Unable to view Scores under Evaluation tab

[saisankalpj]

Describe your question

Under Observability → Tracing, I can see the RAGAS scores being computed successfully.
I evaluated the RAGAS metrics using custom Python code and pushed the scores to Langfuse. The scores are attached as in image.

However, when I navigate to Evaluation → Scores, I do not see any records displaying these scores.
Similarly, under Scores → Analytics, no graphs are shown for metrics such as Answer Relevancy when I select the corresponding dataset run.

Currently, I am using tags and have created custom dashboards based on those tags, where the score visualizations are displayed correctly. However, the same score data does not appear in Scores → Analytics.
Could you please help me understand if there is any additional configuration required for custom RAGAS scores to appear in the Evaluation → Scores and Scores → Analytics sections?

Langfuse Cloud or Self-Hosted?

Self-Hosted

If Self-Hosted

v3.199.0

If Langfuse Cloud

No response

SDK and integration versions

No response

Pre-Submission Checklist

• I have checked for existing issues/discussions and consulted Langfuse AI.

[saisankalpj]

@inkeep.

Any suggestions

[inkeep-ai-assistant[bot]]

Based on your description, your scores are attached to traces correctly (visible in Tracing), but not appearing in Evaluation → Scores or Scores → Analytics. Here are the key things to check:

1. Verify Score Data Type

Score Analytics only supports numeric, categorical, and boolean scores. Text scores are not supported in Score Analytics because free-form text cannot be meaningfully aggregated or compared.(1)

When you push scores via the SDK, ensure you're specifying the correct data_type:

langfuse.create_score(
    name="answer_relevancy",
    value=0.9,
    trace_id="your_trace_id",
    data_type="NUMERIC",
)

(2)

2. Check Object Type Selection in Analytics

In Score Analytics, you need to select the correct object type (Traces, Observations, Sessions, or Dataset Run Items) that matches where your scores are attached.(1)

If you attached scores directly to traces, select "Traces" as the object type. If you're looking at dataset runs specifically, select "Dataset Run Items."(1)

3. Verify Time Range

Ensure your selected time range in Score Analytics includes the period when the scores were created. Use the date picker to set an appropriate range (e.g., "Past 90 days").(1)

4. Score Name Matching

When selecting a score in the Analytics dropdown, the score name must exactly match what you used when creating the score via the SDK.(1)

5. Dataset Run Scores

If you want scores to appear on dataset runs specifically, you need to link the score to the dataset run item, not just the trace. When scoring production traces, langfuse.create_score attaches the value to the trace you pass.(3)

For dataset experiments, scores attach to each item's trace and to the dataset run when using the experiment SDK.(3)

Recommended Troubleshooting Steps

1. Go to Scores → Analytics
2. Select your score name from the first dropdown
3. Set object type to Traces (since your scores are visible on traces)
4. Expand the time range to ensure it covers when scores were created
5. Check if the Statistics card shows any total count(1)

If scores still don't appear, verify in the Langfuse UI under Evaluation → Scores (the table view) that the scores are listed there first before checking Analytics.

📚 Sources:

1. Score Analytics
2. Scores
3. Run Ragas evaluations on Langfuse experiments and traces

---

Have another question? Just tag @inkeep.

[fsystemweb]

Hi, I ran into a very similar (maybe the same) issue. I fixed it by changing how evaluation is run.

In my case, the problem was that scores were being attached to each individual trace instead of to the experiment/dataset run itself. Once I stopped doing that and attached the score to the experiment, the issue went away. Worth checking your Langfuse setup, you might need to update it too.

The key change: when you run the experiment, pass an evaluators argument (not create_score). Each evaluator should be a function that returns an evaluation object containing your metric (e.g. name, value, comment), and Langfuse will handle attaching it correctly to the experiment run:

from langfuse import get_client

langfuse = get_client()

def my_evaluator(*, input, output, expected_output, **kwargs):
    return {
        "name": "accuracy",
        "value": 1 if output == expected_output else 0,
    }

result = langfuse.run_experiment(
    name="my-experiment",
    data=my_dataset,
    task=my_task,
    evaluators=[my_evaluator],  # <-- this is the fix, avoid create_score
)

print(result.format())

This is documented in Langfuse's guide on run-level evaluators that assess the full experiment results and compute aggregate metrics, which when run on Langfuse datasets get attached to the full dataset run for tracking overall experiment performance here: https://langfuse.com/faq/all/retrieve-experiment-scores

Hope that helps!

[CodeCLS]

Hi @saisankalpj, did @fsystemweb's answer help?

[saisankalpj]

Hi @CodeCLS
yes it worked and @fsystemweb Thank you for the solution