The Conversation pages let you connect to the Dialogflow project, Upload intents from a CSV file, Download a Dialogflow agent, and modify intents/response directly within the Dialogflow project.

Edit in Dialogflow at #

The link to the currently active Dialogflow agent is displayed.  This allows you to confirm you are connected to the correct agent as well as access the agent directly.

NOTE: The Edit menu is not accessible until a Security Key has been uploaded to connect the Codebaby avatar to the Google agent.

Download Agent #

NOTE: If the conversation engine is set to LLM in Conversation/Connect, this option does not display.

Download Agent downloads the current Dialogflow agent.  This allows you to preserve any edits that were made within Dialogflow.  There are two download options:

• CSV – downloads a csv file which can be imported into a spreadsheet program such as Excel.
• JSON/ZIP – downloads a JSON directory in a zip file.  Each intent is a separate JSON file.

Upload Agent #

NOTE: If the conversation engine is set to LLM in Conversation/Connect, this option does not display.

Upload Agent allows you to upload intents directly into Dialogflow from a csv file.  The csv file is formatted with the following information:

Upload Agent CSV File for Dialogflow:

This CSV file allows you to upload intents directly into Dialogflow. The file should be formatted with the following columns:

Cell A1 –

Column A: IntentID

• Each intent should have a unique IntentID.
• The sequence should start with ‘1’ and continue incrementally.
• The same IntentID should be assigned to all UserInputs associated with a common ResponseText.

Column B: IntentName

• This is a unique identifier for each intent within the Dialogflow agent.

Column C: UserInput

• This column contains questions, including training questions/phrases, that the user might say.

Column D: ResponseText

• The answers or responses that Dialogflow should provide when the corresponding UserInput is recognized.

Column E: PayloadJson

• This column allows for rich responses within Dialogflow, utilizing JSON formatted data.

Column F: InputContext

• InputContext controls intent matching.
• While contexts are active, Dialogflow is more likely to match intents that have input contexts configured, which are a subset of the currently active contexts.

Column G: OutputContext

• This column controls the active contexts.
• When an intent is matched, any output contexts configured for that intent become active.

Column H: Lifespan

• This defines the number of conversational turns for which the context remains active.

Column I: ResetContexts

• This is a command in Dialogflow that tells the system to clear any data stored during an interaction.
• It is used to reset conversations and start them fresh when needed

Column J: isFallback

• Indicates whether the intent is a fallback intent.
• Fallback intents are invoked when the bot doesn’t recognize the user input as an intent after a configured number of attempts for clarification when the conversation is started

Note: Errors related to Dialogflow displayed in Codebaby can also be seen in the GCP Activity log at Google Cloud Console Activity.

Restore / Replace #

This option overwrites all intents in the Dialogflow agent.

Import/Merge #

This option appends the intents in the Dialogflow agent. What about duplicates?

1. Ensure the csv file to be imported adheres to the required format.
2. Select either ‘Restore / Replace’ or ‘Import / Merge’.
3. A ‘Working’ message displays while the data is being import.  When the import is completed ‘Saved to Dialogflow’ displays.
4. Click on the ‘Edit in Dialogflow at’ link to see the new intents in Dialogflow.
5. If the csv file cannot be imported error messages display.

Refresh Current Chat #

Refresh Current Chat initializes the conversation and clears the conversation log.

Initialize Conversation #

You can add a string that will pass as an intent to your conversation that will fire when your avatar is loaded. This allows you to prompt the avatar to begin the interaction without the user having to type or speak to the avatar first. For example, you can use <> and add that to your training phrases for your default greeting intent.

Input and Output Contexts: #

Contexts and inputs are the foundation upon which natural language processing (NLP) systems are based. Inputs serve as the data that machines understand and draw on, such as text or audio. Contexts are the conditions and frameworks within which an NLP system makes decisions, such as sentence structure and context. Having a well-defined set of context and inputs is essential for building an effective and accurate NLP system. With a clear understanding of the necessary contexts and inputs, machines can comprehend the language being used, accurately interpret the meaning of words, and draw the right conclusions to solve complex problems.