AI Gateway

Playground

Test endpoints with an interactive chat interface before routing production traffic.

Overview

The Playground is a chat interface for testing your endpoints before sending production traffic through them. Send messages, watch responses stream in, and check that guardrails catch what they should.

Using the Playground

  1. Select an endpoint from the dropdown. Only active endpoints appear, filtered by your role.
  2. Type a message and press Enter (or click the send button).
  3. The response streams in real-time with markdown rendering.
  4. Continue the conversation. Full message history is sent with each request.

Playground settings

Click the gear icon to open the settings modal:

  • Temperature: Controls randomness (0 = focused, 2 = creative). Default: 0.7
  • Max tokens: Maximum number of tokens in the response. Default: 4096

Settings are applied per-request and override any defaults configured on the endpoint.

Testing guardrails

If guardrail rules are active, the Playground respects them. A message containing blocked content (e.g., personal data when PII detection is enabled) will be rejected before reaching the LLM. Masked content will be replaced with placeholders.

Guardrail errors
When a guardrail blocks a request, the Playground shows the reason (e.g., "Blocked by guardrail: content_filter: confidential detected"). Good way to verify rules work before going to production.

Cost tracking

Every Playground message is logged in Analytics with cost and token tracking. If a conversation costs $0.02 per exchange, multiply by your expected daily volume to estimate monthly spend.

PreviousEndpoints
NextGuardrails
Playground - AI Gateway - VerifyWise User Guide