Skip to content

Dashboard Guide

The PaperTrail dashboard is a self-contained interactive HTML file for exploring your team's paper landscape. No server required — just open in any browser.

Live Demo: Koo Lab Dashboard

Views

The dashboard has four main views, accessible from the top navigation bar.

Map View

The default view — a canvas-based 2D scatter plot where each dot is a paper.

Map View

Key features:

Region What it does
Scatter plot (center) Papers plotted by semantic similarity. Nearby papers discuss similar topics.
Topic labels Hierarchical cluster labels with colored pill outlines. Zoom in to see subtopics.
Topic lines Lines connecting related topics (configurable thickness, opacity, curve, color).
Channel filter (left sidebar) Toggle channels on/off. Deselected papers fade to ghost dots.
Projection buttons Switch between UMAP (default), t-SNE, and PCA projections.
Color By Cluster (default), Channel, Year, Citations, Engagement, Density, Contributor, Journal.
Scale By Engagement (default), Citations, Year, Uniform. Controls dot size.
KDE Background Density heatmap with configurable bandwidth, opacity, and palette.
Topic Lines Toggle and configure connection lines between topic clusters.
Time slider (bottom) Scrub through time or press play for chronological animation. Papers fade in smoothly.
Tool buttons (right) Pan, lasso select, rectangle select, reset zoom, save as PNG.

Interactions:

  • Scroll to zoom in/out
  • Click + drag to pan
  • Click a dot to see paper details
  • Click a topic label for topic info popup (paper count, top contributors, top papers)
  • Lasso/rectangle to select multiple papers
  • Hover over topics for paper count tooltip

3D View

WebGL-powered 3D scatter plot using Three.js.

3D View

  • Rotate: Click + drag
  • Zoom: Scroll
  • Click: Select paper
  • Same color modes as Map view

Table View

Sortable, filterable spreadsheet of all papers.

Table View

Columns: Title, Year, Journal, Authors, Channels, Shared By, Date Shared, Citations, Engagement, Cluster.

  • Click column header to sort ascending/descending
  • Filter row below headers — type to filter (supports >5, >=10 for numeric columns)
  • Click row to see paper detail panel
  • Export CSV / Export XLSX buttons in top-right
  • Slack icon links directly to the original Slack message
  • Paper icon links to the paper URL

Leaderboard View

Rankings and statistics about your paper collection.

Leaderboard View

  • Top Contributors — Who shares the most papers, with contribution heatmap
  • Most Engaged — Papers with the most reactions + replies
  • Most Cited — Highest citation count papers
  • Topic Breakdown — Cluster sizes and proportions
  • Channel Stats — Papers per channel with bar charts

AI Agent

Click Ask AI in the top-right to open the AI-powered search bar.

What you can ask:

  • "What are the most cited papers about protein design?" — searches papers with tool use
  • "Who contributes the most to genomics?" — contributor analysis
  • "Compare #papers-dl and #papers-genomics" — channel comparison
  • "What's trending in the last 6 months?" — trend analysis
  • @Peter Koo or #papers-dl — autocomplete for people and channels

Configuration (click the gear icon inside Ask AI):

  • Service: HuggingFace (free, default), Anthropic (Claude), OpenAI
  • Model: Qwen 3 8B (default), plus 8 other models
  • Reasoning traces: Optional toggle to show model thinking
  • Chat history: Downloadable in JSON or CSV format

Click Search to open the semantic search bar. Type a query and papers are ranked by content similarity — not keyword matching. Results are highlighted and sized by relevance.

Example queries:

  • single cell RNA sequencing
  • attention mechanism transformer
  • protein folding prediction

Keyboard Shortcuts

Key Action
1 / 2 / 3 Switch projection (UMAP/t-SNE/PCA)
V Pan mode
L Lasso select mode
R Rectangle select mode
/ Open AI agent
? Show all shortcuts
Esc Close panels
Arrow keys Switch views

URL Hash State

The dashboard saves your current view state in the URL hash. Share URLs like:

dashboard.html#p=umap&c=cluster&s=engagement&z=1.5

Parameters: p (projection), c (color mode), s (scale mode), z (zoom level).

Building the Dashboard

From CLI

papertrail build papers_final.json -o dashboard.html --title "My Lab"

From Python

from papertrail.preview import build_preview

build_preview(papers, output_path="dashboard.html", title="My Lab")

Automated Pipeline

The GitHub Actions pipeline automatically builds and deploys the dashboard weekly:

  1. Scrapes all configured Slack channels
  2. Enriches metadata via OpenAlex/Crossref/bioRxiv APIs
  3. Computes embeddings and hierarchical clusters
  4. Builds dashboard and deploys to GitHub Pages

See Deployment Guide for setup instructions.

Deployment

The dashboard is deployed to GitHub Pages at:

https://<username>.github.io/PaperTrail/<workspace>/

Where <workspace> is derived from your Slack workspace URL in config.yml. For example, koolab.slack.com deploys to /koolab/.

The old /dashboard/ path redirects automatically.