Documentation
Learn how to integrate Subtext into your AI products and start gaining insights into user conversations and behavior.
Get Started in Minutes
Follow our quickstart guide to integrate the Subtext SDK and start tracking conversations.
What is Subtext?
Subtext is the first product analytics platform built specifically for conversational AI and LLM-powered products. Unlike traditional analytics that focus on clicks and page views, Subtext helps you understand what your users are actually asking and how your AI responds.
With Subtext, you can:
- Track and analyze every conversation your users have with your AI
- Understand user intent, sentiment, and identify common requests
- Measure AI performance and identify areas for improvement
Core Concepts
Understanding the key building blocks of the Subtext platform will help you effectively track and analyze your AI conversations.
Thread
A single conversation that groups related messages and AI responses. Think of it as a chat session or conversation flow that maintains context across multiple interactions.
Message
Individual user inputs within a thread. Messages represent what users says or wants to achieve, providing context for subsequent AI responses.
Run
AI model executions and their responses within a thread. Runs capture what your LLM generates, including the actual response text, model parameters, and performance metrics.
How They Work Together
Thread → Contains multiple Messages and Runs → Each user Message typically triggers an AI Run → All tracked within the same conversation Thread for complete context and analytics.
How It Works
- Install the SDK: Add the Subtext TypeScript SDK to your project
- Create an API Key: Generate an API key from your Subtext dashboard
- Track Conversations: Use our simple API to log threads, messages, and AI responses
- Analyze Insights: View analytics and insights in your Subtext dashboard