Building a Multi-Modal Code Explainer with Python and OpenAI

Prerequisites

To follow along, you need a basic understanding of Python and how to interact with REST APIs. You will also need active API keys for OpenAI and Eleven Labs. Ensure you have pip installed to manage dependencies.

Key Libraries & Tools

• Streamlit: A framework that turns Python scripts into interactive web apps in minutes.
• OpenAI API: Specifically the gpt-3.5-turbo model for high-speed, cost-effective code analysis.
• Eleven Labs: A high-fidelity TTS platform for generating realistic human voices.
• python-dotenv: For securely managing sensitive API credentials in a .env file.

Code Walkthrough

1. The AI Logic

We use the openai.ChatCompletion endpoint to process the code. Notice how we set the "system" role to define the AI's persona as a developer.

import openai

def retrieve_code_explanation(code):
    response = openai.ChatCompletion.create(
        model="gpt-3.5-turbo",
        messages=[
            {"role": "system", "content": "You are a developer assistant."},
            {"role": "user", "content": f"Explain this code in one paragraph: {code}"}
        ]
    )
    return response.choices[0].message.content

2. Converting Text to Speech

After receiving the text explanation, we pass it to Eleven Labs. We use a while loop to handle potential API latency and ensure the binary data saves correctly as an MP3.

import requests

def save_binary_to_mp3(content, filename="explanation.mp3"):
    with open(filename, "wb") as f:
        f.write(content)

3. The Streamlit Interface

Streamlit handles the UI. We use st.file_uploader for scripts and st.audio to play the generated MP3 directly in the browser.

Syntax Notes

This project relies on environment variables. Using load_dotenv() is a best practice to avoid hardcoding API keys. We also use f-strings for prompt engineering, allowing us to inject user-provided code directly into the API request string.

Tips & Gotchas

Precision in your prompts is vital. If you want a concise answer, explicitly tell the AI to "explain in one word" or "in one paragraph." This prevents the model from rambling and saves on token costs. Additionally, always check the API status of your TTS provider, as these services often experience higher latency than text-based LLMs.