DOC: <Issue related to /evaluation/how_to_guides/evaluation/fetch_perf_metrics_experiment>

[majorgilles]

Hello, in pratice, when we do

results = evaluate(
  lambda inputs: "Hello " + inputs["input"],
  data=dataset_name,
  evaluators=[foo_label],
  experiment_prefix="Hello",
)

resp = client.read_project(project_name=results.experiment_name, include_stats=True)

we get a not found error. If we use the evaluators project_name (a project name we are actually able to find under Tracing projects in the LangSmith UI), we do get results. What is happening here? Is the documentation outdated?

[hinthornw]

Thank you for reporting!

Running that code gives a "not found" error in the "read_project" call?

I just ran it again and it worked as expected:

import langsmith as ls

client = ls.Client()

dataset_name = "Evaluate Examples"


def foo_label(run, example):
    return {"score": 1}


results = ls.evaluate(
    lambda inputs: "Hello " + str(inputs),
    data=dataset_name,
    evaluators=[foo_label],
    experiment_prefix="Hello",
)

resp = client.read_project(project_name=results.experiment_name, include_stats=True)

How are you initializing your client?