Doppelgangers

Why it's ranked here

This is a focused tool with a credible end-to-end workflow: collect repository items, expose topical neighborhoods, and help a maintainer inspect likely overlap. Projection caching, resumable embedding output, date filters, and local-model support make repeated runs practical. The verdict is mixed-positive because clustering only narrows the search. A person still decides which items are duplicates and performs the cleanup.

What's good

It handles issues and pull requests together, while preserving clear visual distinctions for type, state, and selection. Fetching supports open, closed, or all items, plus absolute or relative creation cutoffs. Embedding input can include modified pull-request files, which supplies useful context beyond titles alone. Cached projections avoid needless recomputation when UMAP settings match, and interrupted embedding runs can resume by item URL.

Tradeoffs

The standard workflow requires Node.js 20 or newer, an authenticated GitHub CLI, and an OpenAI API key. Local embeddings are supported, but require a separate package and a compatible GGUF model path. The generated map is exploratory evidence, not duplicate detection with a confirmed verdict. UMAP uses randomness when projections are recomputed, so layouts can vary. Semantic search also embeds vectors into the HTML and increases its size.

How to use it well

Use it when a busy GitHub repository has enough issues or pull requests that manual scanning hides recurring themes. Start with open items or a recent creation window, inspect tight neighborhoods, then open or copy selected groups for deliberate triage. Tune neighborhood and distance settings only when the default map obscures useful structure. It does not replace maintainer judgment, close duplicates itself, or provide a broader issue-management workflow.

Observed

License

MIT

Primary language

TypeScript

Install surface

Global npm package

Interface

Command-line tool producing JSON, JSONL, projection cache, and standalone HTML output

Runtime platform

Node.js 20 or newer

Repository integration

Requires an authenticated GitHub CLI

Embedding backends

OpenAI embeddings, or an optional local GGUF model through node-llama-cpp

Module format

ECMAScript modules