ProjectBuild an AI-Powered Digital Bank with RAG: the explanation · Module 14: Evaluation: measuring quality before customers do · part 16 of 20
Part 16 · Module 14: Evaluation: measuring quality before customers do
Module 14: Evaluation: measuring quality before customers do
25 min read·
About 40 minutes · Needs: Modules 9 (retrieval), 10 (the agent), 11 (routing) and 12 (guardrails) · Code: backend/evals/run.py, backend/evals/metrics.py, backend/evals/judge.py, backend/evals/datasets/
Where this fits
Evaluation (blue) drives the real agent, retrieval, router and models, then reports to observability.
Evaluation calls the real retriever, router and agent graph with real models, then sends its scores to observability. It serves no customer; it decides whether a change may reach one.
The rest of this project is yours to keep
This project is bought on its own, once, and stays readable afterwards, including the parts added to it later.