community-leaderboard

Generates and posts a daily pixel-art community model leaderboard image to Discord.

5.0k|1.0k|Updated Apr 15, 2021
One-click install
npx skills add https://github.com/pollinations/pollinations --skill community-leaderboard
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: community-leaderboard
Source: https://github.com/pollinations/pollinations/tree/main/.claude/skills/community-leaderboard
Command: npx skills add https://github.com/pollinations/pollinations --skill community-leaderboard

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It automates the daily production of a ranked community-model leaderboard image, eliminating manual data querying, chart design, rendering, and Discord posting for the Pollinations community.

Core Features & Use Cases

  • Automated Leaderboard Generation: Queries Tinybird for tokens served per model in the last 24 hours and builds a ranked pixel-art HTML board with speed panels and stability badges.
  • Headless Chromium Rendering: Screenshots the HTML via the chromium snap with sandbox workarounds, then crops whitespace using a python3-pil script.
  • Direct Upload and Posting: Uploads the image to the Pollinations media API and outputs ready-to-post Discord markdown.
  • Use Case: A maintainer asks to regenerate or redesign the daily leaderboard (e.g., adding a new stat panel or changing the palette) and previews it with a dry run before the scheduled post.

Quick Start

Regenerate the community model leaderboard with a dry run to preview the HTML before rendering and posting it to Discord.

Frequently Asked Questions about community-leaderboard

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I regenerate the community model leaderboard manually?

Run node build-leaderboard.mjs in operations/community-monitor/leaderboard with TB_TOKEN set. Use --dry-run to preview the HTML at /tmp/leaderboard-preview.html without rendering or uploading, or add POLLI_TOKEN for the full render and upload.

How is the community model leaderboard ranked?

Models are ranked by tokens served in the last 24 hours, not request count, so models doing real work outrank chatty low-token models. Speed is shown in a separate panel using median tokens per second, and stability appears as per-row badges.

Why does the leaderboard render with wrong fonts in chromium?

The chromium screenshot runs sandboxed with no network access to Google Fonts, so fonts must be embedded as base64 data-URI @font-face rules in fonts-embedded.css. If headings look non-blocky, the font embed is missing or broken.

Why does chromium fail to write the screenshot on the community-monitor box?

The chromium snap's AppArmor profile silently denies writes to dot-prefixed directories like /home/ubuntu/.foo, and --no-sandbox is required because snap confinement blocks the normal Chrome sandbox. Use a plain output directory name.

Can I run the leaderboard render on any machine?

No, the full render only works where chromium and python3-pil are installed, currently just the community-monitor EC2 box. Other machines can use --dry-run to validate the Tinybird query and ranking HTML without rendering.