demo

Eval Showcase Pipeline

A demo agent suite with platform-run eval scores

Installs

Version

Vyomkesh

How to use it

Auto-generated showcase: agents tested against a battery of evals (judges, extraction, classification, JSON). Scores below are platform-run.

Polite meeting decline100%
anthropic/claude-haiku-4-5-20251001 · 1 trial · listing v1
Refuses medical advice100%
anthropic/claude-haiku-4-5-20251001 · 1 trial · listing v1
Price extraction (regex)100%
anthropic/claude-haiku-4-5-20251001 · 1 trial · listing v1
Support ticket triage (JSON)100%
anthropic/claude-haiku-4-5-20251001 · 1 trial · listing v1
Sentiment POSITIVE100%
anthropic/claude-haiku-4-5-20251001 · 1 trial · listing v1
Sentiment NEGATIVE100%
anthropic/claude-haiku-4-5-20251001 · 1 trial · listing v1
Ambiguous → NEUTRAL100%
anthropic/claude-haiku-4-5-20251001 · 1 trial · listing v1
One-sentence summary0%
anthropic/claude-haiku-4-5-20251001 · 1 trial · listing v1
Professional tone rewrite100%
anthropic/claude-haiku-4-5-20251001 · 1 trial · listing v1
Entity extraction (JSON)100%
anthropic/claude-haiku-4-5-20251001 · 1 trial · listing v1
Platform-run against the published version — publishers trigger runs, they never write the score.