Integrating Arize Phoenix with BentoML and LangChain Issues

·Mar 23, 2024 04:05 AM

Hi Team, I'm integrate arize phoenix into bentoml and langchain. The auto instrumentation is not working for langchain. Can advise?

from phoenix.trace.langchain import LangChainInstrumentor
LangChainInstrumentor().instrument()
bento_model = load_bento_model()
itinerary_gen_runner = bento_model.to_runner()

svc = bentoml.Service("model", runners=[itinerary_gen_runner])

6 comments

· Sorted by Oldest

Mikyo
·
Hi Legend H. I’m not super familiar with BentoML so you might walk me through your deployment a bit. Are you in a notebook or using a phoenix image?
Legend H.
·
I'm using phoenix with docker image and I am able to see the interface in localhost:6006, I can't see their tracing being recorded when I fire langchain chain inside the service api
Mikyo
·
It’s a bit difficult to tell what’s wrong from your code but I’d suggest trying migrating to OpenInference as described here https://docs.arize.com/phoenix/tracing/how-to-tracing/trace-a-deployed-app
Mikyo
·
Let me know how it goes
Legend H.
·
Thanks a lot, your doc has inspired me to resolve the issue, bentoml has tracing export config can be used to point to phoenix directly
1
Mikyo
·
So glad to hear!

Mikyo
·
Hi Legend H. I’m not super familiar with BentoML so you might walk me through your deployment a bit. Are you in a notebook or using a phoenix image?
Legend H.
·
I'm using phoenix with docker image and I am able to see the interface in localhost:6006, I can't see their tracing being recorded when I fire langchain chain inside the service api
Mikyo
·
It’s a bit difficult to tell what’s wrong from your code but I’d suggest trying migrating to OpenInference as described here https://docs.arize.com/phoenix/tracing/how-to-tracing/trace-a-deployed-app
Mikyo
·
Let me know how it goes
Legend H.
·
Thanks a lot, your doc has inspired me to resolve the issue, bentoml has tracing export config can be used to point to phoenix directly
1
Mikyo
·
So glad to hear!