ex2codeai

ex2codeai is a Python library designed to streamline the process of generating Python functions and classes based on user-defined specifications and examples. It integrates with Large Language Models (LLMs) to automate the creation of code that adheres to specific behaviors and descriptions.

---

Features

• Function Specification:

  • Define function names, descriptions, and behavior with examples.
  • Automatically generate Python functions that match specified logic.
  • Parse and test generated functions.

• Class Specification:

  • Create Python classes with instance, class, and static methods.
  • Add method definitions with detailed descriptions and examples.
  • Generate class implementations dynamically using LLMs.

• LLM Integration:

  • Seamless integration with langchain to utilize various LLMs.
  • Customizable prompts for tailored code generation.

---

Installation

To use ex2codeai, ensure you have the following installed:

• Python 3.9 or later
• langchain (langchain==0.3.7)
• A compatible LLM (e.g., OllamaLLM) (langchain-ollama==0.2.0)

Install the required dependencies:

pip install langchain
pip install your-llm-wrapper  # e.g., langchain-ollama

---

Usage Workflow

1. Initialize an LLM: Connect to your preferred LLM using supported tools and APIs.

   from ex2codeai import *
   from langchain_ollama.llms import OllamaLLM
   
   llm = OllamaLLM(base_url="https://your-ollama-instance-url", model="your-model-name")

   Note: You can use Google Colab for serving Ollama. See https://github.com/mohammadT77/ollama-colab

2. Define Specifications: Specify the desired functions or classes, including names, descriptions, and examples.

   1. Define Functions:

   # Define a function specification
   _add_spec = FuncSpec("add", "add two integers (a and b)") \
       .add_example({'a': 1, 'b': 2}, "3") \
       .add_example({'a': 3, 'b': 4}, "7") \
       .add_example({'a': -1, 'b': 7}, "6")
   
   print("_add_spec prompt:", _add_spec.prompt())
   
   # Generate the function
   _add_str = _add_spec.invoke(llm)
   print("_add_str", _add_str)
   
   # Parse and test the function
   add = _add_spec.parse(_add_str)
   # Or:
   add = _add_spec.generate(llm)  # calls .invoke and .parse
   
   print(add(1, 2))  # Output: 3

   2. Define Classes:

   # Define a class specification
   _MyClass_spec = ClassSpec("MyClass", "A simple class having methods for arithmetic operations.")
   _MyClass_spec.add_instance_method("add", "add two integers (a and b)",
       Example({'a': 1, 'b': 2}, "3"),
       Example({'a': 3, 'b': 4}, "7"))
   _MyClass_spec.add_class_method("sub", "subtract b from a",
       Example({'a': 5, 'b': 2}, "3"))
   _MyClass_spec.add_static_method("mul", "multiply two integers (a and b)",
       Example({'a': 2, 'b': 3}, "6"))
   _MyClass_spec.add_instance_method("div", "divide two integers (a and b)",
       Example({'a': 4, 'b': 2}, "2.0"))
   
   # Generate the class
   print(_MyClass_spec.prompt())
   MyClassStr = _MyClass_spec.invoke(llm)
   print(MyClassStr)
   
   # Parse and test the class
   MyClass = _MyClass_spec.parse(MyClassStr)
   myClass = MyClass()
   print(myClass.add(1, 2))  # Output: 3

3. Test the Output: Parse and execute the generated code to validate its correctness and functionality.

---

Components

Classes

1. Example:

   1. Represents an example input-output pair.
   2. Converts examples to dictionary format for LLM prompts.

2. Spec:

   1. Base class for all specifications.
   2. Includes methods for invoking LLMs and parsing their outputs.

3. FuncSpec:

   1. Extends Spec for defining and generating functions.
   2. Supports adding multiple input-output examples.

4. ClassSpec:

   1. Extends Spec for defining and generating classes.
   2. Supports instance, class, and static methods with examples.

LLM Integration

• The library uses LangChain for interacting with LLMs.
• Prompts are constructed dynamically to guide the LLM in generating Python code.

---

Example Prompts

The library generates structured prompts for the LLM, like the following:

Function Prompt

You are a Python programmer. Your task is to write a Python function that matches the following examples.
Each example consists of an input and its respective output.

Examples:
Input: (a: 1, b: 2), Output: 3
Input: (a: 3, b: 4), Output: 7
Input: (a: -1, b: 7), Output: 6

Description:
add two integers (a and b)

Write a Python function named `add` that implements this logic. Do not include any additional explanation or comments, just the code.

Class Prompt:

You are a Python programmer. Your task is to write a simple Python class that matches the following description and has the following methods.

Description:
A simple class having methods for arithmetic operations.

Instance Methods:
Name: add, Description: add two integers (a and b), Examples:
Input: (a: 1, b: 2), Output: 3
Input: (a: 3, b: 4), Output: 7

Class Methods:
Name: sub, Description: subtract b from a, Examples:
Input: (a: 5, b: 2), Output: 3

Static Methods:
Name: mul, Description: multiply two integers (a and b), Examples:
Input: (a: 2, b: 3), Output: 6

Write a Python class named `MyClass` that implements this logic. Do not include any additional explanation or comments, just the code.

---

License

This project is licensed under the Apache License.

---

Acknowledgments

This library is built on the langchain framework and leverages the power of modern LLMs to deliver flexible and efficient code generation capabilities.