Improving Trace Display: JSON to Human-Friendly Format

·Jan 31, 2025 03:05 PM

Another question: we instrument our app with openinference-instrumentation-bedrock = 0.1.13, and Phoenix is 7.8.1. The instrumentation successfully submits the traces. But, on the traces page, the input and output columns show JSON pieces instead of plain text. See the screenshot. On the second screenshot, you can see that the individual screen for this trace shows the output in a human friendly way. Maybe we could tweak some options to get the main screen show us nice view of the messages as well? Thanks!

3 comments

· Sorted by Oldest

RunLLM
APP
·
The issue you're experiencing with JSON pieces showing in the input and output columns instead of plain text is likely related to how the data is being rendered in the Phoenix UI. According to <https://github.com/Arize-ai/phoenix/issues/5707|this GitHub issue>, there are plans to make display values for span I/O more readable by using heuristics to display the most relevant parts of the JSON data. However, this feature might not be fully implemented in your current version. To address this, you can try updating to the latest version of Phoenix and the `openinference-instrumentation-bedrock` package, as improvements might have been made in newer releases. Additionally, you can manually parse and format the JSON data in your application before sending it to Phoenix, ensuring that only the relevant parts are included in the input and output fields. Would you like more detail on how to manually format the JSON data before sending it to Phoenix? Sources: - <https://github.com/Arize-ai/phoenix/issues/5707|this GitHub issue>
💡 Hint: Mention RunLLM in the thread for followups.
Roger Y.
·
Thank you for your feedback! This is on our radar and is captured on the ticket mentioned by RunLLM above. cc ticket author Xander S.
💙1
Xander S.
·
Thanks Anatoly S.! Definitely agree this can be improved