polym-eval-db

Name: polym-eval-db
Availability: InStock
Author: byteplus-sa

Import image and video evaluation reports into a local SQLite database.

8|Updated May 13, 2026

Search Tags:#evaluation #model comparison #sqlite #win rate analysis #elo ranking #image report parsing #video gsb import

Ready to run

npx skills add https://github.com/byteplus-sa/polym --skill polym-eval-db

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: polym-eval-db
Source: https://github.com/byteplus-sa/polym/tree/main/skills/polym-eval-db
Command: npx skills add https://github.com/byteplus-sa/polym --skill polym-eval-db

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires pymysql, and includes scripts (resource) components.

What problem does it solve?

This Skill removes the friction of manually organizing and comparing large volumes of image and video evaluation results by storing them in a persistent SQLite database and enabling fast analysis.

Core Features & Use Cases

Import evaluation outputs into SQLite: ingest image eval HTML reports and video eval results into a local eval.db.
Query win rates, dimensions, and examples: compute overall statistics, per-dimension breakdowns, and retrieve example comparisons that match filters.
Elo ranking with caching: calculate and cache Elo scores for overall or specific dimensions across imported report sets.

Quick Start

Run import_report.py on an evaluation HTML URL, then run query.py stats to see overall win/loss/tie for the imported report.

Frequently Asked Questions about polym-eval-db

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

FAQPage Schema

How do I import image evaluation HTML reports into SQLite for model comparison?▼

The Skill ingests image evaluation HTML reports into SQLite by running the import_report.py script on the URL. This stores pairwise comparison data in a local eval.db for analysis.

Can I calculate Elo rankings for specific evaluation dimensions from imported report sets?▼

Yes, the Skill calculates and caches Elo rankings for overall or specific dimensions across imported report sets. It computes Elo scores from pairwise evaluation data stored in the local SQLite database.

What is the best way to query win rates and per-dimension breakdowns for video eval results?▼

The best way to query win rates and per-dimension breakdowns is running the query.py stats CLI script against the local eval.db. It computes overall win/loss/tie statistics and dimension-specific results from imported video comparisons.

Does the video evaluation import support pulling comparisons directly from a MySQL database?▼

Yes, the video evaluation import supports pulling comparisons directly from a MySQL database using the pymysql dependency. It transfers video GSB comparisons into the local SQLite database for unified analysis.

How do I resolve the database storage path when running evaluation queries?▼

You resolve the database storage path by configuring the EVAL_DB_PATH environment variable. The CLI scripts use this variable for deterministic SQLite storage, with an optional bootstrap download of a default database.