Langfuse Setup

Langfuse is an LLM engineering platform for monitoring, debugging, and optimizing AI pipelines. NQRust Analytics can send pipeline traces to Langfuse so you can see how each question is processed. Langfuse runs either self-hosted or in the cloud, and either option is supported here.

Before you start

Make sure that:

A Langfuse instance is up and running.
The langfuse_host value in config.yaml points at that instance.
If NQRust Analytics runs in Docker, the container can reach the Langfuse instance over the network.

Steps

Create a Langfuse account or instance

Sign up at Langfuse, or use your self-hosted deployment. If you are using Langfuse Cloud, note your data region: the EU region uses https://cloud.langfuse.com (the default) and the US region uses https://us.cloud.langfuse.com. Set langfuse_host accordingly.

Create an organization and project

Inside Langfuse, create an organization and a project to hold your traces.

Copy your API keys into the environment

Open your project's settings and copy the public and secret keys. Add them to the installer's environment file so the service can authenticate:

LANGFUSE_PUBLIC_KEY=your_public_key
LANGFUSE_SECRET_KEY=your_secret_key

These variables are already present in the installer's environment template, so you only need to supply the values.

Restart NQRust Analytics

Restart the service so it picks up the new keys and host.

Viewing traces

Once tracing is active, every question you ask in the Analytics UI produces a trace that appears in the Langfuse traces tab within a few seconds. Open a trace to examine each step of the pipeline — the prompts sent, the responses returned, and how long each stage took. This is the fastest way to understand the source of latency or incorrect answers.

Screenshot coming soon.