Download Developing AI-Enabled Database Solutions.DP-800.PassLeader.2026-09-09.54q.vcex

Implementing a Retrieval-Augmented Generation (RAG) pattern with Microsoft SQL Server 2025 allows you to ground generative AI responses in real-time company data without the need for model fine-tuning. This version of SQL Server functions as a native vector database, enabling seamless integration between relational data and AI-powered insights. Incorrect: [Not B] Training a custom language model is not a use case for the Retrieval Augmented Generation (RAG) pattern. In fact, the primary purpose of RAG is to provide a language model with up-to-date information without the need for training, retraining, or fine-tuning. Reference: https://devblogs.microsoft.com/azure-sql/sql-server-2025-embraces-vectors-setting-the-foundation-for-empowering-your-data-with-ai 

Azure OpenAI Service allows users to query structured SQL data using natural language via LLMs. 

Vector embeddings enable similarity search, allowing semantic queries beyond exact matches. 

To disable GitHub Copilot repository-level instructions for yourself without affecting others, you can modify your User Settings in Visual Studio Code. This allows you to override or ignore specific repository-wide configurations like copilot-instructions.md at a personal level. How to Disable Repository-Level Instructions 1. Open User Settings: Press Ctrl+, (Windows/Linux) or Cmd+, (macOS) to open the VS Code Settings editor. 2. Search for Copilot Chat: In the search bar, type github.copilot.chat.customInstructions. 3. Configure Custom Instructions: Find the setting for Github > Copilot > Chat: Custom Instructions. Ensure the User tab is selected at the top to apply changes only to your account and not the shared workspace. 4. Toggle via Chat UI: Open the Chat view (Ctrl+Alt+I or Cmd+Shift+L). Click the Configure Chat (gear icon) at the bottom of the chat panel. Select the Instructions tab and uncheck or remove any active repository-level files to disable their influence on your session. Reference: https://code.visualstudio.com/docs/copilot/customization/custom-instructions

In a Microsoft Fabric SQL database, you can use the REGEXP_REPLACE function (introduced as part of the SQL Server 2025/Always-up-to-date engine updates) to normalize email subaddresses. Solution using REGEXP_REPLACE The command you proposed correctly extracts the email string from the JSON document and applies a regular expression to strip the plus-sign subaddress. SELECT REGEXP_REPLACE(JSON_VALUE(Payload, '$.customer_email'), '\+.*@', '@') AS NormalizedEmail FROM YourTable; Breakdown of the Command JSON_VALUE: Efficiently extracts the customer_email as a scalar string from your Payload column. REGEXP_REPLACE: Searches for the pattern \+.*@ (a literal plus sign followed by any characters until an @) and replaces that entire matched segment with just @. Native Support: Unlike older versions of SQL Server that required complex CHARINDEX and SUBSTRING workarounds, Fabric SQL databases now include this modern regex functionality. Reference: https://devblogs.microsoft.com/azure-sql/exciting-new-t-sql-features-regex-support-fuzzy-string-matching-and-bigint-support-in-dateadd-preview/ 

To initialize the Data API builder (DAB) configuration to read the SQL connection string from your Container Apps secret, use the following dab init command: dab init --database-type mssql --connection-string "@env('MSSQL_CONNECTION_STRING')" Why this command works --database-type mssql: Specifies that you are connecting to an Azure SQL or SQL Server database. @env('MSSQL_CONNECTION_STRING'): This is the built-in DAB function that tells the runtime to substitute the value of the specified environment variable at load time. Since your Container Apps secret is mapped to MSSQL_CONNECTION_STRING, DAB will resolve it automatically when the container starts. --connection-string: This flag sets the data source connection. By using the @env() syntax here, you ensure the secret remains out of the static configuration file. Reference: https://learn.microsoft.com/en-us/azure/data-api-builder/command-line/dab-init

The DISABLE TRIGGER Transact-SQL statement is the correct and appropriate solution for this scenario. Solution Breakdown To prevent a specific database-level DDL trigger from firing without removing the object, you can use the following syntax: DISABLE TRIGGER [TriggerName] ON DATABASE; Key Considerations Object Retention: A disabled trigger remains in the database as an object and is visible in catalog views like sys.triggers, but it will not execute when its programmed events occur. Reactivation: You can re-enable the trigger after your deployment is complete using the ENABLE TRIGGER statement. Permissions: To execute this command on a database-scoped DDL trigger in Azure SQL, you must have at least ALTER ANY DATABASE DDL TRIGGER permission. Reference: https://learn.microsoft.com/en-us/sql/t-sql/statements/disable-trigger-transact-sql

A CHECK constraint is one way to do it. If you use a CHECK constraint (e.g., CHECK (ColumnName IS NOT NULL)), the database will indeed reject new NULL entries. However, the column's metadata will still technically allow NULLs, which can sometimes affect how external tools or APIs interact with your schema. Reference: https://www.postgresql.org/docs/7.0/sql-createtable.htm

To use a Managed Identity with sys.sp_invoke_external_rest_endpoint in SQL Server, you need to configure your database-scoped credential with the following specific values: Required Settings WITH IDENTITY: Must be set to Managed Identity. SECRET: Must be set to SECRET = N'{"resourceid":"<value>"}'. Example Statement CREATE DATABASE SCOPED CREDENTIAL [MyManagedIdentityCredential] WITH IDENTITY = 'Managed Identity', SECRET = N'{"resourceid": "https://azure.com"}'; -- Or your specific resource URI Quick Breakdown Identity: Managed Identity tells SQL Server to use the instance's assigned Azure identity rather than a static key or token. Secret: The resourceid inside the JSON string specifies the Audience (resource) for which the Managed Identity is requesting an access token (e.g., Azure Storage, Key Vault, or a custom API). Reference: https://learn.microsoft.com/en-us/azure/azure-sql/managed-instance/authentication-azure-ad-user-assigned-managed-identity-create-managed-instance 

RAG enhances LLM responses by retrieving relevant SQL data before generating answers. 

To use GitHub Copilot Chat effectively in this environment without exposing sensitive data, you should focus your prompts entirely on the schema and logic rather than the data itself. Since Copilot Chat can "see" your open files, the best approach is to provide the table structures as DDL (Data Definition Language) statements or a simplified description. Steps to generate the stored procedure: Define the Schema: Open a new SQL file in your Codespace. Paste the CREATE TABLE scripts (without any data) for your two tables. Prompt Copilot: Use the Chat view to request the procedure. Example Prompt: "Based on the table definitions in my open file, write a T-SQL stored procedure that joins TableA and TableB on [Join Column]. Include logic to filter by [Parameter] and ensure no PII columns are included in the SELECT statement." Review for PII/Secrets: Before executing, manually verify that the generated code doesn't include hardcoded secrets or call PII columns you intended to omit. Security Check: Because you are in a Codespace, ensure your .env files or connection strings are in your .gitignore so they aren't indexed by Copilot. Reference: https://github.com/orgs/community/discussions/141924