I have a Google Action that returns a basic card as part of a response - e.g.
"basicCard": {
"title": "A bunch of text for the title of this lengthx.",
"formattedText": "*A bunch of formatted text of the following length xxxx xx xxxx, xxxxxxxxxxxx xxxx.",
"image": {
"imageUri": "https://domainname.com/imageurl.png",
"accessibilityText": "Card Image"
}
The card is returned on each call.
Just recently on a Google Nest Hub, I started seeing the following behavior:
We launch the action via “talk to [actionname]” which then asks a "Yes/No" question
After the user says “Yes” when the next response is delivered, it starts to display the associated card (same as the one in the first response) and the following happens:
As the card scrolls into place, the screen locks up
The device sits for about 10 seconds and then starts reading the response (on the server, we’re returning responses in the 150ms range)
It then shows the suggestion chips, but takes in no voice input and eventually shuts down
The action runs without issue on screenless devices, the test harness and on a phone so the issue looks localized to the device.
Has anyone seen anything similar? We have other actions where the cards vary text and images and those are not having any issues. As I said before, this was working fine a few weeks ago.
Thanks,
Sanj
Get Dialogflow agent fulfillment library (NodeJS) by using the following command:
yarn add dialogflow-fulfillment
OR
npm i --save dialogflow-fulfillment
Then require it in your project and add image using a URL:
const { WebhookClient, Image } = require('dialogflow-fulfillment')
const agent = new WebhookClient({ request, response })
agent.add(
new Image('https://avatars1.githubusercontent.com/u/36980416')
)
Related
I have the Plaid Quickstart (node) up and running.
I successfully logged in to one of the sandbox institutions (First Platypus Bank) using the sandbox credentials and got the access_token which it generated.
Per the Link Update Mode docs (https://plaid.com/docs/link/update-mode/), you should be able to "force a given Item into an ITEM_LOGIN_REQUIRED state".
From the docs I linked to above...
Update mode can be tested in the Sandbox using the /sandbox/item/reset_login endpoint, which will force a given Item into an ITEM_LOGIN_REQUIRED state.
I used Postman (in the Sandbox Public environment) to send an API call to this endpoint, https://sandbox.plaid.com/sandbox/item/reset_login, to force the ITEM_LOGIN_REQUIRED state.
This is the body sent with the API call...
{
"client_id": "{{client_id}}",
"secret": "{{secret_key}}",
"client_name": "Insert Client name here",
"country_codes": ["US"],
"language": "en",
"user": {
"client_user_id": "unique_user_id"
},
"access_token": "{{access_token}}"
}
This is the response I got back...
{
"expiration": "2021-02-13T04:13:12Z",
"link_token": "link-sandbox-7d82e9b8-b8a7-4977-9c20-aadbc82ec050",
"request_id": "eE56J8f5XJKpS6L"
}
...which is what I expected.
I then went back in to the Quickstart app and logged in to the "First Platypus Bank" again expecting to be prompted for the MFA challenges, but, instead, it just created a new access_token.
So, how to I get the Plaid Quickstart app into Link Update mode?
Update:
Per Alex's question, here is an image showing the Postman call I made. He's correct, it was a /link/token/create call. What's confusing it that it's labeled Create Link Token - Update Mode.
Update 2:
Here's the /sandbox/item/reset_login call and response. There's no link_token included in the response.
Update 3:
Between Alex's help and some help from Plaid support, I got this working.
I was putting the link token in the wrong place.
For anyone seeing this in the future, here are the steps you need to follow to get the node quickstart app into Link Update Mode.
You can also watch this video to see a demo of the process; HOWEVER, note that in the video, I put the link_token in the wrong place. The instructions below show how to put it in the correct place.
Get the access_token for the institution you want to get into Link Update Mode (this assumes you have the quickstart app installed, up and running. See... https://plaid.com/docs/quickstart/)
Start the quickstart backend (/quickstart/node/start.sh)
Start the quickstart frontend (cd /quickstart/frontend; npm start;)
Once the frontend fires up on port 3000 in your browser, click the "Launch Link" button then click "Continue".
Search for the institution (e.g., First Platypus Bank, First Gingham Credit Union...if in sandbox mode. or your own institution if in dev mode)
Enter your credentials (user_good/pass_good ...if in sandbox mode)
Copy the access_token
Close the Plaid browser tab
In Postman (to set this up, see... https://github.com/plaid/plaid-postman), enter the access token you just copied into the access_token variable (See the first screenshot above. Click on the icon near the upper right-hand side that looks like an eye then scroll down to access_token and enter it.)
Run the /sandbox/item/reset_login endpoint. It's labeled "Simulate ITEM_LOGIN_REQUIRED [Sandbox only]" in Postman. The request body should look like second screenshot above. Once you Send the request, the response body should show "reset_login": true. Now the Plaid Item (i.e., the institution) is in ITEM_LOGIN_REQUIRED mode.
Run the /link/token/create endpoint. It's labeled "Create Link Token - Update Mode" in Postman. The request body should look like the first screenshot above. Once you Send the request, the response body should give you a link_token. Copy it for the next step.
Put the link_token in the Plaid quickstart app.
On line 23 in /quickstart/frontend/src/App.tsx, replace data.link_token with your link_token. So this dispatch({ type: "SET_STATE", state: { linkToken: data.link_token } }); becomes this dispatch({ type: "SET_STATE", state: { linkToken: "link-sandbox-704ef648-2acd-44a2-867b-ea1258e9205c" } }); (Use your own link_token, of course, and don't forget the quotes around your link_token.)
Restart the backend and frontend. When you launch Link, it should now be in Link Update Mode.
The process you describe sounds right, but the API request and response bodies in your post are the request and response for /link/token/create, not /sandbox/item/reset_login. Can you verify that you didn't accidentally call /link/token/create instead of calling /sandbox/item/reset_login?
I have integrated Kommunicate chat bot into my website, however, a lot of traffic is generated on my website, due to which a lot more users are chatting than I have the bandwidth to support (agent wise).
I can't seem to find a way to limit the number of currently active chats one agent/human can handle at any given time. I wish to find a solution for the same.
Can this be done through the webhook integration provided? If so, how?
The solution indeed lied in adding the webhook integration.
First, I spin up a simple flask server, serving a single endpoint: /webhook.
app = Flask(__name__)
#app.route('/webhook', methods=['GET', 'POST'])
def webhook():
logger.debug('Webhook Triggered') #-> we know it's being trigerred.
resp_generated = make_response(jsonify(results()))
logger.debug(resp_generated)#-> always shows 200 anyway.
return resp_generated
if __name__ == '__main__':
app.run(host ='0.0.0.0', port = 5000, debug = True)
Then I use ngrok to create a tunnel to my local server (I plan to host it on GKE in the later stages)
ngrok http 5000
This gives me an HTTPS URL to my Flask server such as https:\\534bbe.ngrok.io
Then I go into DialogFlow -> Fulfillment -> Enter my Webhooks endpoint there:
[Note: Hit the save button at the bottom of the page]
You would have to enable webhook call for the Intents on which you are going to add your server logic, in my case it was when I wanted to transfer to a live agent while limiting the number of chats:
To ensure that the default behavior of Kommunicate is not broken when my server goes down, I have added custom payload (as shown in the image above):
{
"metadata": {
"KM_ASSIGN_TO": ""
},
"platform": "kommunicate",
"message": "---- Redact that sweet sweet company Information. Yeah!!---"
}
[Note: Make sure to hit Save on the top right, once you make your changes in the Intent.]
Then I added the Webhook URL (same as the one used in DialogFlow) in Kommunicate -> Settings -> Developer -> Webhooks
[Note: Hit the save changes button at the bottom of the page]
Once everything is set up, you will start receiving messages in the server and can add your logic as you wish. Do note that this is super tedious as you would have to read a lot of documentation and add a bunch of logic to get it to work as you want it to.
I had written a node js server code so that dialog-flow can connect to my server through web-hook and integrated dialog-flow phone gateway. I am able to get the voice transcript text when made call to that number. But if i type any number on dial-pad i am not getting it. Is there any thing i should do to get the number input from dial-pad on call.
The response is something like this if i say 'hello' on call.
{
"responseId":"b5644842-75bf-41f3-bf84-cf7b4f02c71f",
"queryResult": :{
"queryText":"Hello.","action":
"input.welcome",
"parameters":{},
"allRequiredParamsPresent":true,
"intent":{
"name":"projects/maps-ec9e1/agent/intents/54f5f623-2011-4d95-92a7-b8e873cd55a7",
"displayName":"Default Welcome Intent"
},
"intentDetectionConfidence":1,
"languageCode":"en-us"
},
"originalDetectIntentRequest":{
"source":"GOOGLE_TELEPHONY",
"payload":{
"telephony":{
"caller_id":"REDACTED_IN_STANDARD_TIER_AGENT"
}
}
},
"session":"projects/maps-ec9e1/agent/sessions/l0vQqOpaSsq9rvJc0sMMYA"
}
}
You are looking for DTMF detection. Google's Dialogflow Phone Gateway does not support this, but you can get this capability from some third party Dialogflow Gateway products like Voximplant, Audiocodes' VoiceAI Gateway, and drachtio.org.
There is no standard method for how to do this, but generally the telephony gateway needs to detect a digit and then sends a text query or event to Dialogflow where your bot can handle it.
I wrote reviews at cogint.ai with details on how to do this for Voximplant and Audiocodes.
Hello i'm newbie and im hardly to understand this notification in service-worker, and because my knowledge isn't good yet then probably i will unable to explain my problem clearly.
so here's the code :
// triggered everytime, when a push notification is received.
self.addEventListener('push', function(event) {
console.info('Event: Push');
var title = 'New commit on Github Repo: RIL';
var body = {
'body': 'Click to see the latest commit',
'tag': 'pwa',
'icon': './images/48x48.png'
};
event.waitUntil(
self.registration.showNotification(title, body)
);
});
this is the code that trigger to POP the notification, what I do not understand is where the argument to accept/ receive the data ?
I've been searched a lot: https://auth0.com/blog/introduction-to-progressive-web-apps-push-notifications-part-3/ ,
https://developers.google.com/web/updates/2015/03/push-notifications-on-the-open-web
there's some new data JSON or from git-server or push api, but I still hardly to understand where's to accept the data.
sorry if you still do not understand what's my problem.
Here to make it simple what I want :
Let's say i make a button, and everytime i click the button it will value as 'True' and I want that 'True' value to pass into argument and trigger the push of notication in service-worker.
2nd questions: am I able to trigger notification with header or text in html ? since we can manipulate the text with DOM ?
am I able to trigger notification without GCM, or API cause I just want a simple notification in serivce-worker like above without passing much data.
If you give more advice or maybe notification without service-worker but real time , I am surely happy to read it but I hope Im able to understand.
There are basically two concepts involved that work well together but can be used independently. The first is the visible UI shown to a user that tells them information or prompts them for an action. The second is sending an event from a server to the browser without requiring the user to currently be active on the site. For full details I recommend reading Google's Web Push docs.
Before either of those scenarios you have to request permission from the user. Once permission is granted you can just create a notification. No server or service worker required.
If you want to send events from a server you will need a service worker and you will need to get a subscription for the user. Once you have a subscription you would send it to a server for when you want to send an event to that specific browser instance.
Once you receive a push event from a server you display the UI the same as in the first scenario except you have to do it from the service worker.
Microsoft Bot Framework messages with buttons in Facebook Messenger
My question relates to the question linked aboved. I am writing a bot using node.js that does not use the bot builder sdk. I manually returning a compatible response for the ms bot connector service. This is working fine for a text response, but I wish to return more complicated responses, e.g the buttons/carousel you can return with messenger. Based on the question I linked above, I guessed the format and added the below:
response.attachments = [ { "Title": "Choose One: ", "Actions": [{ "Title": "Postback!", "Message": "Postback from button" }, { "Title": "Postback2!", "Message": "Postback2 from button" }] } ];
The top level title seems to do nothing but the actions render as postback type buttons correctly (they send the Message as the postback content). With messenger you also have the option to return url based buttons, and image urls.
As far as I can tell there is zero documentation on returning attachments using the node bot builder sdk. If there were I'd just write the bot with the sdk in order to obtain the response format.
So my question is, does anyone know how to correctly return both postback and url based buttons to the bot connnector service, including accompanying images, with or without the bot builder sdk?
Update 05/05/2016
So I found the link below and you can see a definition of the attachments property:
http://docs.botframework.com/sdkreference/nodejs/interfaces/_botbuilder_d_.imessage.html
If you follow it to the IAttachment specification, it makes me wonder how/why my code above works at all? As a test of that format I wrote in the following:
var att = {};
att.content = "I am content";
att.contentType = "text/plain";
att.contentUrl = "http://www.google.com";
att.fallbackText = "I am fallback text";
att.text = "I am text";
att.thumbnailUrl = "https://pbs.twimg.com/profile_images/638751551457103872/KN-NzuRl.png";
att.title ="I am title";
att.titleLink = "http://yahoo.com";
Now in slack I get a fairly nice output from this:
However in messenger I get "Service Error:Value cannot be null. Parameter name: source"
I found the info I needed. Not sure if it hadn't been published at the time or whether I was just hunting in the bot builder docs, but it's all detailed fairly well below.
http://docs.botframework.com/connector/message-actions/#navtitle
You have to tweak your message a little for certain integrations, e.g Skype doesn't really seem to support attachments.