Skip to content

Using the OpenAI API in Python

The OpenAI API gives you access to state-of-the-art models like GPT for text and Whisper for speech. This guide walks through authentication, key endpoints, and helpful utilities for robust integration in Python.

1. Setup: Authentication and Environment

Step 1: Install required libraries

pip install openai python-dotenv

Step 2: Store your API key securely

Create a .env file in your project root:

OPENAI_API_KEY=sk-...

Step 3: Load the key in your Python script

from dotenv import load_dotenv
import os
import openai

load_dotenv()
openai.api_key = os.getenv("OPENAI_API_KEY")

✅ Use .env to keep credentials out of source control.

2. Basic Text Completion

Use the Completion endpoint to generate or extend text from a prompt:

response = openai.Completion.create(
    model="text-davinci-003",
    prompt="Explain the benefits of unit testing",
    max_tokens=50,
)

print(response["choices"][0]["text"].strip())

🔍 text-davinci-003 excels at classic single-turn completions.

3. Chat-Based Conversations

For multi-turn interactions, use ChatCompletion.create() with role-based messaging:

chat = openai.ChatCompletion.create(
    model="gpt-3.5-turbo",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Summarize agile development."},
    ],
)

print(chat["choices"][0]["message"]["content"].strip())

🧠 Use system messages to guide the assistant’s behavior.

4. Audio Transcription with Whisper

Transcribe speech to text using the Audio.transcribe() method:

with open("meeting.wav", "rb") as audio_file:
    transcript = openai.Audio.transcribe("whisper-1", audio_file)

print(transcript["text"])

🎙️ Supports WAV, MP3, MP4, M4A, and more.

5. Utility Functions for Reliability and Insight

These helpers improve reliability and give insight into model usage:

import time
import tiktoken

encoding = tiktoken.encoding_for_model("gpt-3.5-turbo")

# Count tokens in a string
def count_tokens(text: str) -> int:
    return len(encoding.encode(text))

# Retry API calls with exponential backoff
def retry(func, *args, max_retries=3, **kwargs):
    for i in range(max_retries):
        try:
            return func(*args, **kwargs)
        except openai.error.OpenAIError:
            time.sleep(2 ** i)
    raise

# Extract message text from chat response
def format_chat(reply) -> str:
    return reply["choices"][0]["message"]["content"].strip()

🛠 Token counting helps manage limits and costs.

6. How It All Connects

sequenceDiagram
    participant Dev as Your Script
    participant API as OpenAI API

    Dev->>API: Request with API key
    API-->>Dev: JSON response
    Dev->>Dev: Parse & display result

Summary

With just a few lines of setup, you can:

  • Generate and continue text
  • Run multi-turn conversations
  • Transcribe audio
  • Count tokens and handle retries

For a handy overview of parameters and model limits, see the OpenAI API Cheat Sheet and the OpenAI API Quick Reference.

Additional Resources

Always keep your API key secure, handle failures gracefully, and monitor your usage to stay within limits.

Ready to go deeper? Try building a CLI assistant or auto-summarizer next.

Follow-Up Exercises

Put what you've learned into practice with these standalone projects:

By learning to authenticate requests, handle responses, and manage errors, you can integrate LLMs into your applications. Building on Lesson 02 – Prompt Engineering's reminder that clear instructions shape reliable outputs, Lesson 04 – Understanding response_format={'type': 'json_object'} in LLM APIs will show how to enforce structured JSON replies for downstream processing.