Gradio

Trace Gradio chat UIs in Muster via the @observe decorator and capture user feedback as scores.

Gradio is a Python library for building web UIs around ML models. Combine it with Muster (langfuse) to trace every chat turn and capture user thumbs-up/down feedback as scores.

Setup

%pip install gradio langfuse openai httpx

import os

os.environ["LANGFUSE_PUBLIC_KEY"] = "pk-lf-..."
os.environ["LANGFUSE_SECRET_KEY"] = "sk-lf-..."
os.environ["LANGFUSE_BASE_URL"] = "https://app.getmuster.io"
os.environ["OPENAI_API_KEY"] = "sk-..."

from langfuse import Langfuse
langfuse = Langfuse()

Implementation

Use the @observe() decorator on the chat handler:

from langfuse import observe
from langfuse.openai import openai

@observe()
async def create_response(prompt: str, history):
    response = openai.chat.completions.create(
        messages=history,
        model="gpt-4o-mini",
    )
    return response

Capture user feedback

Map Gradio thumbs-up/down events to Muster scores:

import gradio as gr

def handle_like(data: gr.LikeData):
    langfuse.score(
        value=1 if data.liked else 0,
        name="user-feedback",
        trace_id=current_trace_id,
    )

Wire handle_like to your Chatbot component's .like(...) event. The score appears on the corresponding Muster trace and can be sliced in analytics.

Gradio

Setup

Implementation

Capture user feedback

See also

On this page