feat: render inline preview of langfuse media (images/audio/video) within trace input/output json viewer

[marcklingen]

This is relevant when using e.g. the python decorator to log multi-modal traces in Langfuse. Currently these are rendered as an attachment that needs to be opened in a new tab.

[marcklingen]

Got another vote for this during todays community session

[sergedc]

This seems related with #4555 I would definitely upvote too. Currently I am using External media (URLs): ! [Alt text] (https://example.com/image.jpg) And even that does not fully renders: i.e. I must click on "Render Image" button on the frontend for the image to appear. By the way does anyone know if we can automate the rendering?

[marcklingen]

This seems related with #4555 I would definitely upvote too. Currently I am using External media (URLs): ! [Alt text] (https://example.com/image.jpg) And even that does not fully renders: i.e. I must click on "Render Image" button on the frontend for the image to appear. By the way does anyone know if we can automate the rendering?

Needing to select "render image" is the desired behavior as we didn’t want to risk loading external (potentially untrusted) URLs as images here. Would you prefer if this was configurable?

[sergedc]

I understand the risk but yes, it would make a world of difference to me if we could get a tick box "auto render images" in the project settings. Without auto rendering, reviewing the accuracy of multimodal traces takes 5 to 10 times as long, especially when there are 15 pictures in 1 prompt.

[marcklingen]

thanks, tracking this here: https://github.com/langfuse/langfuse/issues/5142

[mlcocdav]

Upvoting this as well! Is critical for us!

[github-actions[bot]]

This issue is stale because it has been open for 30 days with no activity.

[github-actions[bot]]

This issue was closed because it has been inactive for 14 days since being marked as stale. Please reopen if the issue persists.