Fix/embedding traces UI

[Vishesh-Paliwal]

This PR aims to solve issue #3573
Problem: Embedding traces incorrectly shown with chat UI in trace drawer
fix : added a check to ensure only span_type != embeddings are shown with chat UI

---

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you all sign our Contributor License Agreement before we can accept your contribution.
1 out of 2 committers have signed the CLA.

✅ Vishesh-Paliwal
❌ Vishesh Paliwal

---

Vishesh Paliwal seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[vercel]

Someone is attempting to deploy a commit to the agenta projects Team on Vercel.

A member of the Team first needs to authorize it.

[mmabrouk]

Hey @Vishesh-Paliwal

I am not sure the check that was done happened in the right place.

Can you please determine/describe the current flow (how and where we discover something is a chat span). If I remember correctly the conditions we had, looked for the span_type. Is that not the case?

Basically can you please update the PR with some research on the source of the issue and the reasoning behind where to put the solution

[Vishesh-Paliwal]

int the file : TraceDrawer/components/TraceContent/components/OverviewTabItem/index.tsx
I could see that OverviewTabItem component , solely relied on structure of content to detrmine if it should be rendered as chat .
inputs?.prompt &&
Array.isArray(inputs?.prompt) &&
inputs?.prompt.length > 0 &&
inputs?.prompt.every((item: any) => "role" in item)
something like this , Although span_type == chat might have helped determine chat directly instead of structure , I thought it could have been a decision based on if there is a possibilty wrong span_type being assigned , so there structure will help identify chat this way .

Coming to the problem at hand . because embeddings result also had structure similliar to chat detection logic , it was being rendered as chat . Hence I decided to just add one more check activeTrace?.span_type !== "embedding" so that embedding is treated differently .

let me know if this looks good ,
and if span type alone is sufficient to call a trace chat , then we can even remove the structural check part if needed .

Also One thing that comes to my mind seeing this kind of pattern is strategy pattern , we can also make common rendering logic available across web repo based on span types , if this is something you feel will be reused at many places now or later .

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no potential bugs to report.

View in Devin Review to see 2 additional findings.

[mmabrouk]

Thanks @Vishesh-Paliwal that makes total sense!