Skip to main content

Understanding TrainLoop Evals

This section provides in-depth explanations of TrainLoop Evals concepts, architecture, and design decisions. These guides help you understand the "why" behind TrainLoop Evals.

Core Concepts​

Architecture Overview​

Learn about the overall system architecture and how components work together.

Data Model​

Understand how TrainLoop Evals stores and processes evaluation data.

Evaluation Engine​

Deep dive into how metrics and suites are executed and results are generated.

LLM Judge System​

Explore the built-in AI-powered evaluation system for subjective metrics.

Design Principles​

Simplicity First​

How TrainLoop Evals prioritizes developer experience and ease of use.

Vendor Independence​

Why and how TrainLoop Evals avoids vendor lock-in.

Type Safety​

The benefits of type-safe evaluation code and how it's implemented.

Composability​

How the modular design enables reusable evaluation components.

Technical Deep Dives​

SDK Architecture​

How the multi-language SDKs work under the hood.

Data Collection Pipeline​

The journey from LLM call to stored event data.

Evaluation Execution​

How evaluations are discovered, executed, and results are stored.

Studio UI Architecture​

The technology stack and data flow in the visualization interface.

Coming Soon​

We're working on comprehensive explanations for each of these topics. Check back soon for detailed guides that help you understand the inner workings of TrainLoop Evals.

Questions?​

If you have specific questions about how TrainLoop Evals works, please: