We offer three ways to test your agent:

​
1. LLM Playground

Test conversation flows and responses by manually chatting with your agent using the LLM Playground.

​
2. LLM Simulation Testing

Automate testing by setting up customer scenarios and evaluating how your agent handles them. This method includes metrics-based evaluation of results.

​
3. Web Call/Phone Call Testing

Make test calls to your agent to evaluate its performance during real phone conversations.

​
Features and Limitations

LLM PlaygroundLLM Simulation TestingWeb Call/Phone Call TestingAudio Simulation Testing
AvailabilityAll agentsConversation flow agents onlyAll agentsAvailable soon
Required effortMediumLowHighLow
Response accuracy testing✅✅✅✅
Function calling accuracy✅✅✅✅
Background noise testing❌❌✅✅
User interruption testing❌❌✅✅
Batch testing capability❌✅❌✅

  1. Start with LLM Playground

    • Begin by manually chatting with your agent to iteratively improve its capabilities.

  2. Progress to LLM Simulation Testing

    • Once you have a functional agent, test it with predefined customer scenarios to evaluate its performance.

  3. Conduct Web Call/Phone Call Testing

    • Test your agent in real-world conditions through actual phone calls to assess its handling of voice interactions.

  4. Implement Batch Testing

    • Create and run multiple test cases to ensure your agent maintains stability and performance across various scenarios, especially after making changes.

Was this page helpful?