Chatbot API BETA
Chatbots can assist your team and take over boring and repetitive tasks such as gathering contact information, allowing your operators to focus on helping customers. They can also be useful as a way to answer simple questions after hours.
At Userlike, we believe in the power of human-to-human contact, but recognize the value service-bots can bring to messaging. This is why we offer you to either integrate your own chatbot with our API or to design your own rule-based flow using the Userlike Bot Language.
This tutorial will walk you through all the details you need to know to integrate a chatbot with our API.
|How to set up a chatbot||How to configure your chatbot to work with our API|
|Example chatbot flow||How a typical chatbot flow can look like|
|API Overview||How our API works|
|Webhook||How our webhook calls are structured|
|Triggerable events||What events you can trigger after receiving webhook calls|
|Keeping state||How to use our API to keep state|
|Chatbot Sessions||How chatbot sessions work|
|Sample Code||Sample code for chatbots|
Chatbot setup in Userlike
We assume that you already have a chatbot (framework) up and running that you now want to connect to the Userlike chat infrastructure. We also assume that your chatbot (framework) is reachable via HTTP. All you have to do now is to create a "proxy" chatbot in our Dashboard: Think of it as your chatbot's representative (or ambassador) in the Userlike chat infrastructure.
In Userlike, bots are a special type of operator. You can create a new chatbot by going to Unified Messaging Config Operators in your Dashboard. You'll find the "Chatbots" section below the list of human operators. Click on "Add Chatbot" to create a new chatbot.
Now you can configure your new chatbot:
The first name, last name and operator group settings are identical to those of a regular human operator. The bot type and bot behavior modes are explained below.
If you want to create a bot that uses our Bot API, you have to select 'API' here. Alternatively you can implement some bot behavior with the Userlike Bot Language.
Bot behavior modes
You can configure your chatbot in four different behavior modes. These modes decide in which situations your contacts will be connected to the chatbot, or whether its human colleagues are the preferred choice.
|Bot behavior mode||Description|
|Service Time||The bot is online only outside of the service time defined on the Widget. During service time, all chats go to your human operators.|
|Failover||The bot is always online but only takes chats when no other operators are available.|
|Human||The bot is always online and takes chats exactly like a normal human operator.|
|Firewall||The bot is always online and takes all chats (e.g. to pre-categorize and forward them).|
After you have configured and saved your new chatbot, it appears in the bot overview, just beneath the human operator table.
From here you can edit it just like any other operator. After you've just added the bot, you need to edit it one time to configure the API settings.
If you edit an existing chatbot, you can additionally configure its API settings, change its profile picture and define how many concurrent chat slots your bot can handle.
This is a URL hosted by you, where we can reach your chatbot via HTTP. We will call this endpoint everyt time a chat-related event happens, e.g. when a conversation starts or when a new message comes in.
This is a unique URL hosted by Userlike, which your chatbot can reach via HTTP. You can call this endpoint any time your chatbot wants to add something to the conversation that is not a direct response to a previous request made to your webhook.
Webhook URL vs. API URL
So your chatbot has two possibilities to send messages to a conversation:
- as a direct response to a chat event that has been sent to your **Webhook URL**.
- as a self-initiated message to your Userlike **API URL**.
In most cases you will just send a response to our calls to your webhook. But when a call makes your chatbot do something that can take a while (like asking another web service), you can acknowledge the request via a response first, and then send the final result to the conversation once it is done.
API Security Token
You need to generate a security token to use our API endpoint and add it here. Each API call you make needs the token specified as "SECURITY-TOKEN" HTTP Header Field.
You can specify HTTP Headers that will be included with each webhook request we send to you. You can use this to make sure the requests really originate from us.
Example Chatbot Flow
Lets say you have a webshop and want to automatically handle some of your customers' requests.
Contact starts a chat
After the bot is set up, you will receive a call to your webhook URL whenever a contact starts a chat with the bot. See request data.
Your bot responds to the webhook request directly and sends a welcome text message, together with a bubbles element to offer the contact some choices: See response data.
This is how it will look for the contact:
Contact selects delivery status
Once the contact clicks on "Delivery status of an order", you will receive a call to your webhook URL, containing all data from the click event. See request data.
Your bot responds by asking the contact for the order id. See response data.
Contact sends order id
When the contact sends the order id, you receive another call to your webhook URL. See request data.
Since looking up all the delivery information on the order can take a while and requires a call to another webservice, your bot schedules an asynchronous job for it. To inform the contact and not keep them waiting, the bot also sends a short info message as a direct reponse to the webhook call. See response data.
Chatbot sends result
After the asynchronous job has finished, the bot makes a self-initiated call to the Userlike "API URL", sending the result of the delivery inquiry. See cURL example.
Your chatbot handled your customer's request automatically, responding to all messages immediately and delivering time-consuming results once they were done.
Whenever a contact starts a chat and your bot is selected as a conversation partner, you'll receive a webhook call with a
start event. From then on you will have the
cid of the chat session, which can be used to trigger events in response.
Whenever the contact writes more messages in the conversation, you will also receive more webhook calls, which will contain the same
cid as the
start event has.
To react to incoming events, you can trigger various events yourself. For example you can send a message or forward the conversation to a human. Those events have a similar JSON structure as the incoming ones. To trigger them you can either directly respond to the webhook call, or you can send a request to our chatbot endpoint.
Authentication and Security
We offer to specify HTTP headers when you create or edit your bot in the Dashboard (see Webhook headers), which will be sent with each webhook call. We recommend that you generate a security token and use it in the headers, so you can verify that the webhook calls really originate from us.
To authenticate on our chatbot API endpoint, you have to generate another security token, which you can also enter when you create or edit a bot in the Dashboard (see API Security Token).
If you use our API endpoint, please note that you can only send 20 requests that use the same
cid per minute. When the limit is reached, we will respond with 429 status code.
Whenever certain events happen, you'll receive a POST Request on your webhook, which contains JSON data describing the event.
The event data looks like this. Note that the contact and widget objects were left out to keep the sample shorter. You can find the full object below.
info object contains general information about the chatbot session, while the
packet object contains event specific information.
Those keys are inside the
|Unique string identifier for the current chatbot session. You will need to include this, if you want to trigger an event in response to the webhook call.|
|Information on the contact. If the contact is visiting your site for the first time, this will not contain much. But otherwise this might contain useful information. Example.|
|You can use the context object to save state for a specific bot session. You can read more about this here: How to use our API to keep state|
|Which approach mode was used to start this conversation. It is set to |
|Information on the widget that was used to start the conversation. Example.|
Those events can be inside the
packet object when your webhook is called:
|A new chat session with a bot was started. You should use this event to send a welcome message to the contact. Example.|
|A new message from the contact was received. Example.|
|A contact clicked on an option of a bubble or carousel element. Please note that options are one-indexed, so the first option is referred to by |
How to respond
We expect you to respond with a 200 status code. You can either leave the response body empty or use the response to directly trigger some events.
To respond to the events sent to the webhook, you can trigger certain events yourself.
This is how a event looks like:
It always needs a cid specified inside the
packet also always needs an event name. The
payloaddiffers for each type of event. In this case we have to provid the content of the message under
Events can be triggered by two in two ways:
- You can include a list of event objects as JSON in your response to our webhook call.
- You can also send the same JSON as a new request to our API endpoint. Example.
Those actions are available:
|Send a message to the contact. Example.|
|Send a message to the contact, together with different buttons to click on. You'll receive a click event when the contact clicks on one of the buttons. Example.|
|Similarly to bubbles, but with images on the buttons. You'll receive a click event as a response. Example.|
|Make the contact navigate to a certain URL. If the contacts browser does not allow navigation, the message from the |
|Update the contact name, email and mobile number. You only need to provide one of |
|Stop bot interaction and continue normal conversation flow. The conversation is visible for and can be routed to human operators. Example.|
|Terminate the bot session and set the 'ended' status on the conversation. Example.|
|Terminate the bot session and forward the conversation to any available operator from the given group. Example.|
|Terminate the bot session and forward the conversation to any available operator. Example.|
Keeping state with the context object
We allow you to keep session-specific state in our API.
On each webhook response or API request, you can provide a
context, which can be any valid JSON object. Just make sure it's not bigger than 10 KB.
After you set a context, we will echo the same object in each webhook request we send to you. Note that we will completely overwrite the previous context object, if you provide a new one. If you send us an empty context, we will ignore it and keep the previous one.
Chatbot sessions are started when a conversation is started on your widget, while a bot is configured with a matching behavior mode.
After the session starts, it stays open for a defined timeout without messages or other chat activity. The timeout can be configured in the widget settings under 'Live conversation timeout'.
Right now there is no way to resume an expired session.
Python Server Example
We prepared a Python sample server for a bot that asks for the contacts name and email, before forwarding him to a human.
This is how the bot interaction looks for a contact: