MetaBot

Definition

The MetaBot is a feature that allows querying multiple bots simultaneously and determining the best response to provide to the end user. The MetaBot is created by establishing a parent bot, to which several child bots are attached.

Regardless of the nature of the user's question, they have a single entry point to ask it. Knowledge base managers only manage the bots for which they are responsible. Deploying a MetaBot is a one-click process; it simply requires having multiple bots and activating the MetaBot option in the dydu solution.

For example, a company might have a general bot (MetaBot) and two child bots: an IT bot and an HR bot. When an employee interacts with the company's general bot, both the IT and HR bots are queried without notifying the user. The user simply receives the response.

Answer arbitration

The arbitration mode determines how matching results from all bots within the MetaBot will be handled. You can choose between two arbitration modes:

Prioritization:

In this mode, child bots are listed in order of priority. All of them are queried by the parent bot. If there are direct matches across multiple child bots, the response from the highest-priority bot in the list is displayed.

Selection:

In this mode, all bots are queried. If there are direct matches across multiple child bots, the end user must choose which bot they wish to query from among those bots.

Management of the MetaBot knowledge base

From content > knowledge:

In a MetaBot context, when adding a reformulation to a response, it's possible to identify which bot the reformulation belongs to.

From learning > conversations

The knowledge manager cannot view, complete, or create knowledge for a child bot on which they do not have rights.

In a MetaBot context, all conversations are recorded within the parent bot's conversations. Conversations are attributed to the child bot if at least one interaction in the conversation originates from the child bot's knowledge.

The name of the bot involved in the interaction is displayed on the conversation. However, the bot's name is not shown if the "concerns only" filter is applied in the open conversation.

From the parent bot, you can filter by child bot. This allows you to display only those conversations where at least one interaction has utilized knowledge from the selected child bot.

From statistics > exploitation > conversations

In the context of MetaBot, it's possible to filter statistics by the child bot. Users have access to general statistics via the parent bot and detailed statistics via the child bot.

For example, when I'm on the parent bot (Documentation), I see 2 conversations. When I filter by Childbot_A, I now see only one conversation. I can find this same conversation in the statistics.

METABOT Configuration

To enable this option, follow these steps:

  1. Go to the Preferences > Bot > General page.

  2. Check the Enable Metabot option.

  3. Click Update. A new Personalities submenu will appear at the top of the page.

Note: you can delete the default social base. This will permanently disable the social knowledge of the social bot.

To add a new bot to your MetaBot, click Add, then configure the fields:

  1. Select the bot you want to attach to your MetaBot.

  2. Define the bot's label. The label will be the title in Selection mode.

  3. Define the type of bot you are adding:

    • Public: allows the bot to respond immediately via the current dialogue box.

    • Social: allows the bot to respond immediately via the current dialogue. This Social bot is not considered a field bot. Interactions are not counted as business interactions.

  4. Click the blue checkmark to confirm.

You can also change the Arbitration Mode of the MetaBot. Two modes are available:

  • Selection Mode: If multiple bots have the answer to the user's query, the user can select the bot from which they want the response.

  • Priority Mode: If multiple bots have the answer to the user's query, the highest-ranked bot in the list will take priority.

Note: When a knowledge with contextual conditions contains responses that are not configured/validated in the branch, the bot returns the closest knowledge with a valid response.

Don't forget to click on Update to save the changes.

Last updated

Tous droits réservés @ 2023 dydu.