Using Evals with Phoenix

Evals + Traces#

Running evals on traces and logging them to Phoenix.

Logging evals as annotations in Phoenix.

Evals + Experiments#

Info

All arize-phoenix-evals Evaluators are drop-in compatible with experiments.

Evals + Prompt Management (Python)#

If your evaluation prompt is versioned in Phoenix Prompt Management, you can fetch it with phoenix-client and convert it into an eval-ready PromptTemplate.

from phoenix.client import Clientfrom phoenix.evals import (    ClassificationEvaluator,    LLM,    phoenix_prompt_to_prompt_template,)client = Client(base_url="http://localhost:6006")prompt_version = client.prompts.get(prompt_identifier="test-prompt")prompt_template = phoenix_prompt_to_prompt_template(prompt_version)evaluator = ClassificationEvaluator(    name="response_quality",    llm=LLM(provider="anthropic", model="claude-sonnet-4-6"),    prompt_template=prompt_template,    choices={"good": 1.0, "bad": 0.0},)

This keeps your eval logic aligned with prompt versions managed in Phoenix while still using the standard arize-phoenix-evals API.