How To Use Different Prompting Techniques in GitHub Copilot

GitHub Copilot is an AI pair programmer that generates code suggestions and solutions based on your prompts. Prompt engineering is the practice of writing clear instructions for a large language model (LLM). By crafting specific prompts, you can steer the model towards the desired output. This article explores various prompting techniques to maximize the effectiveness of GitHub Copilot, specifically for writing Python code.

1. Zero-Shot Prompting
   
   Description: This technique provides minimal context and relies on the model's pre-trained knowledge to complete the task. Think of it as giving a broad instruction like "perform this action" without specifying the details.
   
   Use Case: Zero-shot prompting is ideal for simple tasks where the model's internal knowledge base is sufficient.
   
   Example:
   Prompt: Write a Python program that greets the user by name and asks how they are doing.
   Output: The model constructs a greeting message by combining a string with the retrieved name using f-strings (formatted string literals).
   
   Here is an example of the generated code:
   
   

   def greet_user():
      name = input("Please enter your name: ")
      print(f"Hello, {name}! How are you doing today?")
   
   greet_user()
   

   
   
2. One-Shot Prompting:
   
   Description: Here you would give a single instruction or example to nudge the model in a specific direction, for example, giving a basic outline or a starting point for the model to build upon.
   
   Use Case: One-shot prompting is useful when you want the model to perform a single task, like generating code with a particular functionality.
   
   Example:
   Prompt: Write a Python function that takes a list of numbers and returns the average.
   Output: The model generates a Python function using a loop or list comprehension to calculate the average.
   
   Here is an example of the generated code:
   
   

   def calculate_average(numbers):
      if not numbers:  # Check if the list is empty
         return 0
      total = sum(numbers)  # Sum up all the numbers in the list
      count = len(numbers)  # Count the number of elements in the list
      average = total / count  # Calculate the average
      return average
   
   # Example usage:
   numbers = [1, 2, 3, 4, 5]
   print("The average is: {calculate_average(numbers)}")
   

   
   
3. Few-Shot Prompting:
   
   Description: This approach provides a few examples or instructions to further refine the desired output. Think of it as a small set of examples to guide the large language model (LLM) towards the desired output.
   
   Use Case: Few-shot prompting is valuable for tasks requiring a higher degree of accuracy or specificity.
   
   Example:
   Prompt: Create a Python function named remove_duplicates_ordered that takes a list as input and returns a new list containing only the unique elements from the original list. The function should preserve the order of the first occurrence of each element.
   Output:
   
   

   def remove_duplicates_ordered(data):
      """Removes duplicate elements from a list, keeping the first occurrence."""
      seen = set()   # Store seen elements to check for duplicates. 
      unique_data = []
   
      for item in data:
         if item not in seen:
            unique_data.append(item)
            seen.add(item)
   
       return unique_data
   

   
   etc.
   
   
4. Chain of Thought Prompting:
   
   Description: This technique encourages the model to explain its reasoning process step-by-step. Imagine asking the model to "show its work" while completing a task.
   
   Use Case: Chain of thought prompting is beneficial for debugging the model's decision-making process and understanding its internal reasoning. This can be particularly useful in areas like machine learning where interpretability is crucial.
   
   Example: Write a Python function that checks if a given string is a palindrome (reads the same backward as forward).
   
   Prompt:

• First, define a function called "is_palindrome" that takes a string as input.
• Next, convert the string to lowercase to avoid case sensitivity.
• Then, iterate through the string and compare the characters at the beginning and end.
• If they match, move inwards one character at a time and repeat the comparison.
• If all characters match their counterparts, return True (the string is a palindrome). Otherwise, return False.

Output:



def is_palindrome(text):

   """This function checks if a string is a palindrome."""
   text = text.lower()  # Convert the string to lowercase 
   left, right = 0, len(text) - 1
   while left < right:
      if text[left] != text[right]:
         return False
      left += 1
      right -= 1
  return True
  

GitHub Copilot Training

Our GitHub Copilot course teaches attendees the skills needed in prompt engineering to communicate effectively with AI, optimizing the generation of code snippets, explanations, and solutions.

Let us customize the perfect GitHub Copilot training for your team, using the programming language of your choice, including, Python, Java, C++, Ruby, Go, JavaScript, C#, and more.

Browse all our Generative AI training courses for role-based training, including developers and business users.