On this page |
Section visible by enabling AI Bots on license
◾ OpenAI ChatGPT
AVAILABLE FROM VERSION 3.36.0
This block will allow Voicebot creation via ChatGPT, by configuring an OpenAI account under Cloud Providers section.
Label: here you can type a brief description
Choose an OpenAI bot: select the created bot from the list (mandatory field)
Text you want to send (e.g. the variable {GOOGLE_ASR_TRANSCRIPT}) - (discover below the IVR project built with ASR and TTS blocks)
Sound to play while waiting (available from version 3.40.0): a playback sound, previously uploaded in Tools/Sound section, executed for the customer while waiting for a response from OpenAI, such as, “We are analyzing your request, we will get back to you shortly” , thus improving the user experience and avoiding audio gaps.
This option is not mandatory and by default the sound is set on “No option chosen”. If you decide to select a sound, it is executed right after sending the HTTP request, while the server waits for OpenAI answer. The playback is executed to the end, without being interrupted in case OpenAI's response was already ready before the completion of the audio.
In the Cally Square context, the user can configure the project, by using one of the ASR blocks to understand the conversation, then the ChatGPT block to generate the response, and one of the TTS blocks to play the response.
Requirements for the project:
Set up a Google-Type Cloud Provider with a valid API key
Set up an OpenAI Cloud Provider with a valid API key
Check if available or set a variable in Tools → Variables called GOOGLE_ASR_TRANSCRIPT
Set in the TTS blocks the Google provider
Set in the ARS block the Google API key
Choose in the set Variable block the variable GOOGLE_ASR_TRANSCRIPT
In TTS blocks, set the Google provider
In ASR block set the Google API Key
In Set Variable block, choose GOOGLE_ASR_TRANSCRIPT
In Tools → OpenAI Bots, configure an OpenAI bot with the parameters reported below
Recommended model "GPT-3.5-turbo" for simple bots, otherwise "GPT-4" (slower, but understands better in case of unclear transcripts)
Max tokens: 2000
Exit phrase: "I am redirecting you to a human operator".
ChatGPT prompt:
You are a voice bot assistant. Help users as much as possible.
Answer in a concise way. Answer only in English.
Answer like in a phone conversation.
If the customer asks for a reservation, ask first for the day, then ask for the time.
If the customer asks for warranty information, ask for the customer number, then redirect to a human operator.
If the customer asks for a cancellation, ask for the customer number, and then the reservation number.
If you don't know how to help the user, tell the customer "I am redirecting you to a human operator".Temperature: 0.1
Analyze chat (RECOMMENDED ENABLED):
Answer in a JSON format. The key is 'CHAT_PROGRESS', which is equal to '1' if chat is complete. Else it is equal to '0'. Write a key REASON about why CHAT_PROGRESS should be 1. If you say "Have a good day", CHAT_PROGRESS = 1. If you say "I am redirecting you to an human operator", CHAT_PROGRESS = 1Forward message: “I am redirecting you to a human operator“
Error message: “Sorry, I ran into an error, I am redirecting you to a human operator”
Attachment message: “/“ (a voicebot can not have an attachment sent)
◾ Dialogflow V2
AVAILABLE FROM REL. 2.5.7
This box allows you to build a voice bot using the Google Dialogflow integration
Explore this documentation to find out How to retrieve Google Key for Cally Square blocks
An Internet connection is required for this box to work
Remember:
This software is managed by others. Check if it works properly.
Label: here you can type a brief description
Project ID: Cloud Platform project ID
Client Email: email address associated to Service Account Key
Private Key: private key associated to Service Account Key
Language: the language you want use for the bot
Text: the text you want to send
Please note Dialogflow requires a valid Service Account Key and a sufficient amount of acquired credits.
Furthermore it is pure experimental and it can bring to unexpected behaviour.
The DialogflowV2 block saves the results in the following variables:
DIALOGFLOW_ACTION: Matched Dialogflow intent action name
DIALOGFLOW_ALLREQUIREDPARAMSPRESENT: True if all required parameter values have been collected (true-false)
DIALOGFLOW_ENDCONVERSATION: True when 'end conversation' flag is set for the matched dialogflow intent. It is useful when you want to transfer a call to an agent (true/false)
DIALOGFLOW_FULLFILLMENTTEXT: The text to be pronounced to the user or shown on the screen
DIALOGFLOW_INTENTNAME: The unique identifier of the intent
DIALOGFLOW_INTENTDISPLAYNAME: The display name of the intent
DIALOGFLOW_ISFALLBACKINTENT: True when matched dialogflow intent is fallback intent (true-false)
DIALOGFLOW_LANGUAGECODE: The language that was triggered during intent detection
DIALOGFLOW_QUERYTEXT: User input
DIALOGFLOW_RESPONSEID: The unique identifier of the response
DIALOGFLOW_SCORE: Matching score for the intent (0-1)
DIALOGFLOW_SPEECH: Text to be pronounced to the user
DIALOGFLOW_RESOLVEDQUERY: The query that was used to produce the result.
Exit Arrows
This box provides just one arrow out to the next step
TIP
Click here to learn more about voice bot triggers useful to enable conversation transcript view
◾ Dialogflow
DEPRECATED FROM REL. 2.5.7
This box allows you to build a voice bot using the Google Dialogflow integration (click clip1 and clip2 and learn more about this topic!)
An Internet connection is required for this box to work
Remember:
This software is managed by others. Check if it works properly.
Label: here you can type a brief description
Key: your acquired client api key from the console.dialogflow.com account
Text: the text you want to send
Language: the language you want use for the bot
Please note that Dialogflow requires a valid key from the console.dialogflow.com website and a sufficient amount of acquired credits. Furthermore, it is purely experimental and it can bring to unexpected behaviour.
The Dialogflow block saves the results in the following variables:
DIALOGFLOW_SOURCE: Request source name.
DIALOGFLOW_RESOLVEDQUERY: The query that was used to produce the result.
DIALOGFLOW_ACTION: Matched Dialogflow intent action name.
DIALOGFLOW_SPEECH: Text to be pronounced to the user.
DIALOGFLOW_SCORE: Matching score for the intent (0-1).
DIALOGFLOW_STATUSCODE: Response status code. For more information, please see here https://dialogflow.com/docs/fulfillment#errors
DIALOGFLOW_ENDCONVERSATION: True when 'end conversation' flag is set for the matched dialogflow intent. It is useful when you want to transfer a call to an agent (true/false).
DIALOGFLOW_ISFALLBACKINTENT: True when matched dialogflow intent is fallback intent (true-false).
Release notes
DIALOGFLOW_ENDCONVERSATION and DIALOGFLOW_ISFALLBACKINTENT variables are available from version 2.0.77
Exit Arrows
This box provides just one arrow out to the next step
TIP
Click here to learn more about voice bot triggers useful to enable conversation transcript view
◾ Amazon Lex
AVAILABLE FROM REL. 2.0.77
his box allows you to build a voice bot using the Amazon Lex integration.
For additional information see https://docs.aws.amazon.com/en_us/lex/latest/dg/getting-started.html
An Internet connection is required for this box to work
Remember:
This software is managed by others. Check if it works properly.
Label: here you can type a brief description
Access Key ID and Secret Access Key: AWS security credentials. Required: Yes
(see http://docs.aws.amazon.com/sdk-for-java/v1/developer-guide/credentials.html )Region: AWS regional endpoint. Required: Yes (see http://docs.aws.amazon.com/general/latest/gr/rande.html#pol_region)
Bot name: the bot name. Required: Yes
Text: the text you want to send. Required: Yes
The Amazon Lex block saves the results in the following variables:
AWS_LEX_INTENTNAME: The current user intent that Amazon Lex is aware of. (Read more)
AWS_LEX_MESSAGE: The message to convey to the user.
AWS_LEX_MESSAGEFORMAT: The format of the response message. (Read more)
AWS_LEX_DIALOGSTATE: Identifies the current state of the user interaction. (Read more)
AWS_LEX_SLOTTOELICIT: If the AWS_LEX_DIALOGSTATE value is ElicitSlot, returns the name of the slot for which Amazon Lex is eliciting a value.
AWS_LEX_SLOT_*: The intent slots that Amazon Lex detected from the user input in the conversation. (ex. AWS_LEX_SLOT_PICKUPCITY)
TIP
Click here to learn more about voice bot triggers useful for enabling conversation transcript view
Related topics