Hi! I'm working on a similar project here, but still early on so I haven't implemented much yet and don't have any working examples, but can share a bit about what I've found along the way. And I'm very curious to hear other's responses!
What are you looking to get from Langfuse? From what I've seen, there are two parts to the tool to deal with: the SDK in your application, and the UI/backend.
Langfuse's SDK allows for data export using OpenTelemetry. We have apps that are instrumented with this SDK already and we're looking to it to send OpenTelemetry traces to a collector and then onto dynatrace so that we can alert and visualize the data onto existing dashboards, etc. But there are several libraries that will do the same thing, most are built on the Gen AI OpenTelemetry attributes. One of those is OpenLLMetry, which Dynatrace does have docs for integration.
The UI/backend is where you have to do a lot more work. Langfuse seems to have more built into their UI to parse out spans and extract the data in a useful way. I think this is doable in Dynatrace, but you'd have to spend significant time on dashboards, openpipeline rules, and/or custom applications to get to where they are today.
Theres a new AI Observability App in the hub that looks quite promising, but its setup specifically for OpenLLMetry. If your app needs to use the Langfuse SDK, I'm not sure if you could use this app or not.
Observability, Rocket Mortgage
