How to configure agent settings in VocalLabs

Step by step

1. 1

   First, we click the "View Agent" button to access its configuration. This is the first tab of the agent, which is the Basic Info. This includes changing the gender and the language. We are selecting this voice, specifically Mel, as John's. We can also change the agent's language. We're keeping it Hindi for this example. We can also change the agent's publicity settings, make it a broadcasting agent. We can enable call transfer, enable language transfer, and also configure Lumo. Now, moving on to our second tab, this is the most important tab, the Behavior tab. Here, we have three parts: the welcome message, the static message, and the main prompt.

2. 2

   We are using this prompt for the incoming conversation, which has a variable. Now, the static welcome message will be similar to our dynamic welcome message but without any variables. This message is used if the dynamic welcome message fails. This ensures that a default greeting is always provided to the user, leading to our static welcome message. Now, the most important part, which is the prompt for the agent. we'll mention a tone style, role, some guardrails and simply click on the "Save Changes" button to apply all these configurations.

3. 3

   After this, we can also select one of the options like WhatsApp, Telegram, or Line, if we are having specific app-specific integrations. This expands the agent's reach. Now, we'll go to our next tab, which is the Data Collection tab. Here, it is the data that the agent will collect from the call. We can set any prompt here to describe which data has to be collected. Or we can explicitly mention within the data to be collected from. Here, we can name the field and describe what is to be collected. We can also change the data type, which offers flexibility in data handling. We are skipping it for this example. This is the Call End Conditions tab. We can alter the conditions according to our needs.

4. 4

   For now, we will get the default settings, which cover most of the cases. These settings are designed to be broadly applicable, simplifying initial setup for users. Now, in the Post Call Data Collection, we can collect the data, but unlike the Data Collection tab, it will work after the call. So, the AI model will gather all the call recordings and retrieve the data from them, providing valuable insights post-interaction. Moving to the FAQ section, we can add normal letter cases where the agent can find and use text. And similarly, we can provide any docs for agent context. In the Tool Call section, we can add any tools, which expands the agent's capabilities significantly.

5. 5

   Like book a meeting or check the date of birth. Also, we can create our own tools that could be REST APIs, offering extensive customization. In the Configuration tab, we can change which number the agent will call from. This allows us to manage the agent's outgoing call identity effectively. Along with changing the AI model that we use. In the Timeout and Inactivity section, we can alter data like call duration and initial visibility settings. Next time duration, and all those. Also, there are some other options like silent detection, call end logistics, and all these things. We're skipping all of them for now. This is the DTMF/IVR step, where we can make menus.

6. 6

   For example, we can add language settings when calling, providing a more tailored user experience from the start. And in the Keyword section, there are three tabs: TTS, STT, and AGENTkeyword. Each serves a specific purpose in managing agent responses. TTS defines how a keyword agent pronounces. STT allows for speech-to-text replacement, which can improve accuracy. Agent keywords are for long-term value. These are all crucial for keyword management. And the agent keyword will be doing something specific on some specific keywords.

7. 7

   And then, last but not the least, is the Agent Data section. Here, we will see all the agent logs. This includes the conversation between the agent and the user. Configured the agent and settings. Next video shows effective campaign use. This concludes our overview.