.AI framework. It extends the `BaseWorkflow` class and implements a process for generating and executing custom code tools to solve complex tasks. This workflow is designed to be used within an Agent instance and leverages a large language model (LLM) to dynamically create Python functions that can assist in completing the given task.
+
+## Table of Contents
+
+- [Class Attributes](#class-attributes)
+- [Methods](#methods)
+- [Usage](#usage)
+- [Components](#components)
+- [Dependencies](#dependencies)
+- [Contributing](#contributing)
+- [License](#license)
+
+## Class Attributes
+
+- `llm`: The language model used by the workflow.
+- `task`: The task that the workflow is designed to perform.
+- `verbose`: A flag used for enabling verbose mode.
+- `memory`: The memory component of the workflow.
+- `channel`: The collaboration channel for the workflow.
+- `max_attempts`: The maximum number of attempts for code generation.
+
+## Methods
+
+### `__init__(self, llm: Any, task: str, memory: BaseMemory, verbose: bool = False, max_attempts: int = 3) -> None`
+
+Initializes a CodeToolWorkflow instance.
+
+**Parameters:**
+
+- `llm`: The language model used by the workflow.
+- `task`: The task assigned to the workflow.
+- `memory`: The memory instance used by the workflow.
+- `verbose`: A flag for enabling verbose mode. Defaults to `False`.
+- `max_attempts`: The maximum number of attempts for code generation. Defaults to `3`.
+
+### `_run_task(self) -> None`
+
+Internal method that executes the main CodeToolWorkflow process.
+
+### `run(self) -> str`
+
+Public method to initiate the workflow execution.
+
+## Usage
+
+Here's an example of how to use the `CodeToolWorkflow` class with an Agent in ReXia.AI:
+
+```python
+import os
+from rexia_ai.llms import RexiaAIOpenAI
+from rexia_ai.agents import Agent
+from rexia_ai.workflows import CodeToolWorkflow
+from rexia_ai.memory import SimpleMemory
+
+# Set up the language model
+YI_LARGE_API_KEY = os.getenv("YI_LARGE_API_KEY")
+llm = RexiaAIOpenAI(
+    base_url="https://api.01.ai/v1",
+    model="yi-large",
+    temperature=0,
+    api_key=YI_LARGE_API_KEY,
+)
+
+# Define the task
+task = "Calculate the sum of squares for the first 100 prime numbers."
+
+# Create an Agent instance with CodeToolWorkflow
+agent = Agent(
+    llm=llm,
+    task=task,
+    workflow=CodeToolWorkflow,
+    verbose=True
+)
+
+# Run the agent (which will use the CodeToolWorkflow)
+response = agent.invoke()
+
+# Print the response
+print("Task:", task)
+print("Response:", response)
+```
+
+## Components
+
+The `CodeToolWorkflow` primarily utilizes the following components:
+
+1. **Large Language Model (LLM)**: Generates Python code based on the given task.
+2. **ContainerisedCodeTester**: Executes the generated code in an isolated Docker container.
+
+### Large Language Model (LLM)
+
+The LLM is responsible for generating Python code that can solve the given task. Key features include:
+
+- Receiving a task description and generating appropriate Python functions
+- Handling complex tasks by breaking them down into manageable steps
+- Iteratively improving the generated code based on execution results
+
+### ContainerisedCodeTester
+
+The ContainerisedCodeTester class is responsible for executing the generated Python code in isolated Docker containers. Key features include:
+
+- Initialization with a specified Docker image (default: "python:3.12-slim") and execution timeout
+- Execution of code in a controlled Docker environment
+- Writing code to a temporary directory
+- Running the Docker container with the provided code
+- Parsing and returning the execution results
+
+Key methods:
+
+- `execute_code(code)`: Executes the provided code in a Docker container
+- `_write_files(tmpdir, code)`: Writes the code files to a temporary directory
+- `_run_container(tmpdir)`: Runs the Docker container with the provided code
+- `_parse_output(output)`: Parses the output from the Docker container
+
+The ContainerisedCodeTester ensures that the generated code runs in a secure, isolated environment with:
+
+- Limited memory (128MB)
+- Limited CPU quota
+- No network access
+- Read-only access to the code files
+
+## Dependencies
+
+- `typing`
+- `docker`
+- `tempfile`
+- `os`
+- `json`
+- ReXia.AI components (`BaseWorkflow`, `BaseMemory`, `CollaborationChannel`, `TaskStatus`, `Component`)
+- ReXia.AI Agent class
+
+Ensure all dependencies are installed and properly imported.
+
+## Contributing
+
+We welcome contributions to improve the ReXia.AI framework. Please follow these steps to contribute:
+
+1. Fork the repository.
+2. Create a new branch (`git checkout -b feature-branch`).
+3. Make your changes.
+4. Commit your changes (`git commit -m 'Add new feature'`).
+5. Push to the branch (`git push origin feature-branch`).
+6. Create a new Pull Request.
+
+## License
+
+This project is licensed under the Apache License 2.0. See the [LICENSE](../LICENSE) file for details.

examples/code_tool_workflow_example.py

@@ -0,0 +1,32 @@
+"""Example agent in ReXia.AI using CodeToolWorkflow for a complex mathematical task."""
+
+import os
+from rexia_ai.agents import Agent
+from rexia_ai.llms import RexiaAIOpenAI
+from rexia_ai.workflows import CodeToolWorkflow
+
+YI_LARGE_API_KEY = os.getenv("YI_LARGE_API_KEY")
+
+# Create an instance of the ReXiaAI LLM
+llm = RexiaAIOpenAI(
+    base_url="https://api.01.ai/v1",
+    model="yi-large",
+    temperature=0,
+    api_key=YI_LARGE_API_KEY,
+)
+
+# Define a complex mathematical task
+task = "Calculate the first 100 prime numbers and then find the sum of their digits."
+
+# Create an instance of the ReXiaAI Agent with the specified task, LLM and workflow
+agent = Agent(
+    llm=llm,
+    task=task,
+    workflow=CodeToolWorkflow,
+    verbose=True,
+)
+
+# Generate the response from the agent
+response = agent.invoke()
+
+print("Response:", response)

obj/Debug/net8.0/rexia_ai.AssemblyInfo.cs

@@ -13,7 +13,7 @@
 [assembly: System.Reflection.AssemblyCompanyAttribute("rexia_ai")]
 [assembly: System.Reflection.AssemblyConfigurationAttribute("Debug")]
 [assembly: System.Reflection.AssemblyFileVersionAttribute("1.0.0.0")]
-[assembly: System.Reflection.AssemblyInformationalVersionAttribute("1.0.0+8a4ca504b66337d9f7ed81ca32b5762568effe96")]
+[assembly: System.Reflection.AssemblyInformationalVersionAttribute("1.0.0+1abf9bea62a5b05a8990de615198fc729e8cf50c")]
 [assembly: System.Reflection.AssemblyProductAttribute("rexia_ai")]
 [assembly: System.Reflection.AssemblyTitleAttribute("rexia_ai")]
 [assembly: System.Reflection.AssemblyVersionAttribute("1.0.0.0")]

obj/Debug/net8.0/rexia_ai.AssemblyInfoInputs.cache

@@ -1 +1 @@
-6860acc8e31b2544d30eaef17b437cc430337473f1b17c86cf26ef5c164d9de7
+c9b03f83e6ca55a1b9af8c9f5dd15f306e3470a2632dea9dd3178d35347e02c2

setup.py

@@ -6,7 +6,7 @@
 
 setup(
     name='rexia_ai',
-    version='0.2.3',
+    version='0.2.4',
     author='Robyn Le Sueur',
     author_email='[email protected]',
     description='ReXia.AI: An advanced AI framework for agentic processes',

src/rexia_ai/agents/workers/__init__.py

@@ -6,12 +6,14 @@
 from .finalise import FinaliseWorker
 from .team_work import TeamWorker
 from .tdd import TDDWorker
+from .llm_tool import LLMTool
 
 __all__ = [
     "ToolWorker",
     "Worker",
     "PlanWorker",
     "FinaliseWorker",
     "TeamWorker",
-    "TDDWorker"
+    "TDDWorker",
+    "LLMTool"
 ]