feat: phoenix client get_spans_dataframe() and query_spans()
#2151
Conversation
dosubot
bot
added
the
size:XL
RogerHYang
changed the title
get_spans_dataframe() and query_spans()
src/phoenix/server/query_handlers.py
Outdated
| @@ -0,0 +1,127 @@ | |||
| import asyncio | |||
There was a problem hiding this comment.
Nit, these files are many times nested under routes in traditional REST https://github.com/Arize-ai/openinference/blob/main/python/examples/llama-index/backend/app/api/routers/chat.py
There was a problem hiding this comment.
I see. Thanks for the tip.
src/phoenix/server/query_handlers.py
Outdated
| ) | ||
|
|
||
|
|
||
| class QuerySpansHandler(HTTPEndpoint): |
There was a problem hiding this comment.
Can you consolidate these two? It seems like they could be just /v1/spans?format=xyz
There was a problem hiding this comment.
I'll give it some thought. But probably will do this in a follow-up PR for refactoring
src/phoenix/server/query_handlers.py
Outdated
| traces: Traces | ||
| evals: Optional[Evals] = None | ||
|
|
||
| async def post(self, request: Request) -> Response: |
There was a problem hiding this comment.
Not blocking but POST is not the most standard for REST. Getting something is usually "GET" - I know we use graphql which breaks that convention but it might make sense to stick with conventions in this case
There was a problem hiding this comment.
There was a problem hiding this comment.
Sounds good. In the interest of time, i'll get this deployed first. Then I'll enumerate all the refactoring tasks in a follow-up PR, because I need time to understand them better myself, but those changes don't impact the user.
src/phoenix/session/client.py
Outdated
| logger = logging.getLogger(__name__) | ||
|
|
||
|
|
||
| class Client: |
There was a problem hiding this comment.
Don't we need a shared protocol with Session so this can be say this complies with all the session.evaluation helpers?
There was a problem hiding this comment.
That's a good idea.
resolves #2091
resolves #2136
resolves #2137
Screenshot shows the client in a console returning dataframes from Phoenix running in the background.