Auto-Classify Images in SharePoint Online library via Flow for Free

Microsoft Flow's most recent update added ability to query and update SharePoint File property.  This is actually really timely, as I wanted to combine this with a few other techniques and built a Document Library Image Auto-Classifier Flow.

Is that a clickbait headline?  Well it's totally real, and we'll build it in a moment.

Steps:

1. Set up your cognitive service account (understand the free bucket)
2. Set up a SharePoint Online document library with Categories
3. Set up the Flow file loop
4. Do a fancy JSON array to concatenated string projection operation with Select and Join
5. Viola, no code.  And pretty much *free*

This is part of a series on Microsoft Flow

Set up your Azure Cognitive Service instance

Follow these simple steps to create an Computer Vision API Cognitive Service in your Azure.  Computer Vision API has a free tier.

1. Create Computer Vision API

View fullsize

2. Scroll down and hit Create

View fullsize

3. Give this service a name, set up the region and select Free pricing tier

4. You need the endpoint url here

5. Also, copy the Name and key 1

You will need the "Name" and a "Key" for the next step.

The free tier of Computer Vision API - first 5000 transactions free per month.

Note the service isn't available in all regions.  Most of my stuff is in Australia East, but for the Cognitive Service API it has to be hosted in Southeast Asia.  YMMV.

View fullsize

Then we need to set up the connection in Flow

1. Find the Computer Vision API action

View fullsize

2. Enter service name, key and the root site url to set up the initial connection

View fullsize

3. Created correctly, you get an action like this

Set up the SharePoint Document Library

My SharePoint document library is very simple - it is just a basic document library, but I added an extra site column "Categories". This is an out of the box field, and is just a simple text field.

This is a simple step

Set up the Flow

I trigger the flow with a Scheduled Recurrence that runs once per day.
Using the new Get Files (properties only), I grab a list of all the files in a document library.
I then run for-each on the list of files.

View fullsize

Inside the for-each, I have a condition that checks if the Categories field is null.  If you type null directly into the field, you will get the string 'null'. 

Tip: To actually get the formulat/expression null, select Expressions and type null there.

View fullsize

If the Categories is null, then we proceed.

Grab the file content via Get file content
Call Computer Vision API with the image content.  Select the Image Source to binary, instead of URL.

Tip: I use a compose to see the debug results

I'll explain the array projection in the next section.

View fullsize

Select projection: JSON array to String array

We have an array of JSON objects:

[{
     'name': 'foo'
},
{
    'name': 'bar'
}]

This default UI maps to:

tags -> [{ specified properties }…]

The result is that we would end up with a new array of (simpler) JSON objects.
Hit advanced text mode.

Here, we can use Expression to say item('Tag_Image')?.name

In this case the UI is smart enough to show Tag.Name as a dynamic content (as well as the Tag.ConfidenceScore property).  So we can select that.

This performs a projection of

tags -> [ names… ]

We now have an array of strings.  Combine them via Join with a comma (,) separator.
Update the file properties with this string.

Lets see the results

I uploaded a few images to the library.
Note the categories field is blank.

View fullsize

Running the Flow

View fullsize

When it finishes, I'm checking the JSON - the picture is identified with a "person" with 99% confidence.
The combined string "person,young,posing" is updated into the File property.

View fullsize

The documents are updated.  When Flow runs tomorrow it will skip them.

View fullsize

The Final Flow

View fullsize

Flow to MS-Todo, then all your tasks to Flow

I'm in a celebrating mood - Flow released Export and Import.  This is a great day.  https://flow.microsoft.com/en-us/blog/grow-up-to-logic-apps/

So I'm going to write about something that I've built in the last week, because I've been looking at a simple Todo app - and Flow has started to really streamline the way I work.

This is part of a series on Microsoft Flow

Microsoft Todo, the app that's just damn simple

I don't really want to get into why I choose Todo instead of Wunderlist or Todolist or Google Keep.  I think I just want a really simple Todo app that I'll complete every day.  

In a nutshell, this is how I use Todo:

Every time you open Todo, you see Today.  This starts every day blank, like a clean slate

Hit suggestions, and it shows you tasks from three categories that you can add to "today"

Sync everywhere

Knock out those tasks in today

Todo doesn't have an API

But it sync with Outlook Tasks - which gives me the idea for this Flow.  (Actually, I'm pretty sure Todo just uses Outlook Tasks as the source).

View fullsize

1. Set up a Web Request
2. Calculate today - NOTE don't include quotes " and "
3. Create Outlook Task with title, description and today as due date

Test with Postman

Remember to set the content-type to application/json 

And we see this in Todo, almost instantly.

What's next?

I now have a web service that takes a simple JSON and I can add tasks anywhere into my Todo.  I'm playing with the idea that different Flows can put tasks into this API, which will populate tasks that I can add into "today".

• Bills from Emails
• Notable Twitter that spun off a crazy blog idea
• SharePoint project tasks
• OneNote tasks
• Planner (? this one I may leave in the Planner app)

I'm also thinking to create different lists for different sources, and watch task changes to write back to source system.  That will make this a really good end to end solution for me.

Building Binary output service with Cognitive Services and Microsoft Flow

We covered how to do binary webservices with Microsoft Flow.  A question then lingers in my mind.  If you can push binary data into a Flow, and within the Flow you can pass it around...  Can you output a binary file from Flow?

This question bothered me so much in my sleep, I decided to test it and write this blog.  And thus, we have probably the simplest example of the series.

1. So we will first build a service endpoint that can return binary data.
2. Then we will send it through cognitive services and tag some data as we go.

This is a post in a series on Microsoft Flow.

1. JSON cheatsheet for Microsoft Flow
2. Nested-Flow / Reusable-Function cheatsheet for Microsoft Flow
3. Building non-JSON webservices with Flow 
4. One Connection to Proxy Them All - Microsoft Flow with Azure Functions Proxies
5. Building Binary output service with Cognitive Services and Microsoft Flow

Build a Flow to output non-text Output

The method needs to be set to GET.  Take a image that's authenticated in SharePoint, and set that to be the response output.

Test this with Postman

View fullsize

A few things to note:

1. The request is a GET request.
2. It replies with image/png (content type was automatically worked out)
3. ... and that's it, there's not a lot to say

Add Cognitive Services - Computer Vision

You'll need to create a Cognitive Services in your Azure Subscription.  The free tier offers 5000 images per month, at 20/minute.

We are taking the output of the tag action and adding that to the tags header in the service response.

View fullsize

And here we have the same image, but now with tags in the output.

Smart dogs.

Why do we need this?

1. This means - we can post image in, and we can get image out
2. May be you need to proxy a resource within SharePoint that is authenticated - but you want to use it directly as a file.  If you use a SharePoint Sharing link it'd take you to a page.
3. With this direct link to the file, you can use this as an anchor within HTML, or use this to upload a file to an external system (via URL).
4. May be this isn't a file, but a generated ZIP file that you want to copy somewhere else.  Or it is a docx file.
5. Or perhaps you want to send a picture to a Flow, then resize it or run it through cognitive services before getting back the result.
6. May be you are just mad and want to auto-tag every image in your SharePoint?
   That actually sounds amazing.

Because Microsoft Flow lets us push binary through actions, I think there's a bunch of interesting scenarios for this.

Also, I think assistant branch manager and branch manager are awesome.

One Connection to Proxy Them All - Microsoft Flow with Azure Functions Proxies

Office 365 licensed PowerApps and Microsoft Flow lets you have 1 custom connection.  I think that's cute.  I also think that limitation is absurd, considering how many times we need to break out of existing capabilities every time we want to do something interesting.  So, custom connections shouldn't be a limit, if the BAP team wants us to innovate the crap out of this platform, this is counter productive.  I hope the team reconsider.

Regardless, it is pointless, because we bypass that right away: Azure Functions has Proxies.

This is a post in a series on Microsoft Flow.

1. JSON cheatsheet for Microsoft Flow
2. Nested-Flow / Reusable-Function cheatsheet for Microsoft Flow
3. Building non-JSON webservices with Flow 
4. One Connection to Proxy Them All - Microsoft Flow with Azure Functions Proxies
5. Building Binary output service with Microsoft Flow

In this post, we talk about:

• Azure Functions proxy for Microsoft Flow connections
• Azure Functions proxy for itself
• Azure Functions proxy for bypassing CORS
• Azure Functions proxy for 3rd party APIs
• One Simple API for me

So in one Connection to proxy them all, in one Request to Bind them.  And in Serverless something about we pay nothing - hey I'm a blogger not a poet.

Azure Functions Proxy for Microsoft Flow

In an earlier post - I talk about we can turn Microsoft Flow into webservices with the Request trigger.  This is awesome, but it leaves you with an endpoint that looks like sin.

https://prod-18.australiasoutheast.logic.azure.com:443/workflows/76232c77d84d424b8e56ab2f88b672c4
/triggers/manual/paths/invoke?api-version=2016-06-01
&sp=%2Ftriggers%2Fmanual%2Frun
&sv=1.0
&sig=FAKE_KEY_NXpUV_FEyhl1BKgKftQ0-rcOPcE

Lets clean this up with Azure Functions Proxies.  First, enable it.

View fullsize

Create a Proxy: set the name to "uploadBinary" and template to "upload".  Chose your methods (POST only for this one), and paste in the backend URL.

Hit create - our proxy URL is generated.  https://funky-demo.../upload

We had a Swagger OpenAPI file for this binary web request - lets tidy it up 

{
  "swagger": "2.0",
  "info": {
    "description": "",
    "version": "1.0.0",
    "title": "UploadBinary-proxy"
  },
  "host": "funky-demo.azurewebsites.net",
  "basePath": "/",
  "schemes": [
    "https"
  ],
  "paths": {
    "upload": {
      "post": {
        "summary": "uploads an image",
        "description": "",
        "operationId": "uploadFile",
        "consumes": [
          "multipart/form-data"
        ],
        "parameters": [
          {
            "name": "file",
            "in": "formData",
            "description": "file to upload",
            "required": true,
            "type": "file"
          }
        ],
        "responses": {
          "200": {
            "description": "successful operation"
          }
        }
      }
    }
  }
}

We can delete all the additional parameters sp, sv, sig, api-version.  We pretty much say we just care about the formData.

Go back to our PowerApps custom connectors - delete the old one and add this new OpenAPI file.
This is what PowerApps now see: the URL is only /upload
and there are no hidden query parameters - just waiting for the body.

View fullsize

Delete the old connection and recreate a new connection.
The method now on the new connection only takes 1 parameter "file".

The result in PowerApps and SharePoint

This test picture is called "Azure Function is On Fire!"

View fullsize

So in using AzureFunctions Proxy for a Flow WebRequest - we've simplified the query string in the URL and kept the functionality.

Azure Functions proxy for Azure Functions

Some of our connections were to Azure Functions, so let's tidy that up too.

View fullsize

This is a pretty simple example.  /listEmail now is hooked up, and activation code is hidden.

Azure Functions proxy for bypassing CORS

Everything you proxy in Azure Functions can be made available to CORS - just need to turn on Azure Functions CORS with your domain

View fullsize

CORS' design goals are to prevent browser based injection/consumption of a service.  But services are made to be consumed.  So CORS blocks a browser, but never blocks a server-based call.

Azure Function runs server, and proxy works because it doesn't see CORS at all.  Switching on Azure Functions CORS is so that your domain is accepted to call this Azure Function.

Anyway, if you see CORS and got annoyed - just use Azure Functions Proxy to keep going.

Azure Functions proxy for 3rd party APIs

Let's add an external Open Weather API

View fullsize

Azure Functions Proxies has additional means of passing through parameters to the underlying proxied URL.  See Documentation: https://docs.microsoft.com/en-us/azure/azure-functions/functions-proxies

So the call to the SAMPLE open weather API is:

http://samples.openweathermap.org/data/2.5/weather
?lat={request.querystring.lat}
&lon={request.querystring.lon}
&appid=b1b15e88fa797225412429c1c50c122a1

So we are exposing two query string parameters in our end point to be passed down.

One simple API Connection for me

{
    "swagger": "2.0",
    "info": {
        "description": "",
        "version": "1.0.0",
        "title": "One-Funky-API"
    },
    "host": "funky-demo.azurewebsites.net",
    "basePath": "/",
    "schemes": [
        "https"
    ],
    "paths": {
        "upload": {
            ...
        },
        "listEmail": {
            ...
        },
        "weather": {
            ...
        }
    }
}

In all the examples, we clean up the API we are calling to hide away apiKeys and such and make our end point really easy to consume from Flow, PowerApps or JavaScript CORS.

And because both Microsoft Flow and Azure Functions are Serverless, we still don't care about servers ;-)

Building binary (non-JSON) webservices with Flow

This is a short(er) blog post, as I'm packing to get ready to sleep and then wake up early and fly to Brisbane for this weekend's Office 365 Saturday.

We will build upon what we discussed with building Flow as a Service today, and build a non-JSON webservice - we will build it to handle multi-part/form-data which is perfect for handling sending binary files to Flow, from the likes of PowerApps.

This is the third post of the series on Microsoft Flow.

1. JSON cheatsheet for Microsoft Flow
2. Nested-Flow / Reusable-Function cheatsheet for Microsoft Flow
3. Building non-JSON webservices with Flow
4. One Connection to Proxy Them All - Microsoft Flow with Azure Functions Proxies
5. Building Binary output service with Microsoft Flow

In this post, we will cover:

1. Define a Flow service to handle multi-part/form-data
2. Creating the OpenAPI (Swagger) definition
3. Call this from PowerApps

History

This is a problem that was set out by Paul Culmsee - as I understood the requirements - he wanted to use PowerApp's Add Media control, which does not provide the image as a dataUri and thus our previous solution of sending dataUri (as string) doesn't work.

The Add Media control sends a blob.  A binary file.
Paul actually did lots of research and dug up lots of reference links: 

• https://powerusers.microsoft.com/t5/PowerApps-Forum/native-camera-amp-flow/m-p/43635
• https://community.powerapps.com/t5/PowerApps-Forum/Retrieving-Data-URI-from-an-image-added-via-the-quot-Add-picture/m-p/40247

The second solution is Solved.  But it requires a C# service.

Define a Flow service to handle multi-part/form-data

First - I recall seeing a @triggerMultipartBody() Logic App Workflow Definition.  We build upon the previous post of creating a Flow Service.

First difference - we don't use the body JSON schema.  We assume the body will be multi-part/form-data, so it won't follow a JSON schema.  This is not a JSON service.

Save this Flow, we need the Request URL.

Set up Postman to poke this service

View fullsize

Use the Request URL, set the method to POST, the Body type is form-data.  And for the key value - set it to "File", and select a real picture.

Hit send.

View fullsize

You should see this.  This is quite good.  It means:

1. The form-data works.
2. We have the image/png, it looks like it's in base64 string.

View fullsize

We don't really care about all the bits of part 0 of the multi-part form-data.  We just want the $content.
Select that and wrap it in base64StringToBinary()

Send the picture again from Postman.

And the picture is in SharePoint!

Creating the OpenAPI (Swagger) definition

This part is hard and annoying.  But I started with the OpenAPI example from https://community.powerapps.com/t5/PowerApps-Forum/Retrieving-Data-URI-from-an-image-added-via-the-quot-Add-picture/m-p/40247 

At the end, I ended up with this:

{
  "swagger": "2.0",
  "info": {
    "description": "",
    "version": "1.0.0",
    "title": "UploadBinary3"
  },
  "host": "prod-18.australiasoutheast.logic.azure.com",
  "basePath": "/workflows",
  "schemes": [
    "https"
  ],
  "paths": {
    "/76232c77d84d424b8e56ab2f88b672c4/triggers/manual/paths/invoke": {
      "post": {
        "summary": "uploads an image",
        "description": "",
        "operationId": "uploadFile",
        "consumes": [
          "multipart/form-data"
        ],
        "parameters": [
          {
            "name": "api-version",
            "in": "query",
            "default": "2016-06-01",
            "required": false,
            "type": "string"
          },
          {
            "name": "sp",
            "in": "query",
            "default": "/triggers/manual/run",
            "required": false,
            "type": "string"
          },
          {
            "name": "sv",
            "in": "query",
            "default": "1.0",
            "required": false,
            "type": "string"
          },
          {
            "name": "sig",
            "in": "query",
            "default": "Mb25lHIBI6Sq4xXkNXpUV_FEyhl1BKgKftQ0-rcOPcE",
            "required": false,
            "type": "string"
          },
          {
            "name": "file",
            "in": "formData",
            "description": "file to upload",
            "required": true,
            "type": "file"
          }
        ],
        "responses": {
          "200": {
            "description": "successful operation"
          }
        }
      }
    }
  }
}

Arguments such as "api-version", "sp", "sv", "sig" are part of the Request URL.  I specify the default value, and mark them as not required.

The one required parameter is "file", the type is "file" and is in "formData".

Regression Bug / Note on Custom Connections

Changes to the Custom Connections in PowerApps and Flow has prompted me to mention a workaround - if we specify "not required arguments with default" for arguments "api-version", "sp", "sv" and "sig" - they don't seem to be populated properly when calling from PowerApps. This seems to be a regression bug.

Instead, go into your Custom Connections screen in PowerApps/Flow, and set the optional values to "hidden" (keep the defaults). This will clean up the PowerApps connections and keep the webservice working.

Call this from PowerApps

Use the Add media control.  Configure the button to call our OpenAPI-configured connection.  Which shows the correct arguments: file is required, and the others are optional.  You can ignore them.

Final test

Final Thoughts

So this is an end to end example of sending form-data from PowerApps to Flow to SharePoint.  There are some exercises I think would make this a better application.

Flow should take a second parameter, which could be triggerMultipartBody(1) - it should be a string, and should be used as the filename.  Swagger should be adjusted as well.  Alternatively, you can do this in the URL as part of the query string.

We probably can make the multi-part form-data handle more than one file at a time.  And make the first argument required, but subsequent arguments optional.

In the DataUri example we had issues with the Windows application since it didn't use DataUri - it used Blobs.  With this example, we are able to make Windows app work as well.

If I sit back and really consider what this example means - we've built a webservice that can handle binary files, without writing a single line of 'developer' code.  It makes me wonder what sort of product is Flow, and just how powerful the entire platform is.

Everyone needs to understand Flow, it has serious potential.