AI-Powered Duplicate Detection in Issue Submission Flow

Preflight Checklist

• [x] I have searched existing requests and this feature hasn't been requested yet
• [x] This is a single feature request (not multiple features)

Problem Statement

Problem Statement

Users frequently submit duplicate feature requests and bug reports because:

• GitHub's keyword search doesn't understand semantic similarity
• Users don't know which search terms to use
• Existing issues may use different terminology
• The barrier to "just submit" is lower than thoroughly searching

This creates problems:

• User time wasted - Writing duplicates instead of contributing to existing discussions
• Maintainer time wasted - Triaging and closing duplicates
• Fragmented engagement - Votes and comments split across multiple duplicate issues

Proposed Solution

Add AI-powered duplicate detection that runs before issue submission:

Flow:

1. User writes issue
2. AI searches for similar issues (semantic search, not just keywords)
3. If similar issues found → Show matches + suggest upvoting
4. If no matches → Submit new issue

Alternative Solutions

_No response_

Priority

Critical - Blocking my work

Feature Category

CLI commands and flags

Use Case Example

When a user writes: "Add a progress bar showing context window usage"

The system would show:

🤖 I found similar requests: #5739: progress bar for context window used/remaining 👍 8 upvotes • Aug 14, 2025 Similarity: 95% match ⭐ #7111: Restore Time & Token Indicators + Add Context Usage 👍 12 upvotes • Sep 4, 2025 Similarity: 87% match What would you like to do? 1. 👍 Upvote #5739 (recommended) 2. 💬 Add your perspective as a comment 3. ✏️ Edit your request and search again 4. 📝 Submit anyway (I have a unique angle)

Implementation Approach

Semantic Search:

• Use embeddings to find similar issues (not just keyword matching)
• Compare against issue titles, descriptions, and top comments
• Show top 3 matches with similarity scores (>70% threshold)

User Control:

• User makes the final decision: upvote, comment, edit, or submit
• Non-blocking: adds ~5 seconds to submission workflow
• Can override if they believe their request is genuinely unique

Integration:
This needs to be built at the platform level (GitHub native feature) or into the repository's issue template so it works for all users automatically. A browser extension would require installation and miss most users.

Additional Context

Success Metrics

• Duplicate reduction: Target 60-80% fewer duplicate issues
• Engagement consolidation: More upvotes/comments on existing issues
• User satisfaction: Survey users who interact with the feature
• Maintainer time saved: Track hours spent on duplicate triage

Benefits

1. Reduces duplicate issues - Consolidates similar requests
2. Strengthens existing issues - Concentrates community support
3. Improves discoverability - Users find relevant discussions
4. Saves maintainer time - Less duplicate triage
5. Better UX - Contributors feel their voice is heard

Related Issues

• #7740 mentions "Maybe we should be asking Claude to tell us if an issue has already been reported?" but doesn't detail implementation
• The claude-code-action repo already does POST-submission duplicate detection; this would be PRE-submission

Alternative Approaches

Option A (Best): GitHub native feature in issue creation flow
Option B: Custom JavaScript in GitHub issue templates that searches as user types
Option C (Fallback): GitHub Action that comments on new issues with matches (reactive, not proactive)

Meta Note

This feature request would benefit from the feature it proposes! 😄 (I did manually search for duplicates before submitting this.)