docs: add langfuse observability guide (#1047)

Co-authored-by: angiejones <jones.angie@gmail.com>
This commit is contained in:
Jannik Maierhöfer
2025-02-05 16:37:14 +01:00
committed by GitHub
parent fe4931910f
commit a04b528798

View File

@@ -0,0 +1,40 @@
---
description: Integrate Goose with Langfuse to observe performance
---
# Observability with Langfuse
This tutorial covers how to integrate Goose with Langfuse to monitor your Goose requests and understand how the agent is performing.
## What is Langfuse
[Langfuse](https://langfuse.com/) is an [open-source](https://github.com/langfuse/langfuse) LLM engineering platform that enables teams to collaboratively monitor, evaluate, and debug their LLM applications.
## Set up Langfuse
Sign up for Langfuse Cloud [here](https://cloud.langfuse.com) or self-host Langfuse [Docker Compose](https://langfuse.com/self-hosting/local) to get your Langfuse API keys.
## Configure Goose to Connect to Langfuse
Set the environment variables so that Goose (written in Rust) can connect to the Langfuse server.
```bash
export LANGFUSE_INIT_PROJECT_PUBLIC_KEY=pk-lf-...
export LANGFUSE_INIT_PROJECT_SECRET_KEY=sk-lf-...
export LANGFUSE_HOST=https://cloud.langfuse.com # EU data region 🇪🇺
# https://us.cloud.langfuse.com if you're using the US region 🇺🇸
# https://localhost:3000 if you're self-hosting
```
## Run Goose with Langfuse Integration
Now, you can run Goose and monitor your AI requests and actions through Langfuse.
With Goose running and the environment variables set, Langfuse will start capturing traces of your Goose activities.
_[Example trace (public) in Langfuse](https://cloud.langfuse.com/project/cloramnkj0002jz088vzn1ja4/traces/cea4ed38-0c44-4b0a-8c20-4b0b6b9e8d73?timestamp=2025-01-31T15%3A52%3A30.362Z&observation=7c8e5807-3c29-4c28-9c6f-7d7427be401f)_
![Goose trace in Langfuse](https://langfuse.com//images/docs/goose-integration/goose-example-trace.png)