Hacker News Reader

One issue you will encounter is that many of the LLM calls are hidden and nested below abstractions in tools like LlamaIndex. You will need to integrate with these libraries or users will not be able to get the granularity needed.

See LangFuse as an example. It is what I am using

View on HN

Quick question, what do you use such an observability platform for? Can't be the price as you get it straight from Openai's dev dashboard. So what else do you use Languse for?

I don't use OpenAI because it is an unreliable service, they don't know how to run production.

LangFuse supports all providers, allows for user and session level metrics, and is self hostable. Custom spans are really helpful as is the ability to take feedback on responses and build data sets from them.

How do you include your RAG queries and non-ai ops into your overall chat observability?

I'm somewhat of a mind to switch to an Open Telemetry based system, since that is the standard now

Hey I am building an open telemetry based open source tool for metrics and logs from LLM based apps. Give it a try at langtrace.ai

Will do, thanks for posting

One question I would have is why use a custom UI and storage setup, rather than the existing otel setup for the rest of the infra?

i.e. how much is being reinvented or duplicated from my grafana stack?

Reply on news.ycombinator.com