We are very excited to announce the Public Preview for the workflow assistant in Azure Logic Apps!!
The workflow assistant is a chat interface to answer any questions about Azure Logic Apps, from directly within designer. You can also use it to describe existing workflow or prompt for specific answers in the context of your workflow. The chat interface provides access to Azure Logic Apps documentation and best practices without requiring you to navigate documentation or search online forums.
About workflow assistant
The workflow assistant uses Azure Open AI and ChatGPT to query diverse knowledge sources related to Azure Logic Apps, delivering curated information to your. This system uses Azure Logic Apps documentation from various reputable sources, such as Microsoft Learn, connector schemas, tech community blogs, in addition to the internet data used to train GPT 3.5-Turbo. These documents are processed into a vectorized format and are then accessible through a backend system built on top of Azure App Service. The queries are triggered based on user interaction through the workflow designer in Azure Logic Apps.
When you provide an input prompt, the Azure Logic Apps backend performs pre-processing and forwards the prompt to an Azure Open AI large language model. This model generates responses based on the given context (in the form of workflow JSON) and your prompt.
With the workflow assistant at your side during workflow development, our objective is to ensure that all the information that you require is readily available, presented in a curated manner, and without the need for context-switching.
The workflow assistant is suitable for many use cases including,
- Describe workflow
- This is useful when you are working on workflows built by other developers or collaborating with other developers on shared workflows.
- Azure Logic Apps has 1000+ connectors and so much choice can be overwhelming. The workflow assistant can provide recommendations on connectors or actions, provide best practices to use a connector, provide comparisons between connectors, and much more.
- Recommended patterns
- Like code, well written workflows and integration applications can also follow standard patterns. The workflow assistant can provide responses that you can leverage to apply best practices, such as for error handling, testing, and other optimizations.
- The workflow assistant can recommend step by step guidance on how to build workflows based on the scenarios – including connectors to use, how to configure them and how to process the data.
- Contextual response
- The workflow assistant builds its responses based on your workflow opened in the designer. This means that you can ask very specific questions on ‘how’ to do things in the context of an action. For example, learning about configurations for specific actions and then getting more help on how to configure them. Recommendations on the data input or output for an action and how to test them and so on.
These are just some examples; we look forward to hearing how you use the workflow assistant to improve developer productivity.
How to use workflow assistant in Azure Logic Apps Standard
The workflow assistant is only available in Azure Logic Apps Standard for now. To use the workflow assistant, you can open an existing workflow or create a new workflow.
- The workflow opens in the workflow designer. From the workflow menu , select Assistant. This should open workflow assistant pane to the left of your workflow.
- In the text box, you can ask any questions related to Azure Logic Apps capabilities or specific questions about the workflow.
- You can close this workflow assistant at any time using the X button. Chat history is not preserved when you close the assistant.
Depending on factors such as the number of workflow actions or workflow complexity, you might experience different levels of performance with the workflow assistant. The workflow assistant has been trained on workflows of different complexities but still has a limited scope and might not be able to handle very large workflows. These limitations are primarily linked to the token constraints in the queries sent to the OpenAI systems. We are committed to continuous improvement and will work to enhance these limits in iterative updates.
The workflow assistant might generate responses that might be valid but might not be actually semantically correct or capture the intent of the developer prompt. As the language model is trained on more data, over time the responses will only become better. Make sure to carefully review the recommendations before applying them in your workflows.
We value your feedback and are eager to hear from you, especially if you encounter unexpected responses or have any concerns with the workflow assistant. Providing feedback is easy; simply select the thumbs-down icon in the workflow assistant’s response and share your insights with us. We greatly appreciate detailed feedback, so please include information about your workflow, the specific question, and the response you received from the workflow assistant. If you have any other inquiries or need assistance, feel free to reach out to us directly or through this form.
> Is the workflow assistant available for all developers?
The workflow assistant is only being launched on Azure Logic Apps Standard at Ignite. It’s also available only through Azure portal for now, and is not available in Visual Studio Code.
> Is the workflow assistant available in all regions and languages?
The workflow assistant is being launched in all regions where Azure Logic Apps Standard is available at Ignite. However, in all regions, the workflow assistant supports only English for input prompts and responses. We do plan to support more languages in the future. The workflow assistant deployment will complete in all regions by 11/17, so if you can’t see it yet, you should get it soon.
> Does Azure Logic Apps own the workflow suggested by the workflow assistant?
The Azure Logic Apps workflow assistant does not own the suggestions that workflow assistant provides you or the workflow built with those suggestions. You own and are responsible for the workflow that you author with the workflow assistant’s help.
> What data does the workflow assistant collect?
The workflow assistant relies on the workflow definition ( workflow JSON) to provide contextual responses. The workflow JSON is sanitized to ensure no customer data or secrets are passed as context. The workflow definition is only used to scope the responses and is not stored anywhere. We do collect some telemetry on the UI interactions for troubleshooting purposes and that does not include any customer or personal data.
> How is personal or customer data in workflow assistant used and shared?
The workflow assistant does not collect or share any personal data. We also do not collect or store customer data provided in the workflow assistant questions or responses
> Where can I learn about responsible and ethical AI practices at Microsoft?
The workflow assistant follows responsible and ethical AI practices in accordance with the [Microsoft responsible AI principles and approach](https://www.microsoft.com/ai/principles-and-approach).
> Where can I learn about privacy and data protection?