What is the best way to build Teams messaging extension using the bot framework without ngrock. My organization is not even allowing me to use it for testing or development. My web search pointed me to Azure Service Bus, Azure Relay and a dedicated NGINX Linux server with public IP. Just wanted to know what would be the ideal way and the costs associated it if someone has already developed an app?
Related
I am working on MS Teams development. I installed the MS Teams toolkit in VS Code, set up my subscription with Azure and sideloading is active in my tenant.
When I run the app, it tries to install ngrok. This step fails as my organization does not allow running ngrok or other words tunnelling from our company laptop. We can run this on a VM to go around this but VM is not always available.
I am looking for a resolution for below scenarios:
Is there a way to debug MS Teams application without ngrok?
If we need a https URL, is it possible to configure a web app to facilitate that?
I tried removing install ngrok step from: /.vscode/tasks.json, but there are subsequent steps it the file dependent on that
I've done quite a bit of research on this question myself as I'd been getting a lot of pushback from our IT department regarding the security threats that come with using a tunneling service like ngrok. It eventually led me to this video posted on the MS forums from a Microsoft engineer who explains it clearly.
What it comes down to is that the Teams client (browser/desktop) approaches webservices (configured in the manifest file) differently depending on the type of interaction. If you're testing configurable tabs, task modules or configuration pages, then you can easily route the app to those sites running on your localhost through the manifest. The Teams client will approach them directly. Problems start to arise when you want to debug what happens when you use a bot or message extension, outgoing webhook or MS Graph change notifications (just quoting the video here, there might be other scenarios).
Basically, what happens is that the Teams client goes through a Microsoft-hosted service first, called Microsoft Teams Services, which will then approach your bot framework cloud service (typically an Azure Bot resource). This then forwards any incoming messages to whatever endpoint you have configured. What happens in these separate stages isn't completely clear to me, but what I do know is that whatever is typed by the user in the Teams client is translated to a JSON structure that can be interpreted by your server-side bot code (for C# apps, this is typically your CloudAdapter-derived class working with your TeamsBot-derived class). These messages are then routed to the relevant TeamsBot class method based on properties in the JSON.
Now the issue that ngrok solves is that, when the Teams client goes onto the public internet to reach the MS Teams Services server and then the Azure Bot resource, it then needs a public address to route the traffic to. It doesn't know about your local network anymore. As ngrok sets up a TCP tunnel between their server and your local PC, it is able to route traffic coming to their server to your PC. The Azure Bot now has a public address to send the messages to.
To my knowledge, there is no way to circumvent this as long as Teams client inner workings always make it go outside of your local network. For chat scenarios, the Bot Framework Emulator might offer a solution for unit testing. As far as I can see it performs the translation of chat input to the JSON message model of the Bot Framework and routes it to a local address for your chatbot to process it. Unfortunately, this doesn't work for chat message extension type messages.
As for the question whether ngrok can be avoided, I think the answer is definitely yes but you would need an alternative. There's several alternatives around that you might be able to host yourself if you have the technical know-how. Depending on your IT department, being in control of the public-internet-facing server might be a more viable solution for them. Another option is to host ngrok on a VM or cloud machine with less access to your internal network's resources than your PC/laptop has and test the code there.
TL;DR: If the the feature you're testing is approached directly by the Teams client, you can enter localhost in the manifest and debug it. If you're testing a feature that the teams client approaches through Microsoft Teams Services and the Bot Framework, you need to find a way to expose your code to the public internet. You can use ngrok or host your own alternative depending on requirements.
use mkcert to generate a certificate for ex. localhost.test
add losthost.test to your host file
use https://localhost.test for debugging
I don't want to use bot framework in pop-up or chat window.
Can I use the web API I created myself.
When I do a call, I want the api to request it to list the returned data.
Can I select multiple data listed?
From your description you seem to want to create a Teams Message extention, specifically a search extention that will be used from the compose box.
this is indeed powered by a webservice that you will write and host,
however the communication channel between Teams and your webservice must use the bot framework for communication.
Note that you do not need to write a chatbot as such;
I like to think of it as the 'communication channel' between all the Teams clients that have your app installed , and your webservice.
High level steps:
Create and deploy your web service (while developing use a tunneling
service like ngrok to run it locally)
Register your web service with
the Bot Framework
Create your Teams app package
Install your App into Teams
More steps and details in the messaging extensions documentation
I want to develop a product which simplifies the way of creating conversational chat bots using Microsoft BotFramework SDK. As we can build and host web application on-premise completely.
So can we build the similar thing using bot framework on a self hosted environment?
Here is the link for Microsoft Bot Framework: https://dev.botframework.com/
You can, indeed, build a bot entirely on-premises/self-hosted, provided that you don't need to connect to a Channel like Teams.
Background
Basically, bots communicate via:
User interaction with DirectLine Client (like Web Chat) ->
DirectLine API receives data and sends to bot ->
Bot hosted anywhere receives message
Note:
WebChat is just a DirectLine client. If you want to build your own WebChat, you can clone the repo and build it yourself. However, that all this does is make it so you can host your own webchat.js file, if you really want to be self-hosted.
Steps
With that in mind, all you need is:
A Direct Line client like Web Chat (specify domain property with your own endpoint when you call createDirectLine), or you can write your own client
You own DirectLine API implementation like offline-directline (more info)
Host your bot somewhere
Security
With this approach, you won't be able to use the typical appId/password setup.
You can manually implement our security protocol by following this document
Even though it is possible to manually implement the security protocol, we recommend reusing our existing open-source libraries
Based on answers, putting a more complete answer here. Since the option exists to have internet connectivity, I'd definitely suggest creating the solution as a "regular" Bot Framework bot, using the Azure Bot registration. Importantly, there are TWO options to do this - the most obvious one creates a "registration" for the bot but ALSO defaults to hosting it in an Azure Web App, which has a cost associated that can be quite significant. HOWEVER, there's another option, to create just a "Bot Channel" registration, which means you get a bot registered in Azure, but not necessarily hosted in Azure.
I've described this more in this post and you want to use the "Bot Channels Registration" in Azure, not the "Web App Bot". Then, on the "settings" screen you can define what the bot's actual address is (see here for a view of the Settings screen).
So, overall that should help - basically the bot just needs to have an HTTPS, internet-accessible endpoint. HOWEVER, that said, it's possible to host your bots inside Azure Functions, rather than Azure Web Apps, and the cost is dramatically cheaper (cents vs dollars, especially with low traffic). It's a bit of work, especially as the default samples default to the more standard options, but it works quite well.
I've recently launched a new blog, by the way, and I've got a post describing some of what I've learned so far about how bots work, that would be good to know as part of this. See How Bot Calls Actually Work.
We're in the process of migrating from an entirely on-premise solution with our own DirectLine server to Azure Bot Channels Registration because of the requirement for MS Teams integration.
This question has probably less to do with actual programming and more to do with environment setup.
I'm developing an application for a Windows CE device, which will use a service hosted in Microsoft Azure. Obviously, this is all under development, and the service itself has not yet been uploaded to Azure. So I'm using the emulators provided by MS to deploy the service locally. Also, I don't think uploading the service to Azure just to debug it is a good idea, as that could net us a bill for Azure we don't yet want to pay.
Also, please note, I'm using VS2008 for the Windows CE project and VS2010 for the Azure project (thank you MS for dropping support for mobile devices -_-).
The problem is, the service seems to be accessible only via 127.0.0.1 or localhost, and if I physically connect a Windows CE device or use an emulated one, it becomes a new entity on the network, and cannot access that service any more.
How can I debug my Window CE application and have it see the service, whilst still being able to debug the service itself?
You are correct, the development fabric (the compute emulator that allows you to build an azure application and debug it locally) is only meant for local development. There are some hacks that allow you to get around that, but I wouldn't recommend it.
My recommendation would be to spin up the service in a more traditional hosting environment, at least in these early stages. You can define it as a web app just as you always would have, and get it functioning. Then, when you know its mostly complete, create a cloud service project and do an add existing to bring your web app into the cloud service solution. Once in, its a simple matter to add the web app as a web role.
From there, you can complete testing of the service in both the local and hosted azure environments as needed. This allows you to minimize your development costs while still leveraging the power of the cloud. As an upside, you also have done most of the basic work to ensure your service is compatible with multiple deployment scenarios giving you a greater degree of choice for its final production state.
OK, I don't know if this was intentional, or if I found one of the mentioned hacks, but I saw that IIS hosts the Azure site I created on port 5100, and the binding for this site is *, so it accepts all connections.
Using this I could access the service from my emulator, and I could still debug all Azure related stuff.
Real application how can i use this app fabric ??
How can i put my business logic in this and this logic use in my windows azure application??
Thanks
The Azure AppFabric is a collection of services that allow to you leverage functionality traditionally provided on premise by infrasture components common to most networks. Currently, it consists of the following:
Azure AppFabric Service Bus - allows for connection of applications by providing a centralized relay point in the cloud. Applications create outbound connections to the rendezvous location, thus helping mitigate the challenges posed by security measures like firewall restrictions on in-bound connections and IP masking via NAT layers. This feature includes both 'real time' options as well a 'message buffer' dynamic to allow for more disconnected style communication.
Azure AppFabric Access Control Service - the "ACS" allows WIF applications to quickly access various identity providers and consume a single format of claims token. Used in conjunction with products like ADFS, it allows cloud hosted applications to authenticate against on-premise identity stores.
Azure AppFabric Cache Service - currently in public testing, this service brings the "Velocity" style functionality to applications. This provides them with a distributed cache system as well as a new session provider.
There's more features/services coming in 2011, but these are the hot ones currently. Regarding hosting your business logic, this is not something that is currently available in the Azure AppFabric. There's been mentions that we may eventually see the potential for placing applications "on the edge", meaning the servers that front the Azure AppFabric connections, but no ETA or even firm commitment that this will happen.
You can implement your business logic in Windows Azure, in a web or worker role depending if you need it to be synchronous or asynchronous.
You can surface the business logic using the service bus, though you could also implement your logic on premise and surface them via the service bus.
AppFabric is not a business logic layer. Think of AppFabric as cross-cuts, or glue between different parts of your application.
For now Business logic goes in components like a web or worker role, or an on premise app which you could expose on the internet using AppFabric Service Bus.
In a future release, AppFabric will release "Composite Apps" which in a nutshell seem to allow you to deploy managed WCF/WF workflow services, which makes for a better "business engine". But for now I think you could probably just use Workflow services in a web role.