Skip to end of metadata
Go to start of metadata

You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 9 Next »

This page provides instructions for how to create a chatbot on Open Chat Studio. 

Note on terminology: On Open Chat Studio, a chatbot is synonymous with an "experiment". As such, you will see these terms being used interchangeably in the documentation below. 

Get started by watching the OCS 101 video (a refresher from the earlier section of the wiki).

Video: How to Build an OCS Chatbot

OCS 101.mp4

Creating a Chatbot 

This section assumes you have: created and tested a chatbot prompt on prompt builder, saved your prompt and identified and created source material as needed. It also assumes you have created any pre- and post-surveys and consent forms you'd like to use in your chatbot. 

Navigate to the Experiments page and click the "Add New" button on the top right.

The following steps cover the basic inputs necessary to create a chatbot (i.e. experiment). More detailed information on the other options will be added to this wiki soon. 

  • Give your chatbot (experiment) a name and description. 

  • Select your LLM type

    • Base Language Model

      • Select an LLM provider

      • Select a model (e.g. GPT-4)*

      • Select a temperature (can keep to 0.7 for now, we will provide more context soon)

      • Specify the prompt for the bot

    • Assistant

    • Pipeline (this is only available when you have the pipelines feature flag enabled)

  • Add a seed message (e.g. "hi"): every time a new chatbot session is initiated (e.g. by a user on WhatsApp/Telegram/web), this message will be sent to the chatbot behind the scenes so that the bot knows to kick off the conversation. It's important to add the seed message because this enables the first message of any interaction to be sent by the chatbot. 

  • Select any pre- and post-surveys and consent forms. 

  • Click the create button at the end. 

LLM Types

  • Base Language Model

  • Assistant

    • To understand assistants better, please refer to this and this. Instead of directly sending a query to the Language Model (LLM), you have the option to have an assistant address the user's inquiry, following a similar approach to the Base Language Model. Before utilizing this functionality, it is essential to set up an assistant in OpenChatStudio. To create an assistant, navigate to the “Assistants” section found in the sidebar. Assistants are equipped with the ability to access and search through files. Therefore, any source material that needs to be linked to the bot should be uploaded during the assistant creation process.

  • Pipeline

    • This option is only available if you have the Pipelines feature flag enabled. This allows you to run a pipeline on the user input. In this case, the pipeline will define your bot.

Note: Please contact ocs-info@dimagi.com for information about LLM providers and models. 

  • No labels