LangChain as a Coding Assistant: Fibonacci & Unit Testing

Abstract

LangChain provides a powerful way to generate both code and unit tests automatically using LLMs. This publication explores how LangChain generates a Python function for the Fibonacci sequence and its corresponding unit tests, showcasing AI-assisted software development.

Methodology

1️⃣ User Provides Arguments

When you run the script:

python main.py --language python --task "generate Fibonacci sequence"

• --language specifies the programming language (e.g., Python, JavaScript).
• --task specifies the code to generate (e.g., "generate Fibonacci sequence").

2️⃣ Generating Code Using LangChain

LangChain takes the arguments and constructs a prompt using a PromptTemplate:

code_prompt = PromptTemplate(
    input_variables=["task", "language"],
    template="Write a standalone {language} function that will {task}. "
             "Only provide the function implementation with a proper function signature and return statement. "
             "Do not include any test cases, assertions, or example usages."
)

Example Prompt Sent to OpenAI:

Write a standalone Python function that will generate a Fibonacci sequence.
Only provide the function implementation with a proper function signature and return statement.
Do not include any test cases, assertions, or example usages.

Generated Code Example:

def generate_fibonacci_sequence(n):
    if n <= 0:
        return []
    elif n == 1:
        return [0]
    elif n == 2:
        return [0, 1]

    fibonacci_sequence = [0, 1]
    for i in range(2, n):
        next_num = fibonacci_sequence[i-1] + fibonacci_sequence[i-2]
        fibonacci_sequence.append(next_num)

    return fibonacci_sequence

3️⃣ Generating Unit Tests with LangChain

Once the function is generated, it is passed into test_prompt:

test_prompt = PromptTemplate(
    input_variables=["code", "language"],
    template="Write a full unit test in {language} using unittest for the following function. "
             "Include an import statement assuming the function is defined in a separate module named `your_module.py`. "
             "Only provide the test class and methods, do not include the function implementation itself:\n\n{code}"
)

Example Prompt Sent to OpenAI:

Write a full unit test in Python using unittest for the following function.
Include an import statement assuming the function is defined in a separate module named `fibonacci.py`.
Only provide the test class and methods, do not include the function implementation itself.

Generated Unit Test Example:

import unittest
from fibonacci import generate_fibonacci_sequence

class TestGenerateFibonacciSequence(unittest.TestCase):
    def test_generate_fibonacci_sequence_with_n_0(self):
        self.assertEqual(generate_fibonacci_sequence(0), [])

    def test_generate_fibonacci_sequence_with_n_1(self):
        self.assertEqual(generate_fibonacci_sequence(1), [0])

    def test_generate_fibonacci_sequence_with_n_2(self):
        self.assertEqual(generate_fibonacci_sequence(2), [0, 1])

    def test_generate_fibonacci_sequence_with_n_5(self):
        self.assertEqual(generate_fibonacci_sequence(5), [0, 1, 1, 2, 3])

    def test_generate_fibonacci_sequence_with_n_10(self):
        self.assertEqual(generate_fibonacci_sequence(10), [0, 1, 1, 2, 3, 5, 8, 13, 21, 34])

if __name__ == '__main__':
    unittest.main()

Results

Run the Fibonacci Function

python fibonacci.py

Run the Unit Tests

✅ Method 1: Run Directly

python test_fibonacci.py

✅ Method 2: Using unittest

python -m unittest test_fibonacci.py

✅ Method 3: Run All Tests in the Directory

python -m unittest discover

Conclusion

LangChain successfully generated both a function and unit tests, reducing manual effort. However, human oversight is essential to refine AI-generated code, especially for edge cases.

✅ Key Takeaways:

• LangChain can accelerate boilerplate code writing & testing.
• AI-generated code still requires verification & debugging.
• Future improvements in LLMs may lead to fully autonomous coding agents.

---

✍️ Author: Mary Ann Dizon
🔗 Connect: LinkedIn Profile / GitHub
📌 Publication on Ready Tensor