Skip to content

Similar Issues

Surfaces related issues from your repository using vector search across issue titles and bodies.

/similar_issue

MergeMate indexes all past issues once, then retrieves the closest matches whenever you call the tool — handy for spotting duplicates or finding earlier discussions.

Trigger it from the CLI:

python -m mergemate.cli --issue_url https://github.com/owner/repo/issues/42 similar_issue

Vector database options

Choose your backend in the config:

Database Best for
LanceDB (default) Zero-config, embedded — works out of the box.
Pinecone Managed, scales well for large repos.
Qdrant Self-hosted or cloud, open-source.

Pinecone

# .secrets.toml
[pinecone]
api_key = "..."
environment = "..."

# configuration.toml
[pr_similar_issue]
vectordb = "pinecone"

Qdrant

# .secrets.toml
[qdrant]
url = "https://your-instance.qdrant.io"
api_key = "..."

# configuration.toml
[pr_similar_issue]
vectordb = "qdrant"

Get a free Qdrant instance at cloud.qdrant.io.

Automation

Add /similar_issue to the pr_commands list to run it automatically when a new issue is opened.

Tips

  • Index happens once. The first call builds the vector index from all existing issues. Subsequent calls are fast.
  • Use it for triage. If a similar issue is already in progress, link them and save duplicate work.
  • Qdrant or Pinecone make sense for repos with thousands of issues; for smaller repos, LanceDB is the simplest path.