To create a Python script that fetches a Git Pull Request (PR) from a repository and then uses the OpenAI API to perform a code review, you’ll need to follow these steps:

  1. Set Up Your Environment: Make sure you have Python installed, along with the necessary libraries (openai, requests, or a Git library like PyGithub if you’re using GitHub).

  2. Get Your OpenAI and GitHub Credentials: You’ll need an API key from OpenAI and, if the repository is private or requires authentication, credentials for GitHub (like a personal access token).

  3. Install Required Libraries: 
    pip install openai PyGithub requests
  4. Write the Python Script:

This script provides a basic structure. Depending on the complexity of your PRs and the specific requirements of your code review process, you might need to add more sophisticated handling of the PR data.

from github import Github

# Replace these with your token and repository details
access_token = 'your_github_access_token'
org_name = 'organization_name'
repo_name = 'repository_name'
pr_number = 1  # The number of the PR you want to fetch

# Initialize GitHub client
g = Github(access_token)

# File to write the review results
review_file = 'pr_review.txt'

def get_review_comments(file_patch):
    # Placeholder for your review comments function
    # Replace this with actual logic to get review comments
    return "Review comments for the file."

try:
    # Fetch the organization
    org = g.get_organization(org_name)

    # Fetch the repository from the organization
    repo = org.get_repo(repo_name)

    # Get the PR
    pr = repo.get_pull(pr_number)

    with open(review_file, 'w') as file:
        file.write(f"Review for PR: {pr.title} (#{pr_number})\n")
        file.write(f"Created by: {pr.user.login}\n\n")

        files = pr.get_files()

        # Review each file in the PR
        for file_data in files:
            retries = 2
            while retries > 0:
                try:
                    review_comments = get_review_comments(file_data.patch)
                    file.write(f"File: {file_data.filename}\n")
                    file.write(f"Review Comments:\n{review_comments}\n\n")
                    break  # Break out of the loop if successful
                except Exception as e:
                    retries -= 1
                    print(f"Error fetching review for {file_data.filename}: {e}")
                    if retries == 0:
                        file.write(f"File: {file_data.filename}\n")
                        file.write("Review Comments: Error fetching review, skipped.\n\n")
                        print(f"Skipping review for {file_data.filename} after retries.")

        print(f"Review written to {review_file}")

except Exception as e:
    print(f"An error occurred: {e}")

In this script:

  • The while retries > 0 loop is used to attempt fetching review comments up to two times.
  • If an exception occurs in get_review_comments, it is caught, and the retry counter is decremented.
  • If the function fails twice (retries == 0), it skips to the next file and writes a message indicating the review was skipped for that file.
  • Replace get_review_comments(file_data.patch) with your actual function that returns review comments for the given file.

This approach ensures that each file gets up to two attempts for fetching review comments, and in case of persistent failures, the script will move on to the next file.

Important Considerations:

  • Rate Limits and API Costs: Be aware of the rate limits for both the GitHub and OpenAI APIs. Also, consider the cost associated with OpenAI API usage.
  • Security and Privacy: Ensure that you handle any sensitive data securely. Avoid printing or logging sensitive information.
  • Error Handling: Add robust error handling to deal with API failures, rate limits, or invalid responses.
  • Contextual Reviews: The quality of the review will depend on the context you provide to the AI. You might need to adjust how you format and send the PR data for better results.