----------------------------- **While I generate an answer, here are some sources that might be helpful**: * [Web Calling]() * [Orchestration Models]() * [Tools Calling]() * [List Assistants]() -----------------------------

### Answer It sounds like you're encountering significant issues with hallucinations and repeated responses in your Vapi integration. Here are some potential solutions and considerations: 1. Hallucinations at Low Temperature Settings: Even with the temperature parameter set low, hallucinations can still occur. This might be influenced by other factors such as the underlying model's training data or its current state. If you're consistently seeing this across different models, it might be helpful to review the prompts you're using and ensure they are as clear and structured as possible. Experimenting with different prompt designs can sometimes yield better results. 2. Repeating Responses: The issue of the assistant repeating responses might be related to how the state management is handled within Vapi. Ensuring that the assistant's state is reset or managed correctly between interactions can help prevent this issue. It might also be beneficial to look into how responses are being queued or managed within your application's logic to ensure that duplications are not occurring due to the application itself rather than Vapi. 3. Cross-Testing with Other Providers: Since you mentioned that this issue has not occurred with other providers, it suggests that there may be specific configurations or operational nuances in Vapi that are contributing to these problems. It might be useful to closely compare the integration details between Vapi and other providers to pinpoint any differences in implementation that could be adjusted. If these issues persist and are impacting your use case significantly, I recommend reaching out to Vapi's support team with detailed logs and descriptions of the issues for more targeted assistance. They might be able to provide insights or adjustments specific to your setup that could resolve these anomalies. For further guidance on s

etting up and troubleshooting your Vapi assistant, you might find the documentation on orchestrating models and managing assistant behavior helpful [(1)]() .

Marking it as complete as there is no follow up.

@Sahil we are actually now having these problems without KB

d44992de-fe2f-4629-b2ca-7f527fcd7f43 and dc873045-527f-4c35-97ff-89ccfdd6c27e. It's pronouncing things strangely