Using LLM to generate markdown based off GitHub issues

generate.py

import os
import json

from dotenv import load_dotenv
from openai import AzureOpenAI
import requests

load_dotenv(".env", override=True)


# Use Azure OpenAI to generate a markdown for the issue
openai_client = AzureOpenAI(
    api_version="2023-07-01-preview",
    azure_endpoint=f"https://{os.getenv('AZURE_OPENAI_SERVICE')}.openai.azure.com",
    api_key=os.getenv("AZURE_OPENAI_API_KEY"),
)


def get_all_issues(repo_name):
    issues = requests.get(
        f"https://api.github.com/repos/microsoft/{repo_name}/issues"
    ).json()

    issue_numbers = [issue["number"] for issue in issues]
    return issue_numbers

def get_issue(issue_num):
    # Get the issue from the GitHub API
    issue = requests.get(
        f"https://api.github.com/repos/microsoft/AI-Chat-App-Hack/issues/{issue_num}"
    ).json()

    # Get the issue body
    issue_url = issue["html_url"]
    issue_body = issue["body"]
    return issue_url, issue_body


def extract_hack_properties(chat_completion):
    response_message = chat_completion.choices[0].message

    if response_message.tool_calls:
        for tool in response_message.tool_calls:
            if tool.type != "function":
                continue
            function = tool.function
            if function.name == "describe_hack":
                arg = json.loads(function.arguments)
                return arg
    return {}

def describe_hack(hack_content):
    tools = [
        {
            "type": "function",
            "function": {
                "name": "describe_hack",
                "description": "Describes a hackathon submission based on the provided content.",
                "parameters": {
                    "type": "object",
                    "properties": {
                        "name": {
                            "type": "string",
                            "description": "Name of the hackathon project, based on Project name heading",
                        },
                        "one_sentence_description": {
                            "type": "string",
                            "description": "One-sentence description of the hackathon project, based on Description heading, but worded in a way that will get people excited to learn more.",
                        },
                        "technology_used": {
                            "type": "string",
                            "description": "A comma-separated list of Azure services or other technologies used, which may be mentioned in the description - such as Azure AI Search, Azure OpenAI, Azure App Service, etc. If none are mentioned, don't list any.",
                        },
                        "video_url": {
                            "type": "string",
                            "description": "URL to the project video, based on Project video heading",
                        },
                        "repo_url": {
                            "type": "string",
                            "description": "URL to the project repository, based on Project Repository URL heading",
                        }
                    },
                    "required": ["search_query"],
                },
            },
        }
    ]

    chat_response = openai_client.chat.completions.create(
        # Azure Open AI takes the deployment name as the model name
        model=os.getenv("AZURE_OPENAI_DEPLOYMENT_NAME"),
        messages=[
            {
                "role": "system",
                "content": "You are a helpful assistant that is preparing copy for a blog post about an exciting hackathon to make AI chat apps using RAG and Azure services. When you describe the hacks, be very supportive and encouraging.",
            },
            {
                "role": "user",
                "content": f"Please describe following hackathon submission: {hack_content}",
            },
        ],
        tools=tools,
        temperature=0.3,
        max_tokens=400,
        n=1,
        stream=False,
    )
    return extract_hack_properties(chat_response)


def describe_all_issues(repo_name):
    hack_descriptions = []
    all_issues = get_all_issues(repo_name)
    for issue_number in all_issues:
        issue_url, issue_body = get_issue(issue_number)
        hack_properties = describe_hack(issue_body)
        hack_properties["submission_url"] = issue_url
        print(hack_properties)
        hack_descriptions.append(hack_properties)

    # write to file
    with open("hack_descriptions.json", "w") as f:
        json.dump(hack_descriptions, f, indent=2)

def generate_markdown():
    markdown_str = ""
    with open("hack_descriptions.json", "r") as f:
        hack_descriptions = json.load(f)
        markdown_str += "# Hackathon Submissions\n"
        for hack in hack_descriptions:
            markdown_str += f"## {hack['name']}\n"
            markdown_str += f"{hack['one_sentence_description']}\n\n"
            markdown_str += f"Technology used: {hack['technology_used']}\n\n"
            markdown_str += f"See: [Video]({hack['video_url']}) | [Repo]({hack['repo_url']}) | [Full Submission]({hack['submission_url']})\n\n"
        # save to file
        with open("hackathon_blog.md", "w") as f:
            f.write(markdown_str)