Research & Papers

ScholarScout turns 3-min pipeline wait into real-time SSE-driven pixel art game

Each paper spawns an in-game item based on a real SSE event from the backend.

Deep Dive

ScholarScout v1.5.3 features a pixel art owl running through a parallax forest during its 2-3 minute pipeline. The animation is driven by real SSE events: each paper found spawns a dot on screen, color-coded by source (white=arXiv, green=PubMed, purple=Crossref). Papers drip-feed at 600ms intervals from a queue. When the pipeline finishes, the owl celebrates. The release also adds Review Mode with paper clustering (k-means on embeddings, Jaccard fallback), per-cluster synthesis, and a paper freshness system that prioritizes least-used papers in the cache. Built with vanilla JS canvas and a 32×32 sprite sheet (12 frames).

Key Points
  • Real-time SSE events drive an in-progress game: each `paper_found` event spawns a colored dot in a parallax forest.
  • Review Mode uses k-means clustering on embeddings (Jaccard fallback) with per-cluster synthesis and cross-analysis.
  • Paper freshness system prioritizes least-used papers, auto-widens date ranges, and all thresholds are in config.yaml.

Why It Matters

Gamifying pipeline waits with real data feedback boosts user engagement while adding practical clustering and freshness features.