Roger Yang

duplicate of https://github.com/Arize-ai/phoenix/issues/5082 (currently blocked)

One thing we can try is to customize the Pydantic serialization and deserialization process specifically to and from the database, so that the function parameters are stored as a (json)...

Thank you for bringing this issue to our attention. As discussed offline the recommendation right now is to upgrade from Phoenix 7.12.2 to Phoenix 8.26.2+ in order to take advantage...

Thank you for the suggestion. We'll be tracking this in our experiment enhancement milestone.

Thanks for this PR @daavoo. I think the current monkey-patching approach might run into problems when users create their own callbacks (as they're encouraged to do by the docs [here](https://google.github.io/adk-docs/callbacks/))....

@daavoo if you don't mind, i think we can keep this PR open as we update the approach here given what we have so far

> Is there a retention limit on traces in Phoenix? I just noticed that my projects no longer have any traces except for my most recent project. Currently there is...

FWIW, below is my understanding of this debate. ## If we want `NULL` to equal `''` (empty string) Then use a _single_ index with `COALESCE`, because `COALESCE(NULL, '')` equals `COALESCE('',...

Thank you for filing this issue. Because this actually relies on LangChain's internal plumbing for trace_id propagation, it indicates a difference with how trace_ids are propagated internally between the sync...

Thanks for the additional info. Can you try passing along `config` in every node? For example, in the answer_node you could do something like below. ```python async def answer_node(state: MessagesState,...