"No feedback details found" message when expanding individual evaluation metrics of a record (answer relevance, context relevance, answer faithfulness), with OTEL disabled.

[emmanouilvoitsidis]

I am running a RAG-based application and using standard evaluation metrics (answer relevance, context relevance, answer faithfulness). I have disabled OTEL as it was causing issue [https://github.com//discussions/2303]. All my metrics scores seem to default to 1, and when I try to expand a metric for a specific record, I see the following message on the dashboard: "No feedback details found".

New to TruLens and have tried to resolve this unsuccessfully for quite some time, so any help is greatly appreciated. Thanks.

[dosubot[bot]]

The "No feedback details found" message usually means the dashboard can't find or process the feedback call data for that metric. This often happens if the feedback call data is missing, empty, or not in the expected format—a list of dicts. In older versions (like 2.3.1), a known bug caused the dashboard to break or show missing details if the feedback call was a string or a single dict instead of a list. The fix is to wrap single dicts in a list before processing, as described in this discussion and here.

If all your metric scores default to 1, it's likely that the feedback functions aren't being executed or their results aren't being stored/retrieved correctly, so the dashboard falls back to a default value. This can also happen if there's a mismatch between OTEL and non-OTEL record formats: if you disabled OTEL but your database only contains OTEL-format records, the dashboard won't find the right data to display and will show missing details or default scores (source).

Here are some steps to resolve this:

1. Make sure you're running a recent version of TruLens, as this bug is fixed in newer releases.
2. If you must stay on an older version, patch your code by wrapping single dict feedback calls in a list before they're processed (see the code snippet in this discussion).
3. Double-check that the environment variable TRULENS_OTEL_TRACING is set to "0" (as a string) before any TruLens imports to fully disable OTEL (source).
4. Use the "refresh data" button in the dashboard after opening it to ensure the latest records and feedback are loaded (source).
5. Confirm that your feedback functions are actually running and storing results—check your logs or database for evidence of feedback execution.

If you still see only default scores or missing details after these steps, it's likely a data format or version mismatch. Upgrading is the most reliable fix.

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}