LLM Tasks

LLM tasks leverage generative AI to perform advanced capabilities, such as rephrasing confusing chatbot user messages, checking whether NLP matched passages are helpful, and generating handover transcript summaries for your livechat agents.

Each task performs a specific function using the connector (the provider and LLM model) and prompt you specify, along with additional configurations like the timeout period. Some tasks may also need to be enabled before they'll appear in your live chatbot.

You need an administrator or publisher role in your team to create and edit tasks. See Create tasks.

Tasks and usages

Some features require several tasks to work together. Tasks that work together are grouped into a usage, such as the TrueIntent usage which contains tasks for rewriting chatbot user utterances, checking if NLP matches are helpful, and using an LLM to find an answer to your chatbot user's question. 

Each usage type contains a pre-defined set of tasks: you create the usage, then configure the tasks it contains. Tasks can't be moved from one usage to another. See Create usages to configure tasks. 

You can create multiple usages to experiment with different prompts and connectors for tasks, but the only the primary usage is used by your live chatbot.

Task prompts

Each task is preconfigured with a default prompt, making them ready to use out-of-the-box. You can view or customise this prompt from the task's tab in its usage page. Each task is chatbot-wide, using the same prompt for the whole chatbot, but some tasks allow you to add additional prompt instructions per-use as additional context, such as to customise transcript summaries for a specific passage.

Timeouts

Sometimes an LLM model may take a long time to respond to a prompt, such as if there's a temporary network issue. To prevent your content creators or chatbot users waiting too long, the task will wait for the model to respond for a set period of time. If it doesn't receive a response from the model within that time, it will continue without the model's response. For example:

• If the model takes too long to generate suggested training phrases, the inGenious AI platform notifies the content creator.
• If the model takes too long to rewrite an utterance, the chatbot continues NLP matching with the chatbot user's original utterance.

Enabled tasks

Some tasks, such as generating handover transcript summaries or rewriting chatbot user utterances, can indirectly impact your chatbot user experience. To ensure these tasks are thoroughly tested before you publish them in your live chatbot, the tasks are disabled when you first create them.

Tasks must be enabled to test them in TestBot, but they won't run in your live chatbot until you publish the change. You can see if a task is enabled at the top of the task's tab. 

Only tasks for chatbot usages like TrueIntent and handover transcript summary need to be enabled: tasks from platform usages such as generating training phrases or refining copy don't directly impact your chatbot experience and are always enabled.

Create usages to configure tasks

Your chatbot has basic tasks for generating suggested training phrases and refining chatbot message text. These tasks are already configured out-of-the box: if you want to use them as they are, you don't need to do anything.

If you want to customise your tasks or use advanced tasks such as generating handover transcript summaries, you'll need to create the appropriate usage then configure the task. Contact inGenious AI support to enable advanced generative AI features in your chatbot.