Comment on page

Artificial Intelligence

Customise Released's language model
The advanced AI prompts of Released have undergone meticulous fine-tuning to produce release notes from an array of Jira issue descriptions. However, achieving a universally perfect prompt catered to every content type remains challenging.
To guarantee that the AI output meets your expectations, you have the option to personalize the prompt, molding it to suit your distinct content requirements.
Custom prompts are not a guarantee It's important to note that the resulting description might not strictly follow your input. Your personalized prompts serve as navigational tools for guiding Released's content creation in the desired direction, leading to reduced editing efforts. However, it's essential to understand that a custom prompt does not ensure an infallible or guaranteed result.

Base prompt

The base prompt is the initial instruction provided to the language model to guide its response. It sets the context for the AI's subsequent language generation. The quality and clarity of the base prompt significantly influence the AI's output. To incorporate broad instructions that apply to all types of issues, you can easily customize the base prompt according to your needs.

What makes a good custom prompt?

  • Instructive Language: Utilize custom prompts as instructional tools, employing metaphors to elucidate intricate ideas. Rather than using requests like "please incorporate metaphors" or "this changelog needs metaphors," focus on guiding language like "utilize metaphors to clarify complex concepts."
  • Negation Avoidance: Keep in mind that language models can struggle with negations. To enhance results, it's advisable to rephrase custom prompts. For instance, instead of "do not include details that aren't relevant," you're likely to achieve better outcomes with "avoid details that aren't relevant."

Issue specific prompts

Enhance your prompt by incorporating issue-specific instructions, customizing the desired response length, adjusting the tone of voice, or requesting the AI to provide additional context as needed.
Issue specific prompts provide you the flexibility to adopt a playful tone while crafting feature descriptions and maintain a professional tone when communicating security vulnerabilities.


The formatting option empowers you to tailor the length of the generated description. For instance, bug-fixes usually warrant succinct, one-sentence explanations, while major features can be elaborated upon with a couple of paragraphs.

Tone of voice

The Tone of Voice settings let you customize how Released writes. You can configure the tone of voice via the issue specific prompt settings. There, you'll see an option that looks something like this:
Incorporating a tone of voice will guide Released towards a particular writing style, yet it doesn't ensure that all your content will strictly adhere to that tone. For instance, opting for the Humorous tone will infuse elements of humor into the article, but Released can not guarantee that the output will be a comedic masterpiece.

Custom prompt

With the custom prompt option, you can go beyond the pre-configured choices to personalize formatting and tone of voice according to your preferences. Additionally, feel free to include any extra instructions you'd like.

Writing a good custom tone

Express custom tones of voice as adjectives.
  • Good: motivational, philosophical, and cheeky.
  • Bad: injecting humor, projecting authority, and emulating Shakespearean style.
Incorporating conjunctions (such as 'and') is acceptable, although brevity in your tone is preferable for optimal results.
  1. 1.
    Good: lively and engaging.
  2. 2.
    Bad: example is: lively, engaging, concise, inspirational, and authentically genuine.