Skip to Content
GlossaryPlayground - Glossary

Playground

Back to GlossaryDevelopment Tools

An interactive, real-time interface for testing conversational endpoints and converting live conversations directly into tests.

Also known as: interactive playground

Overview

The Playground is a built-in testing environment that lets you interact with your configured endpoints in real time. Unlike running a full test set, the Playground is designed for exploration, debugging, and rapid test creation from real conversations.

Key Features

Real-Time Testing: Send messages to your endpoint and see responses immediately via WebSocket communication. There is no need to configure a separate testing environment.

Split-View Interface: The Playground uses a split-view layout—your conversation on one side and the response with trace details on the other—so you can see exactly what your application returned and how it got there.

Trace Linking: Every Playground session generates a trace that you can inspect directly. This makes the Playground a powerful debugging tool: send a message, see the response, and immediately drill into the execution trace.

Test Creation from Conversations: Convert any Playground conversation into a formal test with a single click. Rhesis uses an LLM to pre-fill the test fields (expected behavior, category, tags) from the conversation content, reducing manual entry.

Single-Turn and Multi-Turn Support: Create both single-turn tests (one message/response pair) and multi-turn tests (full conversation sequences) from Playground sessions.

How to Use the Playground

  1. Navigate to your endpoint in the platform
  2. Open the Playground tab
  3. Select the endpoint you want to test
  4. Type a message and press Send
  5. Review the response and the trace
  6. Optionally, click "Create Test" to save the conversation as a test

WebSocket Communication

The Playground uses WebSocket connections for real-time communication, enabling low-latency responses from your endpoint. The connection includes an automatic retry mechanism to handle intermittent network issues gracefully.

Best Practices

  • Use the Playground before creating a formal test set to understand how your endpoint behaves in practice
  • Create tests from Playground conversations to capture real examples of both correct and incorrect responses
  • Review the linked execution trace for every Playground session to diagnose unexpected outputs
  • Switch between multiple endpoints in the Playground to compare behavior across environments or model versions

Related Terms

TermsAcknowledgments