Tutorial: Analyzing Chat Data with Kura¶

What You'll Build

By the end of this tutorial, you'll have built a complete analysis pipeline that processes 560 real user queries, identifies the top 3 query categories that represent 67% of all requests, and creates production-ready classifiers for automatic query categorization.

This tutorial showcases an entire end to end flow of how to use Kura for your specific application.

Prerequisites¶

Install Kura in a virtual environment with uv pip install kura
Set your OPENAI_API_KEY environment variable to use OpenAI's GPT-4o-mini model
Download the tutorial dataset

Download Dataset

Tutorial Series¶

Step 1. Cluster Conversations¶

Discover user query patterns through topic modeling and clustering. You'll learn to identify that three major topics account for 67% of queries, with artifact management appearing in 61% of conversations.

Start Clustering Tutorial

Step 2. Better Summaries¶

Transform generic summaries into domain-specific insights. Build custom summarization models that turn seven vague clusters into three actionable categories: Access Controls, Deployment, and Experiment Management.

Start Summaries Tutorial

Step 3. Building Classifiers¶

Convert clustering insights into production classifiers. Build real-time systems that automatically categorize new queries and scale your insights.

Start Classifiers Tutorial