Skip to main content

Arize AI Community Icon

Home
Events

Arize AX Releases
Arize AX Support
Arize News
Arize Observe
Discussions
Introductions
Phoenix Support

Powered by Tightknit

Compatibility Layer Discussion for Llama-Index in Phoenix | Arize AI Community

Phoenix Support

Phoenix Support

Compatibility Layer Discussion for Llama-Index in Phoenix

·Dec 11, 2024 08:27 PM

Hey Rusty - we've thought about having that compatibility layer with llama-index but just haven't got to it since we have our own context managers to do that https://docs.arize.com/phoenix/tracing/how-to-tracing/customize-spans#using_attributes

6 comments

· Sorted by Oldest

RunLLM
APP
·
To add custom attributes to spans in Phoenix, you can use the OpenTelemetry (OTEL) framework, which Phoenix supports. This allows you to inject metadata into spans, such as user-specific data or other contextual information. You can achieve this by modifying the span creation or processing logic to include additional metadata. This is a common practice in distributed tracing and is supported by Phoenix's integration with OpenInference and OTEL.
If you need more detailed guidance on implementing this, please let me know.
Sources: - Phoenix Tracing with LlamaIndex
💡 Hint: Mention RunLLM in the thread for followups.
Mikyo
·
This readme might also be helpful. https://github.com/Arize-ai/openinference/tree/main/python/openinference-instrumentation
Mikyo
·
But happy to also support llama-index tags if that's the desired ergonomics. They really are 1-to-1. We were talking with Logan about it recently.
😄1
Rusty
·
Thanks. Yeah, I've been trying to follow the llama index instrumentation format, so we can layer in other consumers as well. It's probably still too early to say what's better or not though.
Rusty
·
I can use the OpenInference methods in the meantime 👍🏻
Mikyo
·
Cool cool, yeah it should be low lift to support so let us know.
😎1