Additional settings

Once you have specified the main Bot Instructions, entered the Bot Name, and set the Timeout in seconds, you can proceed to the additional settings. To do this, expand the settings menu by clicking the Additional settings tab:

A menu with a number of additional settings will expand before you:

Let's break down each setting individually.

1. Message History (Context):

Transfer message history - this parameter determines how many recent messages are transmitted to the model for processing. In other words, how many of the last messages the bot remembers.

• Disabled - in this case, the bot will operate in Question-Answer mode.

• Enabled - the entire dialogue over time will be passed to the model. The model will better understand the context of the conversation, but this will increase token consumption, and in some situations, may worsen understanding of the context. For example, in the beauty industry, when a client books a service once a month and asks the same questions at a certain frequency. In this case, we recommend limiting the history context by time.

• Use last X messages - this means a certain number of messages defined by the user will be stored in the bot's memory.

• Use last X minutes - this means the message history will be stored in the bot's memory only for a certain time.

1. Save function messages

Save function messages - after the initial function call, the bot will keep this function in memory and reference it later without spending additional tokens on a new call. This setting increases the cost of the whole dialogue but reduces the cost of an individual message.

1. Knowledge Base Options:

• Keywords - since the knowledge base fundamentally works on the semantic understanding of context (the bot decides when to select a particular file for response), there can be situations where it doesn't invoke the knowledge base. This setting allows you to specify words that will force the bot to definitely reference the knowledge base even if it's not necessary from the context.

• Increased heat of the knowledge base - a setting allowing to specify the level of necessity to reference the knowledge base. If the switch is on, the bot will more frequently reference the knowledge base rather than use its own knowledge.

How to add keywords:

1. Specify the keyword or phrase.

2. Press "Enter".

1. Dialogue Interruption/Resumption:

• Do not respond when interrupted by an empolyee - if an employee intervenes in the dialogue depending on the channel the request comes from, the bot will stop responding.

• Ignore first messages from the employee - activating the flag allows you to select the number of employee messages for which the interruption will not be executed.

• Specify phrases for wich the interrupt will not be triggered - the dialogue will not be interrupted when the selected phrases are received.

• Interruption by control phrase - if the specified phrase appears in the dialogue (from an employee or client), the bot will stop responding and hand over the dialogue to a human (administrator). There are situations where an employee writes in the final channel and CRM systems can't always recognize this message as from an employee. In this case, this setting helps pause the dialogue from the final channel.

• Return control to the bot by timer - setting which allows resuming the bot's communication after a set time.

• Spcify resume phrases - a phrase after which the dialogue will resume.

In some channels, there may be service messages or messages that the bot cannot process. This setting allows you to specify text that will be sent if an unrecognized message is received.

1. Image Recognition:

This settings section allows managing the process of image recognition by the bot. You can choose one of the image processing modes:

• Disabled - the model will not recognize images.

• If not Found in KB - in this mode, the bot will first search for an image match in the knowledge base. If no matches are found, the bot will act according to the instruction specified under the drop-down list.

• Always - in this mode, the bot will always recognize images and act according to the instruction specified under the drop-down list.

• Transmit to the model - advanced mode where the model recognizes images without returning text and immediately understands the image's meaning.

1. Spam Protection:

Maximum number of messages in dialogue - a setting that allows limiting the length of the dialogue in terms of message count.

Maximum cost of the dialogue - allows setting a limit on the dialogue cost.

Since every message has a cost, there is a risk of "draining" the budget in some intentional or unintentional situations when a client writes too many messages. This setting helps protect against unnecessary balance loss.

1. Messages Before and After:

This is the text that will be sent Before and/or After a message from the neural network. For example, it can be used to sign that it is a bot.

For example: The bot responded with the phrase

Good day, how can I help you?

And in the channel (and consequently to the end client) will be sent:

Message from the bot:

Good day, how can I help you?

Text written by Suvvy

If in these settings you specify Message from the bot: and Text written by Suvvy respectively.

1. Bot Response Splitting:

Splitting the bot's response into several messages depending on the chosen settings.

How to set it up?

1. Select the type of response splitting.

2. Set the number of sentences/paragraphs for one part of the split response.

1. Working Hours:

This section allows specifying the time zone within which the bot operates and responds. It also allows setting the weekly and daily work schedule.

How to set the bot's work schedule?

First, enable the setting Specify working hours:

Next, set the desired working time for the bot.

Let's discuss an example.

Suppose we need the bot to respond only outside of working hours. That is, from 8:00 PM to 10:00 AM, then the time settings will look as follows:

That is, we split the bot's working time into two time segments within one day.

From 00:00 to 10:00 and from 20:00 to 23:59.

1. Replacement Patterns:

Replacement patterns are used to substitute unwanted words or phrases in the bot's response with desirable ones.

For example, if you want the bot to use "company" instead of "firm" in responses, fill in the fields as follows:

The enabled setting Pattern can be part of another word allows replacing part of a word in the response:

1. Send Follow-Up:

Follow-up - this is sending additional messages to the client if they do not respond to the bot's last message.

To configure this section, click New Message:

Message settings will open:

• Time after which the message will be sent - when the time expires, the message will be sent.

• Reschedule the event if the time for wich it was scheduled turns out to be out of order — if the follow-up falls on non-working hours, it will be rescheduled and sent during the designated working time.

• Set exact time - send the message at an exact time.

• Set call limit in dialogue — the number of times a follow-up can be successfully executed within one sending cycle.

• LLM instruction — used to determine whether certain actions should be performed in the dialogue with the client based on the information provided in the conversation.

• Fixed message — if this function is selected, a specific message must be entered in the input field below, which will be sent to the user.

• Message generation by existing bot — if this function is selected, an instruction for additional bot actions and message generation can be entered in the input field below. For example: "Remind about yourself based on the dialogue context and call the function get_file_text(file=Administrator)".

• Message generation by LLM model — if this function is selected, an instruction for message generation must be entered in the input field below. For example: "Remind about yourself based on the dialogue context."

• Call Knowledge Base file — allows sending a specific knowledge base file. If this setting is enabled, the file to be called by the follow-up must be selected in the "Called file" field below.