LLM Eval Frameworks

Why it matters

Eval frameworks shape iteration. Understanding shapes tooling.

The architecture

Define datasets + metrics.

Run against models.

Track scores + regressions.

How it works end to end

RAGAS: RAG-specific.

TruLens: application-level.

promptfoo: prompt A/B.

LangSmith + Braintrust + Langfuse: eval + observability combined.