The transcript only contains the user's side of the conversation and does not provide any evidence of the AI's behavior. Test Suite ID: d417732d-c158-44cc-89d3-d85524277e25. Run 5/1/2026, 20:26:27 (id: 589b054c-b2d8-4db0-88f6-945b6851d0a9) https://cdn.discordapp.com/attachments/1457818985151201539/1457818985616773194/Captura.PNG?ex=6979bb7b&is=697869fb&hm=8540bda65e9a09660df1df5040b9fdc534663f05fb422251802d4100a67887af&

Thank you for sharing the details and error messages. To help diagnose why the AI did not participate in your Test Suite run (ID: 589b054c-b2d8-4db0-88f6-945b6851d0a9), could you confirm if the phone number used in this test has an assistant currently assigned to it in your dashboard? This is a common requirement for successful test execution according to the documentation. If possible, please check the phone number assignment and let me know if the assistant is correctly linked. Source: - [Test Suites Documentation](https://docs.vapi.ai/test/test-suites)

Message marked as helpful by @weeds! 🎉

Assistant is linked. Regular calls and "Talk to assistant" tool on Dashoard are working

Hi weeds, Please check these in your dashboard and paste any exact errors you find: 1. Observe > Call Logs — open the run matching Run ID 589b054c-b2d8-4db0-88f6-945b6851d0a9 (timestamp 2026-05-01 20:26:27). Copy the Call ID(s), endedReason values, and any call-level error messages. 2. Observe > API Logs — filter around 2026-05-01 20:26 and paste any failing request/response pairs or HTTP errors. 3. Test > Test Suites > \[that suite\] > Runs — open the run and expand each attempt; copy the LLM reasoning / failureReason shown (if present). 4. Confirm the phone number used by the test is assigned to the assistant (Dashboard > Phone Numbers) and paste the phone number and assigned assistant ID. 5. Tell me whether the test type was Voice or Chat and whether transcripts for other runs show both sides normally. Quick likely causes based on the docs: - Phone number not actually assigned to assistant for that test (verify in Phone Numbers). See: https://docs.vapi.ai/test/test-suites - Call created but agent never answered (endedReason ≠ mockConversation.done or transcript only user-side) — check Call Logs and API Logs for STT/TTS or provider errors. - Role-swapping or transcription settings causing missing assistant audio — confirm transcriber and voice provider are compatible (Voice tests only). See troubleshooting in Debugging: https://docs.vapi.ai/debugging

1. The call can not be found in "Call Logs" as the Test Suite is running over Chat. No information shown in "Chat Logs". 2. No errors shown in any HTTP method on "API Logs". 3. Reasoning 1: "The AI did not participate in the call transcript at all. There was no greeting, confirmation of reservation details, checking of availability, or polite closure. The ideal outcome requires the assistant to interact with the user, confirm and repeat the reservation details, and end the conversation politely. Since none of these behaviors were present, the test failed." *Reasoning 2: * "The transcript only contains the user's side of the conversation and does not provide any evidence of the AI's behavior. There is no indication that the AI politely informed the caller that the requested time was unavailable, offered an alternative, or completed the reservation with a confirmation and proper call ending. Without the AI's responses, it is impossible to determine if the ideal outcome was met. Therefore, the test fails due to insufficient information about the AI's actions." **Reasoning 3: **"The AI assistant behaved as expected." Although the Call Transcript does not show any Assistant transcript, the request has been completed successfully. 4. Phone number +34936942105. Assistant ID: ef6b31bb-be9f-417b-a0f9-305648960b59. LINKED 5. Test type is via CHAT, other runs DO NOT WORK. We've created several Test Suites an different Runs with different inputs but the Assistant's Transcription is never shown. We can not run Voice Test Suites because we are using an international number (+34). Thank you in advance.

Hi, Thanks for your response. Can you try to Verify Assistant Configuration? Check your assistant (ID:

e0de4a31-daf0-4111-aa7a-4fcc479c42b3

) for these common issues: Model Configuration: • Ensure a valid LLM model is selected and configured • Verify API keys for your chosen model provider are active • Check that model parameters (temperature, max tokens) are set appropriately System Prompt: • Confirm the assistant has a system prompt/instructions • Verify the prompt doesn't contain conflicting directives • Ensure the prompt encourages active conversation participation Step 2: Test with Simple Evaluation Create a basic smoke test to isolate the issue:

{
  "name": "Assistant Response Test",
  "type": "chat.mockConversation",
  "messages": [
    {
      "role": "user",
      "content": "Hello"
    },
    {
      "role": "assistant",
      "judgePlan": {
        "type": "regex",
        "content": ".+"
      }
    }
  ]
}

This minimal test checks if the assistant responds at all to a simple greeting. Step 3: Check Tool Configuration Since you mentioned calendar integration, verify: Tool Setup: • Calendar tools are properly configured with correct endpoints • Authentication credentials for GHL integration are valid • Tool schemas match expected input/output formats Function Definitions: • Tools have proper JSON schemas defined • Required parameters are correctly specified • Tool descriptions are clear and actionable

Hi, Thanks for the quick rosponse. Model Configuration: • Ensure a valid LLM model is selected and configured 🆗 • Verify API keys for your chosen model provider are active 🆗 • Check that model parameters (temperature, max tokens) are set appropriately 🆗 System Prompt: • Confirm the assistant has a system prompt/instructions 🆗 • Verify the prompt doesn't contain conflicting directives 🆗 • Ensure the prompt encourages active conversation participation 🆗 Tool Setup: • Calendar tools are properly configured with correct endpoints 🆗 • Authentication credentials for GHL integration are valid 🆗 • Tool schemas match expected input/output formats 🆗 Function Definitions: • Tools have proper JSON schemas defined 🆗 • Required parameters are correctly specified 🆗 • Tool descriptions are clear and actionable 🆗 The assistant is working fine for production workflows. The only issue is regarding "Test Suites" as we can not see the assistant transcript.

Hey weeds, I am looking into your issue and will update you soon.

Is there any update regarding this issue? Would be nice to have an automatized test before we deploy our service. Thanks in advance

Hey weeds, I am sorry for delated response, I'll check it today for the exact issue.

Help