Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

This page provides an overview of how to select deployment channels. 

Overview of Deployment

...

Channels 

As of now, there are 3 options for deploying your OCS chatbot, with more options coming soon. These 3 options are:

  1. OCS Web

  2. WhatsApp

  3. Telegram

To select a deployment channel, simply navigate to the home page of your chatbot (experiment), select the "+" button and choose the channel you'd like to use. Note that "web" is automatically offered and so does not need to be selected. (You can test this by navigating to the top right-hand menu and select "public session" to interact with your bot as if you were a user without access to OCS). Image Removed

...

OCS Web

Deploying an OCS chatbot on the web is a convenient way to share chatbots, especially with individuals or teams who will likely use the chatbot on their laptop or desktop computers. (Other channels such as WhatsApp and Telegram would be better suited to users interacting with OCS bots on smartphones). Using an OCS bot on the web also allows for consent forms and pre- and post-surveys to be made visible to users as they progress through the chatbot. 

WhatsApp

WhatsApp is an ideal channel to use, given its large user base and its popularity as an application, especially in LMICs. However, there are multiple steps involved in procuring a WhatsApp for Business number that you can use for your chatbotan organisation will need in order to use a chatbot on that number. There are also multiple providers you can approach to purchase this line. from whom your organisation can purchase this line. Each interaction on a WhatsApp Business number is also charged, where the exact cost depends on a mix of factors including the volume of messages, geographies and type of message (as categorised by Meta).

Currently, these are supported providers:

Turn.io

The steps to link your Turn.io account to OpenChatStudio and forward WhatsApp messages are as follows:

  • Get an auth token from Turn.io by going to Settings > API & Webhooks.

  • Create a messaging service with this auth token on OpenChatStudio by going to teams and creating a new messaging service.

  • Create a whatsapp channel on your experiment of type "turn.io". After you created the channel, a URL will be provided.

  • Use this URL to create a webhook on Turn.io. Go to Settings > API & Webhooks to do this.  Note that this URL is unique per experiment. Once you link the Turn.io channel type to another experiment, you'll have to update the webhook as well.

Please note that the turn.io integration only supports text messages at the moment.

If you'd like to use WhatsApp as a deployment channel, please contact ocs-info@dimagi.com for more information and support. 

Telegram 

Among multiple advantages, deploying on Telegram is a fairly simple process that does not require approvals (as above for WhatsApp). Additionally, as of this writing, Telegram is free to use, and thus makes for an ideal tool for iteration. One drawback of Telegram is that it is not as widely used as WhatsApp, and therefore might not be an ideal channel if you are hoping to scale organically to many users, especially in LMICs. 

 In order to deploy your OCS bot on Telegram, first it is necessary to create a chatbot on the Telegram app itself. Review the following steps to do so:

  • On Telegram, search for "BotFather". 

  • Once found, send a message to "BotFather" and follow the steps outlined to create a new Telegram channel. 

  • Give this channel a name as if it were your chatbot- for example if your chatbot persona is called Bob and your chatbot is called "Coach Bob", you can label your Telegram channel "Coach Bob". 

  • Once your channel is created, save the "token" provided. 

  • On Open Chat Studio, select Telegram as an option and add this token. 

  • Final step: Test your chatbot by talking to it on Telegram. 

...

Facebook (coming soon)

Although Facebook is visible as an option, please note this feature is not yet functional.