Hello I'm having an issue with squads, where if bring in two assistants that seem to work perfectly well on their own, don't work as expected in the context of the squad. At first the problem seemed to be that I needed to override the assistant prompt/model/voice in the squad config for each assistant. I copy/pasted the system prompt and adjusted model/voice model to be identical as the assistant's (which seems like something we shouldn't need to do) and even though it works, the assistant's tone and personality is completely different. Is this something I should be configuring via the API vs the UI? You can hear the difference by checking this squad : 0d7e9a84-0145-4659-bd55-5bb661c730ef Vs the assistant : babf8de3-b6e9-4759-9981-db1b6b37812c Oh and small side note : in my second assistant, which the call is being transferred to, if I use any Cartesia voice, the voice speaks extremely slowly. Other voice models don't seem to have this problem

Hey there! 👋 I understand your frustration with the squad vs. solo assistant behavior differences. Let me help explain what's happening and provide some solutions. Unfortunately, I can't directly listen to your specific squad and assistant examples, but I can definitely address the issues you're experiencing. ## Why Squad Assistants Behave Differently This is actually a known behavior! Several factors can cause this: 1. **Global Overrides**: Squad configurations use

membersOverrides

which act as "house rules" for all assistants in the squad, overriding individual assistant settings. According to the [documentation](https://docs.vapi.ai/api-reference/squads/create-squad), both

membersOverrides

and

members[n].assistantOverrides

can be used together, with

membersOverrides

taking precedence. 2. **Different Parameter Handling**: Parameters like

maxDurationSeconds

are handled differently in squads - they must be applied at the squad level rather than individually per assistant. 3. **Instruction Impact**: Specific instructions in squad configurations (like "When transferring to another assistant, do not mention the transfer to user, just do it") can cause immediate transfers that wouldn't happen in solo mode. ## API vs UI Configuration Based on the issues you're experiencing, configuring via the API might give you more precise control. The API allows you to: - Explicitly set all parameters including system prompts, voice settings, and experimental controls - Configure

firstMessageMode

options like 'assistant-speaks-first' - Adjust specific voice parameters that might not be easily accessible in the UI Some users have reported UI bugs with squad configurations, particularly when renaming assistants within a squad. ## Cartesia Voice Speaking Slowly For your Cartesia voice issue, you can try these solutions:

Hey! To help track down this issue, could you share: - The call ID - When exactly this happened (the timestamp) - What response you expected to get - What response you actually got instead This would really help us figure out what went wrong!

Hey @Shubham Bajaj I wasn't using a call ID because I tested this just by using "Talk with Squad" and I don't see any record of those interactions in the Call Logs. The squad ID is : 0d7e9a84-0145-4659-bd55-5bb661c730ef The assistant ID is : babf8de3-b6e9-4759-9981-db1b6b37812c But the issue is that you have to listen to how the assistant delivers the exact same system prompt vs how the squad does it. They are both using Claude 3.7, but the assistant follows the "tone" and personality I developed in the system prompt, whereas the Squad version of the same assistant (with the system prompt assigned in the assistant and at the squad level, identical) takes on a different ton and personality. I'm going to try to setup the same thing via the API to see if this is a configuration issue related to the UI

Hey Ouranos, Whenever you click on "Talk with a Squad," it creates a call ID which is under your VAPI dashboard > Observe section. You can find it over there. Sorry, I cannot help you with anything as I don't know exactly what could go wrong for you. And I do not agree with any of your claims until I check the logs using the call ID.