Triaging Issues & Creating Fix Sessions

Review discovered issues in a purpose-built triage interface, then dispatch AI agents to fix approved findings automatically.

Triage Mode Overview

Triage Mode is a widescreen two-column interface designed for reviewing issues at speed. The left panel shows a scrollable list of issues with severity indicators. The right panel shows the full detail view: source code with syntax highlighting, the AI's explanation, and action buttons.

For each issue, you can:

Approve — Mark the issue as valid and eligible for a fix session
Reject — Dismiss the issue as a false positive or not worth fixing
Skip — Move to the next issue without making a decision
Tag — Add custom tags for organization (e.g., sprint-42, critical, team-backend)

Keyboard Navigation

Triage Mode is built for keyboard-driven review. You can navigate through issues and take actions without touching the mouse:

Key	Action
j or ↓	Next issue
k or ↑	Previous issue
a	Approve current issue
r	Reject current issue
s	Skip current issue
t	Open tag editor
Esc	Exit triage mode

Tip: Combine a (approve) with automatic advance to fly through issues. Each action automatically advances to the next unreviewed issue.

Creating a Fix Session

After triaging, you will have a set of approved issues. To create a fix session:

Navigate to the Issues tab and filter by Approved status.
Select the issues you want to fix (use checkboxes or Shift-click to select a range).
Click the Create Fix Session button in the toolbar. This opens the fix configuration dialog.
Choose the AI backend for fixing (defaults to your discovery backend), optionally set a branch name, and click Start Fixing.

The fix session dispatches AI agents to implement solutions for each approved issue. Progress is tracked in real-time.

Human-in-the-Loop Intervention

Sometimes the AI agent needs guidance while implementing a fix. When this happens, the fix session pauses and an intervention banner appears. You can read the agent's question, type a response, and let it continue.

Animated demonstration of responding to an AI intervention during fixing — Responding to an intervention: read the question, type a response, and resume.

Note: Fix sessions create changes on a dedicated branch. You can review the diff, run tests, and merge when you are satisfied with the results. VibeRails never pushes to remote repositories automatically.

Reviewing Fix Results

After a fix session completes, you can review the results:

Status per issue — Fixed, Failed, or Skipped
Diff viewer — See exactly what changed for each fix
Git log — Each fix is committed separately with a descriptive message
Intervention log — Full history of any human-in-the-loop interactions

Warning: Always review AI-generated code changes before merging. While VibeRails verifies fixes against the original issue context, automated changes should be treated like any other code review.

End-to-End Summary

You have now completed the full VibeRails workflow:

Add a project — Register a local git repository
Run a review — AI discovers issues across your codebase
Triage issues — Approve, reject, or tag each finding
Fix issues — Dispatch AI agents to implement solutions

Explore the Features and Reference sections for in-depth documentation on every aspect of VibeRails.