Module 7: Working with Text Completion in OpenAI

In this module, we’ll explore how to use the Text Completion API to generate text from a prompt using models like text-davinci-003.

This API is useful when you want the model to complete a sentence, paragraph, or idea based on a given start.

---

What is a Text Completion?

A completion means you’re asking the model to “complete the text you started.”
For example, if your prompt is:

"Once upon a time, in a galaxy far away,"

The model will try to continue that story in a logical and coherent way.

---

Common Use Cases

• Story writing
• Explanation or reasoning
• Translation
• Grammar correction
• Code generation

---

Basic Setup

Install the OpenAI package and import the required libraries:

!pip install openai

import openai
import os

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

---

Making a Completion Request

Here’s a basic example of using the Text Completion endpoint:

response = openai.Completion.create(
    model="text-davinci-003",
    prompt="Write a tweet about climate change.",
    max_tokens=60)
print(response.choices[0].text.strip())

---

Key Parameters

Parameter	Description
prompt	The text you want the model to complete
model	The completion model (e.g., text-davinci-003)
max_tokens	Max length of the output
temperature	Controls randomness (0 = deterministic, 1 = very random)
top_p	Controls diversity via nucleus sampling
n	Number of completions to generate
stop	Stop generation at specific token(s)

---

Example with All Parameters:

response = openai.Completion.create(
    model="text-davinci-003",
    prompt="Explain the concept of gravity in simple words.",
    temperature=0.7,
    max_tokens=100,
    top_p=0.9,
    n=1,
    stop=["\n"])
print(response.choices[0].text.strip())

---

Use Case: Grammar Correction

prompt = "Correct this sentence: 'She no went to the market today.'"
response = openai.Completion.create(
    model="text-davinci-003",
    prompt=prompt,
    temperature=0,
    max_tokens=60)
print(response.choices[0].text.strip())

---

Use Case: Email Drafting

prompt = "Write a formal email to a colleague asking for help on a project due next week."
response = openai.Completion.create(
    model="text-davinci-003",
    prompt=prompt,
    max_tokens=150,
    temperature=0.6)
print(response.choices[0].text.strip())

---

Best Practices

• Be specific in your prompts: vague prompts give vague results.
• Limit token size if you only need short responses.
• Use temperature and top_p to balance creativity and control.

---

When to Use Completion vs ChatCompletion?

Task	Use This API
Long-form generation	Completion.create()
Conversational systems	ChatCompletion.create()
Role-based agents	ChatCompletion.create()
Fill-in-the-blank style	Completion.create()

---

Summary

• Text Completion is ideal for open-ended text generation.
• Use text-davinci-003 for best results.
• Tune parameters like temperature, max_tokens, and stop for optimal behavior.

---