So I am testing out Dialogflow and one of the first questions I have is: how does my bot know who it is talking to? I need to identify a user and keep that information for as long as I can. The basic scenario being:
User starts his/her first conversation.
Chatbot send a fulfillment request to the server trying to match a user within its own database.
The user is found, the information (as a JWT or some other token) is sent back to Dialogflow and stored there for further communication. In reality, this part would involve asking for user email, sending a verification code to that email and then verifying the user with the code.
User then starts chatting with a bot and all fulfillment requests get the unique token stored for this very user, so that my REST API knows which user is being served with the response.
Couldn't find anything about it in the docs (maybe I am looking in the wrong places).
There will be several integrations, like Messenger, Viber, Telegram. I dunno, maybe those APIs add some unique information on the user?...
Thanks for the help!
Sorry, I know it's been a while, but maybe this will help someone else.
The right solution here is a user id, not a session id. A user id is provided by the chat platform (Facebook, Slack etc) and is consistent across sessions for the same user.
To get the user id, go to the Fulfillment tab, enable the editor and use a function like so:
let r = request.body.originalDetectIntentRequest
//this makes sure that you're on an integration
if (r["source"]){
return r.payload.data.sender.id;
}
To tie together ids from different platforms, you probably have to have some kind of log-in process every time you encounter a new id on a platform.
Pop,
Sessions are built in already into DialogFlow requests to your fulfilment service, if you check the payload you will find a sessionId, it remains the same for the same client until it expires.
However if you want to identify the user from any of the clients that you can connect to DialogFlow like Messenger then from the same request payload to you you will notice that there is an object named originalRequest that is only available when requests are coming from those clients.
You can personalize those users response eg using their FB firstname in a message to them.
Related
I have to design an e-learning tutorbot that has multiple possible channels living in parallel:
A live chat (to support learner with FAQ when browsing the e-learning application website)
A IM chat as Telegram (for asyncronous notifications)
An email (for anything else)
...
Take a part the connectors gateway communication logics.
My problem/question is mainly about how to manage user identification.
I want that a user is identified with an unique identifier (call it conversation_id) regardless of channel.
In the mentioned scenario user has one different ID for each channel:
An account_id on the website (e.g. solyarisoftware)
A chat_id on Telegram (e.g. 1234567890)
An email_address (e.g. giorgio.robino#blabla.com)
So, how to let RASA chatbot accept requests coming from the same user (but from different channels)? In other words, how can I map multiple channel IDs into a RASA unique ID, say the conversation_id ?
At the end of the day I need to generate a lookup table that map a UUID with all channel ids. Right?
But is not clear to me what's a good (simple) user experience. At first glance seems to me that that I need a sort of login/registration flow: each time user submit a request on channel X, only for the first time, the user must identify himself somewhere (with his email/account_id?). So by example trough Telegram, when the Telegram connector server receive a /start command, the bot must ask the user email(or account_id), to be able to associate 1-to-1 the TG chat_id with that email. Not perfect in terms of security, I admit.
Any experience/idea/suggestion?
Thanks/giorgio
I don't have an hands-on experience on this but it is something I have been thinking myself for a while.
Indeed the 'merge' of the different accounts (across channels) is something that we need to maintain ourselves, making sure each channel user ID is eventually mapped to a universal user ID (lookup table makes sense to me).
Since you have a website, an email address and multiple channels I would say the assumption is that users first registered on the web site. Each user has a unique email address and obtains an account secured by username/password, as well as a unique secret token.
When chatting on browser (ie Rasa WebChat) you can initialise the plugin with custom parameters (ie userId=1 or token=X), in this case the chat session is assigned to the given user id.
When using Telegram (or other messenger application) I would (at the very first access) ask the user if he/she has already registered. If yes then I would ask the token: this can be used to perform the map and link this Telegram user to the universal user Id. Same approach for other channels.
Depending on your needs you might need to deal with anonymous access (users don't provide the token) or maybe you want to provide a different token for each channel (more secure, but requires more work and an easy way to explain this to users).
Finally something to look at (I haven't tried myself) is the possibility to provide in the web site the option to open Telegram (or others) passing an extra parameter, which would allow you to perform the mapping. I don't know if this is possible, but it would simplify this process for your end users. Or the other way around: the users in Telegram could be provided with a link which opens the website, where they can login and be recognised.
I am implementing subscriptions to a premium service on a website using Paypal as the payment service. I have successfully created a Catalog Product and Billing Plan through the API, and I am able to get to the payment page on Paypal, but it's not clear how I'm supposed to persist a user identifier through the purchase process.
I assumed it would be something along the lines of passing a user id somewhere, but there's nothing in the Paypal documentation about this. I need to be able to let the user make a purchase and have the Paypal webhook send the confirmation to an endpoint on my site, and that's where I'd expect to get their user id to toggle the subscription on their account on my end.
Is there something I'm missing? There has to be a way to do this cause I'd imagine it's a pretty common use case. If anyone has information or has done this before, I'd love to hear. Thanks.
The only truly secure way I've found when using javascript SDK, is to securely generate a unique custom_id on your server side associated with the user.
Then when you create the buttons, the 'createSubscription' function takes custom_id as a parameter.
Then use a webhook to receive events from your subscription and the custom_id will be present in the body of all BILLING.SUBSCRIPTION events under resource.custom_id.
I am able to get to the payment page on PayPal,
You are vague about what you are doing here. There are multiple ways (and some ways have multiple versions) of accepting subscriptions via PayPal, so it is important that you provide full details about the method you are using.
The time to associate a created subscription ID with a user ID is when it is approved, in the onApprove function if you are using a Smart Payment Button: https://developer.paypal.com/docs/subscriptions/integrate/#4-create-a-subscription
Recently google has added multiple user support to the assistant so how would use the API to identify the person by voice?
It depends what you mean by "identify the person".
There is no way for an Action to get the raw audio, so there is no way for it to do voice printing or anything along those lines.
Although each voice has to be reported against a Google User ID, you do not have direct access to that user ID.
What developers do have access to is a UID that is sent along with each request to your fulfillment server. This UID is consistent across requests, although it can be reset by a user (for example, if they reset their Google Home). You can think of this the same way you think of an HTTP cookie - you can track the UID and, if you see it again, have reasonable assurance it is the same user that accessed it last time. This breaks down, however, for the "default" account on Google Home, since anybody who doesn't have an account will map to this user.
Beyond this, you can also use Account Linking to connect a Google Account consistently to an account in your own system. If you have sufficient authentication in place, or are using one from Google or Facebook for example, this can act as an identity.
There isn't an API for developers to identify users by voice.
Since apparently you're not reachable for API questions via email, I guess I'd have to put all the questions in here since it would be tedious to create one for each of them.
So we're planning on creating a webhooks system for our integrations to create a better experience for our users. There are a few pieces that are missing for our end to support having a Podio integration that runs 100% on webhooks.
Support organization level webhooks instead at the space level.
Send full object payload in the webhooks push so we don't need to fetch the resource again. Or at least provide a delta of changes.
Have a secret token header that you send with each request so we can assure the authenticity of the data (that it's actually from Podio and it's not someone trying to hack our system).
Github does a great job with this: https://developer.github.com/webhooks/#delivery-headers.
Ability to unsubscribe from a webhook if the client revokes the OAuth credentials. Do you automatically delete that webhook or do we need to contact the user or you guys to get rid of it?
Trello for instance deletes all webhooks associated with a token if it's revoked or a second option is our server sends a 410 Gone back and the webhook, well you guessed it, it's gone.
Do you think something like this is feasible in the near future?
Thanks for your insights :)
Seems pretty valid point. Are there any particular org operations that you are interested to have webhooks for? Can you also describe some use-cases for org level webhooks?
Sending full object payload is pretty complicated because items in Podio can be extremely big and it might be not needed for webhook to get all details. That's why we keep it small and tiny and let Podio partners build own complicated solutions based on webhooks.
Secret token is not much needed if there is only item_id sent in webhook. Let's review it on item modified scenario.
1) User (or anything else) modifies item
2) item.modified webhook is triggered with item_id
3) 3rd party app receives webhook
4) 3rd party app makes a call to Podio API (if it's needed) and gets all required information from Podio API
So, if webhook request is fake, then Podio API will return same item details and nothing is affected. If you want to have some extra layer of security: why not making webhook URL unique and specific to webhook? Then only Podio will know which URL to use to trigger this webhook.
Ability to unsubscribe from a webhook if the client revokes the OAuth credentials.
Sorry, but I don't understand your question. Can you please re-phrase it as new StackOverflow question?
This is an authentication flow, which logs in the user with only the phone number provided (Whatsapp style). The Steps are:
User enters phone number and sent to server.
Server generated 4 digit random key, and save the pair (phone,key) in DB.
Server asks 3rd party SMS service to send key to phone.
SMS service sends message.
User enters the key from sms, and together with phone, sent to server.
Server checks the pair against the DB.
If pair exists, server sends back a token for further calls.
What I am trying to understand is where services like Twilio and Nexmo fit in (or replace parts of the flow).
From what I understand, looking at Nexmo for example,
I can replace steps 2 and 3 with an API call to:
https://api.nexmo.com/verify/
and save the request_id from the response in the DB as pair (phone,request_id).
And now, when the user enters a 4 digit code and sends it back,
I need to call:
https://api.nexmo.com/verify/check/json
providing it with request_id and code.
But where do I get the request_id?
Do the server needs to send it back to the client, the moment it gets it from Nexmo?
I can't see the benefits of using Nexmo here, what will it save me?
Answer to your first question: the request_id is part of the response to the first verify API call. See: https://docs.nexmo.com/index.php/verify/verify
As to what are the benefits of Nexmo here, I believe you have two options:
Generate your own code, use Nexmo to text it to your user, have the user submit the code back to your application, verify code against your own database.
Use Nexmo verify service to generate and send the code to the user, store the returned request_id in your db, have user submit code to your application, call Nexmo verify API to validate code.
In some ways the first option is easier as it is less API calls. However the benefit of the second option, using Nexmo Verify, is that they provide a whole lot more capabilities into the service to fall back to a voice call if SMS isn't working, filter out virtual phone numbers to prevent spam, you don't have to pay for failed SMS attempts, reporting/analytics, etc. etc.
Hopefully that helps a little.