Assign issue to an AI developer
Release notes
Writing software is time-consuming, and backlogs often grow faster than they can be burned down. By holistically integrating existing AI solutions, issue requirements could be met automatically.
Problem to solve
Software engineers needing to write code to meet the requirements of an issue is now likely unnecessary, but yet still required.
Existing AI solutions integrated into IDEs are isolated from team interaction and feedback. The "session" exists locally between a single AI solution and the developer. Mechanisms for the wider team to experience and join in on the communication between the AI solution and the developer is lacking.
Proposal
Combine the existing available AI solutions, to remove the need for software engineers to write code to meet the requirements of an issue.
- GitLab Duo can create a summary of the issue and comments which could be used as the requirements for code changes
- Code suggestions can be used to start meeting the requirements
- write unit and integration tests to codify the requirements and validate the code
- ensure security needs are met
- ensure high-level design requirements are met
- each change made as a small atomic commit that can be individually reviewed
- requirements that are too large may be automatically extracted into a separate issue
- A merge request is automatically created when all requirements for an issue are met, allowing the team to review and provide feedback.
- The comments on a merge request provide a feedback mechanism where Code Suggestions either make new commits to respond to the comments or first respond via a comment with code suggestions before actually making the changes * I'm not sure what this would look like if multiple files need to be changed
- Any git conflicts should be resolved automatically
I imagine that some form of Constraint Satisfaction Programming (CSP) solution would additionally be needed to complement Large Language Model (LLM) solutions to validate that issue requirements are sufficiently met.
Intended users
Anyone with the capability to validate that the code changes generated by the AI solutions are correct and meet the requirements of the issue.
- Parker (Product Manager)
- Delaney (Development Team Lead)
- Presley (Product Designer)
- Sasha (Software Developer)
- Priyanka (Platform Engineer)
- Simone (Software Engineer in Test)
- Dakota (Application Development Director)
- Amy (Application Security Engineer)
- Isaac (Infrastructure Engineer)
- [Alex (Security Operations Engineer)
- Cameron (Compliance Manager)
Feature Usage Metrics
- Improvement in user satisfaction metrics.
- Improvement in DORA metrics.
Does this feature require an audit event?
Existing audit features should be sufficient. There is no new functionality, besides AI solutions making code changes in response to issues, and comments on issues and merge requests.