Skip to main content

LLM - Document Studio Flows with Azure OpenAI

ACCESS THE CUSTOM STEP FOLDER

LLM - Document Studio Flows with Azure OpenAI

Description​

The "LLM - Document Flows with Azure OpenAI" custom step uses Azure OpenAI’s large language models, such as GPT-4o, to automatically create documentation for a SAS Studio flow.

Video and Blog Post​

How It Works​

  1. Reads and prepares the .flw file content.
  2. Loads a system message from the messages file.
  3. Retrieves Azure OpenAI credentials and endpoint details from the .env file.
  4. Sends the .flw file content and system message to the Azure OpenAI API.
  5. Saves the API response (generated documentation) to a specified .txt output file.

Steps:

  1. Add the custom step to a SAS Studio flow.
  2. Choose the inputs.
  3. Specify the output, the .txt file for the documentation.
  4. Run the flow.

The LLM uses 2500 tokens for documentation by default. Adjust the parameter "max_tokens": 2500 in the Python program to change this limit.

Prerequisites​

1. SAS Viya​

a. A working SAS Viya environment with Python installed and configured. Tested on SAS Viya LTS version 2024.09 and Stable version 2025.01. b. Required Python libraries:

  • python-dotenv: Loads environment variables from .env file.
  • requests: Sends API requests to Azure OpenAI.

c. Display Hidden Files: Enable hidden file visibility in SAS Studio via Options > Preferences > General > check Display hidden files.

2. Azure OpenAI Resource​

  • Azure Subscription: Active subscription required.
  • Deployed GPT-4o Model: Set up an Azure OpenAI resource and deploy a GPT-4o model. Details needed:
    • Endpoint URL
    • API Key
    • Deployment Name

3. Environment Configuration (.env File)​

Create a .env file with:

AZURE_OAI_ENDPOINT='https://my_endpoint.openai.azure.com/' # change my_prefix
AZURE_OAI_KEY='my_api_key' # change my_key
AZURE_OAI_DEPLOYMENT='gpt-4o' # change to match your deployment name

Example available at extras/.env.sample.txt.

4. System Message File​

A system_message.txt file shapes the documentation by guiding the LLM. It defines structure, detail level, and focus areas. Example available at extras/system_message.txt.

5. Output Location​

Specify a file path for the generated documentation. Ensure write permissions.

User Interface​

Options Tab​

Options

About Tab​

About

Usage​

  1. Add the custom step to a SAS Studio flow.

  2. On the Options tab:

    Usage example

Inputs and Outputs​

Required Inputs​

  1. Flow to be documented (.flw file): Full path to the SAS Studio flow file to be documented. Example: document Flow.flw stored in sasserver:/azuredm/code/Flow.flw.
  2. Folder for the .env file: Directory containing the .env file for the Azure OpenAI connection.
  3. File for LLM system message (system_message.txt): Path to the system_message.txt file containing the system message.

Required Outputs​

  1. Output file (.txt): Path to save the SAS Studio flow documentation. The file contains Markdown-formatted text.

Run the SAS Studio flow.

Expected Results​

  • Documentation saved in a .txt file.

    Expected Result A

    Expected Result B

  • The file contains Markdown. View it with a Markdown viewer, e.g., Visual Studio Code.

    Expected Result C

    Expected Result D

Change Log​

  • Version 1.2 (23JUL2025)

    • Small changes to About tab structure to reflect best practices.

  • Version 1.1 (29JUN2025):

    • Implemented reviewer suggestions.

  • Version 1.0 (09APR2025):

    • Initial version.