Auto-Fix agent
The auto-fix agent is an AI agent that can perform the following tasks in succession if it is provided with the issue context:
- Find the files that need update
- Make updates to the files based on the requirements of the issue. It adheres to the best coding practices and generates clean readable code
- Build the code in an Android or iOS environment and check if the code builds properly
- Creates and pushes the changes to a branch on Github
- Generates a PR Title and Description for the changes and submits a PR for review
- Runs the built apk or app on a cloud emulator or simulator respectively and puts the link as a comment on the PR
This can be triggered by just putting a comment @dashwave /fix on any Github issue. We are extending its support to Jira tickets and Slack as well in the coming weeks.
Currently, if the AI generated code has issues or logical errors, you can fix it right away in the cloud workspaces having Android Studio, VSCode, Debuggers, Cloud Emulators and a prompt based interface for code-less debugging and fixing.
Limitations:
- The current capabilities of the agent are limited to small issues only which demand not more than 6 file changes
- The issues should contain good contextual description of the bug
Coming soon:
- Ability to iterate on AI-generated solutions preserving the context to get to the perfect solution
- Ability to give additional context from external sources like Sentry logs, Crashlytics and Customer queries