Dynamic Data
Interact with the real world by connecting your agent to external APIs.
Introduction
With Dynamic Data, you can make external API requests at the start and throughout your phone call. This allows you to load data from your database, or from any other API. You can then use that data in your AI responses, or to define circumstantial behavior for each call.
Some examples of what Dynamic Data enables:
- Maintain conversation history between calls
- Define behavior based on the user’s location
- Handle real-time data like status updates or prices
We’ll cover the following features in this section:
- System variables
- External API requests
- Extracting data from responses
- Variables as parameters
- Chaining requests
System Variables
Variables are defined with double curly braces, like {{variable}}
. System variables are predefined variables that are available in every AI phone call. You can use them to access information about the current call, like the user’s phone number or the current time.
Note: Variables are NOT case sensitive, and outer spaces are trimmed automatically.
Base variables:
{{phone_number}}
- Always the other party’s number{{country}}
- The country code (ex. US){{state}}
- The state or province’s abbreviation (ex. CA for California){{city}}
- The full city name, capitalized{{zip}}
- The zip code{{call_id}}
- The unique ID of the current call{{now}}
{{now_utc}}
{{from}}
- The outbound number in E.164 format{{to}}
- The inbound number in E.164 format{{short_from}}
- outbound number with country code removed{{short_to}}
- inbound number with country code removed
Variables can be used mid-sentence, like this:
External API Requests
External API requests are defined in the dynamic_data
parameter of your call request or inbound agent configuration. The dynamic_data
parameter is an array of objects, where each object represents an API request.
Here’s a simple request that can be used to load public data about the current price of Bitcoin, then store it in a variable called {{bitcoin_price}}
:
Extracting Data from Responses
Rather than using the full response, you can extract specific data from the response using the data
parameter. The data
parameter follows JSON structuring, using dot notation and array indices. For example, if the response is:
Then $.bpi.USD.rate
would return 9,000.00
.
More complex filters can be used if they follow the JSONPath format.
Variables as Parameters
Once defined with response_data
, variables can be used nearly anywhere.
- In the
task
orprompt
parameters - In the
context
parameter ofresponse_data
- In the
body
,headers
and/orquery
parameter of each request
Afterwards, in your webhooks and when retrieving call transcripts, the variables
field will contain all variables that were defined during the call.
By far, the easiest way to test out your dynamic_data
configuration is via the /dynamic_data/test endpoint. It returns the original configuration, every raw response, and the final variables after parsing is applied.
Chaining Requests
Each request is executed in order, and variables defined in previous requests can be used in the next request. For example, if you want to retrieve information from your database or ours, then take additional actions with that data then you could do something like the following:
For this example, imagine a delivery service that offers instant checkout for customers that have signed up to be a member. The first request retrieves their member_id from your database like so:
We just created that {{member_id}}
variable - now we can use it in the next request.
This delivery service also can be called to check on an order status.
Note a difference: The cache
parameter is set to false
, so if the order status changes during the call, the agent will immediately know about it and be able to inform the customer.