Conversation Engine #
The Conversation Engine options are used to specify whether the agent is in Dialogflow ES, Dialogflow CX or connecting to ChatGPT or another LLM.
The Dialogflow Essentials (ES) Edition is a pay-as-you-go edition that provides the standard ES agent type. The Essentials Edition offers production-ready quotas and support from Google Cloud support.
The Dialogflow Customer Experience (CX) Edition is a pay-as-you-go edition that provides the advanced CX agent type.
LLM allows you to select from a list of available large language models. The avatar is connected to the selected LLM to manage the conversation.
Select the appropriate option:
- LLM
- Dialogflow ES
- Dialogflow CX
After selecting an option refresh the browser page using the refresh button next to the address bar in the browser.
Conversation #
In order to edit a conversation, the avatar needs to be connected to a Dialogflow agent or an LLM.
Connected Dialogflow Agent #
To connect your Google Dialogflow conversation to your avatar, you need to upload your service account key .json file here. Need help knowing how to generate that key? Go to here for instructions on how to create your key.
Note: Before connecting to Dialogflow this field says “Create your own conversation”. This link brings you to the Google documentation.
Upload Key #
Select the JSON file created using the Create New Key function in Dialogflow in order to connect the avatar to the Dialogflow agent.
Create New #
Create New automatically creates a Dialogflow Agent and Service Account Key and connects the Dialogflow agent to the CodeBaby Avatar. This greatly simplifies creating a new project.
Additional Settings #
The following fields are required for CX and optional for ES.
Agent ID #
For Google Dialogflow ES, your account will have a single Agent ID, so this field is not required. For Google Dialogflow CX accounts, you can find this value by following the steps listed here: https://codebaby.com/help/connecting-to-google-dialogflow/#gdagent
Environment ID #
For Google Dialogflow CX, enter the Environment ID for the conversation you want to connect. You can find this value by following the steps listed here: https://codebaby.com/help/connecting-to-google-dialogflow/#gdenvironment
Configure LLM Redirect #
A large language model can be used as the conversational engine for your avatar, rather than using Dialogflow or a similar tool. CodeBaby can connect to GPT-3.5, GPT-4.0 or a Custom model. When the user interacts with the avatar the LLM is driving the conversation rather than a pre-defined conversation. If you are interested in using a custom model, contact us at info@codebaby.com for assistance.
Generate Your Avatar Prompt from Web or Documents #
A system prompt is a specific input that is used to guide and instruct an AI language model (like ChatGPT) on what kind of task or response it should generate. Essentially, it’s like providing a starting point or cue to help the AI model produce relevant and accurate results.
Generate Your Custom Organization Profile #
Welcome to our unique prompt-generation tool! Here’s how it works:
- Input a Website URL or Upload Files: You can start by entering the URL of your company or organization’s website. Alternatively, you can upload a documents that contains relevant information about your business.
- We Extract Key Content: Once you provide the URL or documents, our system will intelligently analyze and extract content from up to 20 pages. This includes not just web pages, but also PDFs and text documents linked or contained in the URL.
- Generating a Custom Prompt: Using the information gathered, we will craft a detailed prompt specifically tailored for a Language Model like GPT-4. This prompt aims to encapsulate the essence of your business or organization.
- Accurate Representation by the Avatar: The custom prompt enables the avatar to act as a representative of your organization, reflecting its values, services, and unique characteristics in its responses.
Note: For the best results, ensure the URL or documents you provide are comprehensive and up-to-date, covering key aspects of your organization.
Advanced Settings #
Use Advanced settings if you want more control over the GPT model.
Provider
OpenAI with CodeBaby, OpenAI, Bedrock (Amazon Anthropic), Vertex (Google).
Model
If you have a specific model you would like to use, choose it here. The options include GPT-4.0 Turbo Preview, GPT-3.5, GPT-4.0, GPT-4.0 Turbo, Claude, Llama2 and Custom. If Custom is selected and additional field, Model ID is displayed. Specify the name of the model, for example GPT_3.5_turbo, GPT4, or MyCustomTrainedModel. See this link for additional information.
Results Temperature
Temperature is a value between 0 and 1. A higher temperature promotes more creative and longer responses, lower creates more certainty in answers.
Create Custom Embeddings from a Website or Text Files #
By supplying a website address and/or uploading source documents, the system can create the information needed for the avatar to carry on a conversation. When embeddings are updated, existing embedded data is overwritten. In order to embed multiple documents, they need to be uploaded at the same time. If you need to embed both a website and documents, use the Create Custom Embeddings function to scrape the website first, then download the resulting CSV file, then upload the website CSV and other documents together.
View/Edit Embedding Data in a CSV file format #
The files uploaded in the Embeddings function are converted to a CSV file. The file can be downloaded, edited and then reuploaded.
Download CSV file – after a URL is scraped, or documents are uploaded for embeddings, the Download CSV option is available.
Upload CSV file – upload a CSV file that was originally downloaded and then edited.
