fix: round 2 review - SQL safety, precomputed, em dashes, gitignore

- SQL: use N'' prefix for Unicode string in sys.tables lookup
- SQL: add vector dimensions comment for text-embedding-3-small
- SQL: add timeout/retry rationale comments on OpenAI client
- sample.env: move inline comment to separate line
- README: pre-computed -> precomputed (6 instances)
- README: fix em dash spacing (7 instances)
- .gitignore: expand parent gitignore with .env.*, dist/, build/, logs, OS files

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Author: diberry

samples/features/vector-search/.gitignore

@@ -1,2 +1,20 @@
+# Environment files
 .env
+.env.*
+!.env.sample
+!.env.example
+
+# Dependencies
 node_modules/
+
+# Build artifacts
+dist/
+build/
+
+# Logs
+*.log
+npm-debug.log*
+
+# OS files
+.DS_Store
+Thumbs.db

samples/features/vector-search/vector-search-query-typescript/README.md

@@ -4,16 +4,16 @@ This sample demonstrates how to perform **native vector search** in Azure SQL Da
 
 It uses:
 
-- **[tedious](https://www.npmjs.com/package/tedious)** — Microsoft's Node.js driver for SQL Server, with Azure AD authentication
-- **[openai](https://www.npmjs.com/package/openai)** — Azure OpenAI SDK for generating embeddings (via the `AzureOpenAI` class)
-- **[@azure/identity](https://www.npmjs.com/package/@azure/identity)** — `DefaultAzureCredential` for passwordless authentication to both Azure SQL and Azure OpenAI
+- **[tedious](https://www.npmjs.com/package/tedious)**—Microsoft's Node.js driver for SQL Server, with Azure AD authentication
+- **[openai](https://www.npmjs.com/package/openai)**—Azure OpenAI SDK for generating embeddings (via the `AzureOpenAI` class)
+- **[@azure/identity](https://www.npmjs.com/package/@azure/identity)**—`DefaultAzureCredential` for passwordless authentication to both Azure SQL and Azure OpenAI
 
 ## What the sample does
 
-1. Loads 50 hotels with pre-computed embeddings from `data/HotelsData_Vector.json`
+1. Loads 50 hotels with precomputed embeddings from `data/HotelsData_Vector.json`
 2. Connects to Azure SQL Database using `DefaultAzureCredential` (no passwords or API keys)
 3. Creates a table with `id`, `name`, `description`, `category`, `rating`, and a `VECTOR(1536)` column
-4. Inserts all 50 hotels with their pre-computed vector embeddings
+4. Inserts all 50 hotels with their precomputed vector embeddings
 5. Generates a fresh query embedding using Azure OpenAI `text-embedding-3-small`
 6. Performs a vector similarity search using either **exact kNN** (`VECTOR_DISTANCE`) or **approximate ANN** (`VECTOR_SEARCH` with DiskANN index), based on the `VECTOR_SEARCH_ALGORITHM` environment variable
 7. Displays the top matching results with category, rating, and similarity scores
@@ -109,7 +109,7 @@ This runs the TypeScript code directly using `tsx` with Node.js 20+ native env-f
 ## Expected output
 
 ```
-=== Azure SQL Vector Search — TypeScript Quickstart ===
+=== Azure SQL Vector Search—TypeScript Quickstart ===
 
 Server:     <your-server>.database.windows.net
 Database:   <your-database>
@@ -126,12 +126,12 @@ Connected.
 Creating hotels table (if not exists)...
 Table ready.
 
-Inserting hotel data with pre-computed embeddings...
+Inserting hotel data with precomputed embeddings...
 Inserted 50 hotels.
 
 Searching for: "luxury beachfront hotel with ocean views and spa"
 
---- Search Results — Exact (kNN) via VECTOR_DISTANCE (Top 3 by Cosine Distance) ---
+--- Search Results—Exact (kNN) via VECTOR_DISTANCE (Top 3 by Cosine Distance) ---
 
   Hotel:       Ocean Water Resort & Spa
   Category:    Luxury
@@ -186,9 +186,9 @@ CREATE TABLE dbo.hotels_typescript (
 );
 ```
 
-### Loading pre-computed vectors
+### Loading precomputed vectors
 
-Hotel data with pre-computed embeddings is loaded from `data/HotelsData_Vector.json`. This avoids calling Azure OpenAI for each hotel during the main run, making the demo faster and simpler:
+Hotel data with precomputed embeddings is loaded from `data/HotelsData_Vector.json`. This avoids calling Azure OpenAI for each hotel during the main run, making the demo faster and simpler:
 
 ```typescript
 const hotels = JSON.parse(readFileSync(dataPath, "utf-8"));
@@ -295,7 +295,7 @@ az sql server firewall-rule create \
 
 ### Vector dimension errors
 
-**"Invalid or missing vector dimensions"** — The pre-computed embeddings in `HotelsData_Vector.json` must use 1536 dimensions (matching `text-embedding-3-small`). Re-run `npm run embed` if you changed the embedding model.
+**"Invalid or missing vector dimensions"**—The precomputed embeddings in `HotelsData_Vector.json` must use 1536 dimensions (matching `text-embedding-3-small`). Re-run `npm run embed` if you changed the embedding model.
 
 ## Explore your database
 

samples/features/vector-search/vector-search-query-typescript/sample.env

@@ -4,7 +4,7 @@ AZURE_SQL_DATABASE=<your-database>
 AZURE_SQL_TABLE_NAME=hotels_typescript
 AZURE_OPENAI_ENDPOINT=https://<your-resource>.openai.azure.com
 AZURE_OPENAI_EMBEDDING_DEPLOYMENT=text-embedding-3-small
-# Vector search algorithm: "exact" (default, uses VECTOR_DISTANCE) or "diskann" (uses VECTOR_SEARCH with DiskANN index)
-VECTOR_SEARCH_ALGORITHM=exact # "exact" (default, uses VECTOR_DISTANCE) or "diskann" (uses VECTOR_SEARCH with DiskANN index)
+# Vector search algorithm: "exact" (default, uses VECTOR_DISTANCE) or "diskann" (uses VECTOR_SEARCH with DiskANN index, requires 1000+ rows)
+VECTOR_SEARCH_ALGORITHM=exact
 # Drop table at end of run to clean up artifacts (default: false)
 SQL_DROP_TABLE=false

samples/features/vector-search/vector-search-query-typescript/src/index.ts

@@ -186,7 +186,7 @@ async function main(): Promise<void> {
   console.log(`Loaded ${hotels.length} hotels from data file.\n`);
 
   // Validate vector dimensions for ALL hotels
-  const VECTOR_DIMENSIONS = 1536;
+  const VECTOR_DIMENSIONS = 1536; // text-embedding-3-small output dimensions
   const badVectors = hotels
     .map((h, i) => ({ index: i, id: h.HotelId, dim: h.DescriptionVector?.length }))
     .filter((v) => !v.dim || v.dim !== VECTOR_DIMENSIONS);
@@ -236,7 +236,7 @@ async function main(): Promise<void> {
   console.log(`Creating table dbo.${tableName} (if not exists)...`);
   await executeSql(
     conn,
-    `IF NOT EXISTS (SELECT * FROM sys.tables WHERE name = '${tableName}' AND schema_id = SCHEMA_ID('dbo'))
+    `IF NOT EXISTS (SELECT * FROM sys.tables WHERE name = N'${tableName}' AND schema_id = SCHEMA_ID('dbo'))
      BEGIN
        CREATE TABLE dbo.[${tableName}] (
          id NVARCHAR(50) PRIMARY KEY,
@@ -298,8 +298,8 @@ async function main(): Promise<void> {
     endpoint: config.azureOpenAiEndpoint,
     azureADTokenProvider,
     apiVersion: "2024-10-21",
-    timeout: 30_000,
-    maxRetries: 3,
+    timeout: 30_000,    // 30s timeout for embedding generation
+    maxRetries: 3,      // Retry transient failures
   });
 
   let queryEmbeddings: number[][];