Think about AI support for the development process

[sebastian-sommerfeld-io]

Relates to

• https://github.com/sebastian-sommerfeld-io/configs/issues/145
• https://github.com/sommerfeld-io/.github/issues/15

ChatGPT infos

There are several AI tools and techniques you can use to enhance your software development:

1. Code Review and Quality Analysis: Tools like Codacy or DeepSource use AI to analyze your code for potential issues, ensuring higher code quality and consistency.

2. Automated Testing: Tools like Testim or Mabl leverage AI to create and run automated tests, making it easier to catch bugs early.

3. Code Generation: Beyond GitHub Copilot, you can explore tools like TabNine or Kite, which offer AI-driven code completion and suggestions across multiple languages.

4. Natural Language Processing: You can integrate NLP models (like those provided by Hugging Face) to build features such as chatbots, text analysis, or sentiment analysis within your applications.

5. AI-powered Debugging: Tools like Sentry can help you identify, diagnose, and fix errors in your code by analyzing patterns and common issues.

6. Project Management: AI tools like ZenHub or Clubhouse can help in managing your projects more efficiently by predicting timelines, prioritizing tasks, and providing insights based on your project data.

Exploring these tools can help you streamline various aspects of your software development process. Which of these sounds interesting to you?

:zap: The issue with automated Code Reviews (due to available services)

Using ChatGPT with GitHub Actions like code-review-gpt or using the OpenAI API directly consumes credits for ChatGPT. I already pay for GitHub Copilot. I do not want to pay for a second AI solution. Plus (I assume) it is only a matter of time until Copilot is deeply integrated into GitHub and allows working with PRs and Issues and everything.

Additional non-AI logic

:warning: for Backlog refinement, not Code Reviews

• Each story gets a business value and sorry points (or similar) in a dedicated field (T-shirt sizes for both)
• Score for each T-Short size.
  • High score for high business value
  • High score for low story points
• Calc: business_value x story_points = ranking (which todo should come next? get some sort of prio / ranking)
• Run as GitHub actions workflow … assign and remove labels? Just 8 items as „next“ or similar???
  • Or Better: somehow visualize in a Board / Table based on the same metrics? So the stories that should be implemented next rise to the top of the view.

:thought_balloon: It might be better to not just prioritize issues based on score but to use an AI solution right from the start where the score is combined with some AI recommendations.

[sebastian-sommerfeld-io]

GitHub Copilot Chat

You can store business value and story points for each user story by using custom fields in GitHub Issues. Here’s how you can do it:

Custom Labels: Use labels to represent business value and story points. For example, you can create labels like business-value: high, story-points: 5.

Issue Templates: Create issue templates that include fields for business value and story points.

Querying: Use GitHub’s search and filtering capabilities to filter issues based on these labels.

For instance, to filter issues with high business value, you can search for label:"business-value: high".

Would you like more detailed steps or help with setting up specific filters?

---

When you use issue templates in GitHub, the inputs are stored as a single text block in the resulting issue. However, you can still make the fields queryable by using labels or structured text within the issue.

Approaches:

1. Labels:

   • Use labels to tag issues with specific story points and business values.
   • Example: label:"story-points: 5" and label:"business-value: high".

2. Structured Text:

   • Include structured text in the issue body and use search queries to filter.
   • Example: Use a consistent format like Story Points: 5 and Business Value: High.

To filter issues based on these fields, you can use GitHub's search functionality with appropriate queries. For example:

• To find issues with 5 story points: is:issue "Story Points: 5"
• To find issues with high business value: is:issue "Business Value: High"

Would you like to see an example of an issue template structure for this?

---

Example Issue Template

---
name: User Story
about: Create a new user story
title: "[User Story] "
labels: user-story
assignees: ''

---

**Story Points:** 

**Business Value:** 

**Description**

Provide a detailed description of the user story.

Querying Example

• To filter issues with 5 story points:

  is:issue "Story Points: 5"

• To filter issues with high business value:

  is:issue "Business Value: High"

Using labels to tag issues can make it easier to filter and manage them.

---

Label Naming

• "needs triage": This is a common term used to indicate that an issue needs to be reviewed and prioritized. It is widely understood in agile and software development practices.
• "needs refinement": This also makes sense and may be more specific in indicating that the issue needs further details or planning.

Both are correct, but "needs triage" is more commonly recognized in a broader context.

Marking Story Points and Business Value

• Using Labels:

  • Pros:
  • Easy to filter and search.
  • Can quickly see at a glance.
  • Cons:
  • May clutter the label list if you have many different values.

• As Part of the Issue Description:

  • Pros:
  • Keeps the label list clean.
  • Allows for more detailed information.
  • Cons:
  • Requires structured search queries to filter.
  • Not as immediately visible or standardized.

Recommendation

• Story Points: Use labels (e.g., story-points: 5). This allows you to quickly filter and see the complexity of issues at a glance.
• Business Value: Use part of the issue description. This provides room for more nuanced explanations and prioritization criteria.

Would you like specific examples or further assistance with setting up labels or issue templates?

[sebastian-sommerfeld-io]

needs-triage vs needs-refinement

Yes, that distinction makes sense:

• needs-triage: Indicates that the issue is ready for review and prioritization.
• needs-refinement: Indicates that the issue needs more details before it can be estimated or reviewed.

Using both labels can help clarify the stages of issue processing in your workflow.

[sebastian-sommerfeld-io]

Google Gemini Free Tier

https://ai.google.dev/docs

Request

curl \
  -H 'Content-Type: application/json' \
  -d '{"contents":[{"parts":[{"text":"Explain how AI works"}]}]}' \
  -X POST 'https://generativelanguage.googleapis.com/v1beta/models/gemini-1.5-flash-latest:generateContent?key=<API_TOKEN>'

Response

{
  "error": {
    "code": 400,
    "message": "Gemini API free tier is not available in your country. Please enable billing on your project in Google AI Studio.",
    "status": "FAILED_PRECONDITION"
  }
}