So I finally figured out a way to successfully make detect intent calls and provide an input context. My question is whether or not this is the CORRECT (or best) way to do it:
(And yes, I know you can just call DetectIntent(agent, session, query) but I have to provide a input context(s) depending on the request)
var query = new QueryInput
{
Text = new TextInput
{
Text = model.Content,
LanguageCode = string.IsNullOrEmpty(model.Language) ? "en-us" : model.Language,
}
};
var commonContext = new global::Google.Cloud.Dialogflow.V2.Context
{
ContextName = new ContextName(agent, model.sessionId, "my-input-context-data"),
LifespanCount = 3,
Parameters = new Struct
{
Fields = {
{ "Source", Value.ForString(model.Source) },
{ "UserId" , Value.ForString(model.UserId.ToString())},
{ "Name" , Value.ForString(model.FirstName)}
}
}
};
var request = new DetectIntentRequest
{
SessionAsSessionName = new SessionName(agent, model.sessionId),
QueryParams = new QueryParameters
{
GeoLocation = new LatLng {Latitude = model.Latitude, Longitude = model.Longitude},
TimeZone = model.TimeZone ?? "MST"
},
QueryInput = query
};
request.QueryParams.Contexts.Add(commonContext);
// ------------
var creds = GetGoogleCredentials("myCredentials.json");
var channel = new Grpc.Core.Channel(SessionsClient.DefaultEndpoint.Host, creds.ToChannelCredentials());
var client = SessionsClient.Create(channel);
var response = client.DetectIntent(request);
channel.ShutdownAsync();
return response;
Note: I included the explicit ShutDownAsync
(it's not in an async call) because I was getting some file locking issues when attempting to re-deploy the WebAPI project (and only after having executed this code).
Thanks Chris
Updated 4/25: The most basic way I use this is to integrate the user's name into intent responses:
It can also be read from within the webhook/inline fulfillment index.js:
const name = request.body.queryResult && request.body.queryResult.outputContexts && request.body.queryResult.outputContexts[0].parameters.Name