Skip to main content
Testing Mode runs your scenario as a text-only chat inside the scenario editor. It uses the same nodes and configuration present in the editor so you can validate flow logic before starting a real conversation. It does not use microphone, camera, avatar video, or text-to-speech. Testing sessions do not consume your voice or video conversation minutes (Testing Mode uses a separate monthly allowance instead).

​
Open Testing Mode

  1. Open a scenario in the editor (create or edit from Scenarios).
  2. Open the Scenario Menu (side panel).
  3. Select the Testing Mode tab (next to Settings).

​
Runtime variables

Testing Mode includes a runtime variables editor (JSON object). Values here will be passed in as runtime_vars when you start a test conversation. These variables substitute into {{runtime.*}} placeholders in your role instruction, task instructions, and compatible endpoint templates. For role vs task instructions and how they use {{runtime.*}}, see Role and task instructions. For using {{runtime.*}}, {{secret.*}}, and {{llm.*}} syntax in HTTP headers and bodies, see Templates and variables.

​
Start a test session

  1. Open a scenario in edit mode
  2. Select Visual in the Visual/JSON Toggle in the top right.
  3. Open the left side nav, select the Testing Mode tab, and enter runtime variables if needed.
  4. Click Start Conversation.
  5. You will enter a text-based chat with the LLM, being instructed by your scenario. Each user turn, assistant reply, and tool call/response will appear in the chat transcript as the scenario runs.
You can iterate on the scenario and test how the bot will behave before you save and run a full avatar-based conversation.

​
Monthly testing sessions

Each successful Testing Session counts against a monthly quota for LLM-only testing. The limit depends on your billing plan.
Vision tools do not work in LLM only Testing Mode

​
See also