🔵 🟡 PythonGPT: My Virtual Assistant Rachel 🟡 🔵

Introducing PythonGPT, the name I’ve given to the code for my virtual assistant, Rachel. It’s a chatbot written in Python that leverages OpenAI’s GPT-3 (Generative Pre-trained Transformer 3) model.

🧠 I learned about the release of OpenAI’s public APIs from Davide Severini’s post, where he discussed how a Python GPT chatbot, trained specifically for his needs, could benefit his work at Thauma - Digital Performance Intelligence.

I’m not sure how it will benefit me immediately, but having a virtual assistant on hand that you can ask anything is definitely a plus.

❓ What’s next?

I’ll definitely look into implementing GPT-4 as soon as the public APIs are available. Meanwhile, I’m exploring the feasibility of a version in Java, and perhaps my next post will detail that implementation!

If you’re interested in the code, you can find it below

Code Analysis

The first part of the code demonstrates how the API key for OpenAI is securely managed. Instead of hardcoding the key directly into the script, the key is stored in a separate file called secrets.json. The script reads this file, extracts the API key, and then assigns it to the openai.api_key variable, which allows the program to authenticate and make requests to the OpenAI API.

The core functionality of the script lies in the get_response function, which handles the interaction with the OpenAI API. This function takes a list of messages as input and sends them to the gpt-3.5-turbo model, which then generates a response. The temperature parameter is set to 1.0, which indicates a higher level of creativity in the model’s responses. The function returns the content of the first response choice, which is then used to continue the conversation.

In the main part of the script, the conversation setup begins with a list of messages, including a system-level instruction that establishes Rachel’s identity and her language capabilities. The script then enters an infinite loop, where it repeatedly prompts the user for input. The user’s input is stored and sent to the OpenAI API through the get_response function, and the assistant’s response is printed out, simulating a real-time conversation. The loop continues until the user decides to interrupt it.

Next, the code snippet related to the KeyboardInterrupt exception provides a graceful way to handle user interruptions. If the user decides to stop the execution of the script (for example, by pressing Ctrl+C), the program catches this interruption and prints a personalized farewell message that includes the username, ensuring a smooth and user-friendly exit.