-
Notifications
You must be signed in to change notification settings - Fork 7
Building bots
Building a bot requires setting up an Experiment. An experiment brings together various features and information that ultimately dictates the behavior and knowlege of the bot. To create a basic bot, you'll only need a prompt.
This serves as the initial instruction for your bot. A prompt can be as simple as "You are a helpful assistant".
The temperature parameter influences the randomness of your bot's responses. For a detailed explanation of how to use and fine-tune the temperature parameter, check out this article., but the default value should be fine for most use-cases.
If you have specific material that you want your bot to draw knowledge from, you can create source material within your experiment. This allows your bot to answer questions and provide information based on the content you provide.
The safety layer allows you to review user or bot messages to determine their safety. You can specify a prompt that will be used to assess the safety of the content. If the content is deemed unsafe, the bot will be prompted to formulate an appropriate response.
This allows your bot to use a synthetic voice when responding to voice notes, provided that the user's communication channel supports them. Otherwise, the bot will respond with text.
Only AWS Polly is currently supported. To set up a synthetic voice for your bot, you'll need to configure the following environment variables:
AWS_POLLY_ACCESS_KEY_ID
AWS_POLLY_SECRET_KEY
AWS_POLLY_REGIONConfigure the activity settings to determine when your bot should prompt the user to respond. The activity config dictates the timing and frequency of these prompts, ensuring a seamless interaction with the user. See the no activity ping page for more information.
See the LangChain page on Agents.