Generate Sequence Diagrams for Azure B2C User Journeys with ChatGPT

Dariusz Parys
medialesson
Published in
3 min readJul 31, 2024

Working with Azure B2C is an XML battle, especially when it comes to customizing user journeys using custom policies. A starter pack you can get from GitHub helps you set up your own policy.

Azure-Samples/active-directory-b2c-custom-policy-starterpack at dariuszparys.com (github.com)

The readme is comprehensive and has all the necessary information, but sometimes, it helps to have a flow sequence diagram. Sequence diagrams can be created using Mermaid, given a syntax that can be included in your Markdown documentation. The renderer will then visualize the sequences for you.

Generative AI can help here

There is an excellent way to do this with Generative AI. In my case, I use ChatGPT. The steps are as follows.

First, upload the XML policy files you want to analyze. In my example, I’m interested in the sign-in and sign-up user journey, so I’m only uploading TrustFrameworkBase.xml, TrustFrameworkExtensions.xml, and SignUpOrSignIn.xml.

Next, prompt ChatGPT to generate the sequence diagram based on the uploaded files. In the case I provide here, I used the following prompt:

Create a sequence diagram for the <User Journey Name> user journey in Azure AD B2C. The diagram should integrate the flow of claims directly into the main sequence and use notes and background colors to distinguish between different flows. 

### Sequence Diagram Details:
1. The initial request where the user opens the page and requests the policy.
2. The user decision point where they choose between different options (e.g., Sign Up or Sign In).
3. The detailed flow for each option, including:
- Collecting claims
- Validating and transforming claims
- Interactions with Azure AD for validation and user data retrieval
- Displaying success messages to the user
4. The token refresh flow if applicable.
5. Use notes to highlight claims collection and validation.
6. Use different background colors to distinguish between different options (e.g., Sign Up and Sign In).

Replace <User Journey Name> with the user journey I uploaded, in the example Sign In or Sign Up. The chatbot will start to analyze the files, and as a result, I get back this Mermaid sequence diagram.

sequenceDiagram
participant User
participant Browser
participant AzureAD_B2C as Azure AD B2C
participant IDP as Identity Provider
participant App as Relying Party Application

Note over User, Browser: Step 1: Access Sign Up or Sign In page
User->>Browser: Access Sign Up/Sign In Page
Browser->>AzureAD_B2C: Request Sign Up/Sign In Page
AzureAD_B2C->>Browser: Return Sign Up/Sign In Page

Note over User, Browser: Step 2: Enter credentials or select IDP
User->>Browser: Enter credentials/select IDP
Browser->>AzureAD_B2C: Submit credentials/IDP selection

alt Local Account
AzureAD_B2C->>AzureAD_B2C: Validate credentials
AzureAD_B2C->>Browser: Prompt for additional claims (if needed)
Browser->>User: Display additional claims form
User->>Browser: Submit additional claims
Browser->>AzureAD_B2C: Submit additional claims
else Social IDP
AzureAD_B2C->>IDP: Redirect to IDP
IDP->>User: Display IDP login page
User->>IDP: Enter credentials
IDP->>AzureAD_B2C: Return user claims
end

Note over AzureAD_B2C: Step 4: Process user claims
AzureAD_B2C->>AzureAD_B2C: Validate and transform claims
AzureAD_B2C->>App: Send claims to relying party application

Note over App, User: Step 5: User is authenticated
App->>User: Welcome Page or Application Landing Page

You can look at the result in GitHub, which supports rendering Mermaid diagrams.

Rendered sequence diagram for Sign In or Sign Up user journey

https://gist.github.com/dariuszparys/904d64978f9448f23bfb7698fee0cf6a

As with all generative models, the output may vary depending on how you prompt it and what model you use. It is more about the basic idea that we can transform one format to another, which helps us create documentation we don’t have out of the box. Give it a try.

Originally published at https://www.dariuszparys.com on July 31, 2024.

--

--

Dariusz Parys
medialesson

Passionate about software and technology development, software engineering, photography, and heavy metal. Supporter of Bayern Munich and VfB Stuttgart.