Manual Instrumentation Examples for Retrievers and Rerankers

Hello is there examples about manual instrumentation for a Retriever and a Reranker?

· Sorted by Oldest

Hello is there examples about manual instrumentation for a Retriever and a Reranker?

· Sorted by Oldest

Dustin N.
·
Hi Remi, Thanks for your interest in Phoenix! We don’t have a too many examples on manual instrumentation, they’re mainly focused on our autoinstrumentors. What are you trying to instrument?
Rémi C.
·
RAGatouille
Rémi C.
·
What would the function look like? It's very hard to know what the expected signature of the spans are, is there a place where I can find what the expected types for the attributes are?
Dustin N.
·
yeah! You'd have to refer to OpenTelemetry examples for manual instrumentation, but as long as your spans follow our [semantic conventions](https://github.com/Arize-ai/openinference/tree/main/python/openinference-semantic-conventions) They will render usefully on Phoenix
Xander S.
·
Rémi C. Here's our manual instrumentation example (does not include re-rankers). You might also find it helpful to look at how re-rankers are instrumented in LlamaIndex here. If you do a full-text search for "reranker", you'll see the places it is used.
Rémi C.
·
Awesme! So with this I'm all set 🙂
🫡3
Rémi C.
·
I have to admit I really struggle with understanding the code from the manual instrumentation What is the non-fancy version supposed to look like? Is it a list of strings with alternating role_string and message_string?

Dustin N.
·
Hi Remi, Thanks for your interest in Phoenix! We don’t have a too many examples on manual instrumentation, they’re mainly focused on our autoinstrumentors. What are you trying to instrument?
Rémi C.
·
RAGatouille
Rémi C.
·
What would the function look like? It's very hard to know what the expected signature of the spans are, is there a place where I can find what the expected types for the attributes are?
Dustin N.
·
yeah! You'd have to refer to OpenTelemetry examples for manual instrumentation, but as long as your spans follow our [semantic conventions](https://github.com/Arize-ai/openinference/tree/main/python/openinference-semantic-conventions) They will render usefully on Phoenix
Xander S.
·
Rémi C. Here's our manual instrumentation example (does not include re-rankers). You might also find it helpful to look at how re-rankers are instrumented in LlamaIndex here. If you do a full-text search for "reranker", you'll see the places it is used.
Rémi C.
·
Awesme! So with this I'm all set 🙂
🫡3
Rémi C.
·
I have to admit I really struggle with understanding the code from the manual instrumentation What is the non-fancy version supposed to look like? Is it a list of strings with alternating role_string and message_string?