Run model evaluations with Velvet

Starting this week you can run evaluations with Velvet. Test numerical and sentiment-based metrics across models, settings, and prompts. We’re onboarding beta testers this week, and would love your feedback.

• Run evaluations on top of your datasets

• Test model endpoints, configurations, and prompts

• Set experiments as point-in-time or continuous

Learn how it works, and respond to this email if you’d like to try it out!

2 lines of code to get started

Velvet is a proxy to warehouse LLM requests to your database. Use the gateway to run your AI-powered features; Evaluations, fine-tuning, and continuous monitoring.

Read the docs →

Free up to 10k requests / month

Use Velvet free up to 10k requests per month. Get started quickly with a hosted database, or connect your own database.

Velvet pricing →