For some time now, we have been using OpenAI to process these articles. How does it actually work?
Determining the DevSquad Articles
Currently, we manage all DevSquad articles in Gitlab. The post itself is an issue in Gitlab. Whenever an article is presented internally, we trigger a webhook. This makes a request to our website.
On the website, there is code that reads out the content from Gitlab via an API call. We use Gitlab's GraphQL API to determine the title, content, and author and check whether the issue is confidential.
Utilizing OpenAI
Then we take our defined ChatGPT prompt, which is currently defined as follows:
11. Use markdown as output.
22. Change all existing image links which do not start with https from the original format and apply the target format using the following rule:
3 Existing link: 
4 Target format: 
53. Preserve all source code examples in the original format. Do not create new code examples.
64. Keep all absolute links in the original format unchanged.
75. Translate any German texts in the content to English.
86. Write the blog post in a technical blogger style, using a tone suitable for a tech-savvy audience.
97. Generate a YAML frontmatter that should be a yaml object with the following properties:
10- title: The title of the blog post in Title Case
11- slug: A short slug for the blog post
12- tags: A list of tags
13- category: A category from the following list: ["Blog","Agile","Geodata","AI","Mobile","Web"]
14- summary: A short summary of the content with maximum 285 characters
15- contents: A list of questions that get answered by this post
And we supplement the original article.
We evaluate the information obtained and use the Storyblok API to create a new page.
Informing the User
Finally, we again use the Gitlab API to create a comment and notify the author about the new article.