Merge pull request #53921 from GraemeMalcolm/main

Switch to sentence casing and add explicit prompt

Author: prmerger-automator[bot]

learn-pr/wwl-data-ai/use-generative-ai-tools/includes/03-code-interpreter.md

@@ -12,7 +12,7 @@ Key features include:
 - **Real-time Feedback**: The model sees code execution results and can iterate or fix errors
 - **Complex Problem Solving**: Tackle math problems, simulations, and logic puzzles through executable code
 
-## Common Use Cases
+## Common use cases
 
 | Use Case | Example |
 |----------|---------|
@@ -70,7 +70,7 @@ The general process for using the code_interpreter tool is:
 4. **Code runs**: The code runs in a sandboxed environment with access to common libraries (for example, *pandas*, *numpy*, and *math*).
 5. **Results returned**: The model receives the output and incorporates it into its response.
 
-## Best Practices
+## Best practices
 
 - **Be specific**: Describe the data format and expected output clearly. Many models internally use the name *python tool* to identify the code_interpreter tool - so use this language in your instructions.
 - **Provide context**: Include relevant domain knowledge in your prompts

learn-pr/wwl-data-ai/use-generative-ai-tools/includes/04-web-search.md

@@ -14,7 +14,7 @@ Key features include:
 - **Automatic query generation** - The model decides when and how to search based on user intent
 - **Seamless user experience** - Search and response generation happen in one flow
 
-## Common Use Cases
+## Common use cases
 
 | Use Case | Example |
 |----------|---------|
@@ -61,7 +61,7 @@ The general process for using the web_search tool is:
 4. **Results are reviewed** - Relevant pages are selected and summarized.
 5. **Response is generated** - The model combines search findings into the final answer.
 
-## Best Practices
+## Best practices
 
 - **Ask time-aware questions clearly** - Include words like "latest", "current", or date ranges when needed
 - **Set expectations for sources** - Prompt for reputable or official sources when accuracy matters

learn-pr/wwl-data-ai/use-generative-ai-tools/includes/05-file-search.md

@@ -14,7 +14,7 @@ Key features include:
 - **Citations and transparency** - Include matched results for debugging and traceability
 - **Better enterprise relevance** - Use organization-specific knowledge in model outputs
 
-## Common Use Cases
+## Common use cases
 
 | Use Case | Example |
 |----------|---------|
@@ -46,7 +46,7 @@ client.vector_stores.files.upload_and_poll(
 response = client.responses.create(
     model=model_deployment,
     instructions="You are an AI assistant that provides information from HR policy documents.",
-    input=prompt,
+    input="What's the maximum amount I can claim for a tsxi ride?",
     tools=[{
         "type": "file_search",
         "vector_store_ids": [vector_store.id]
@@ -68,7 +68,7 @@ The general process for using the file_search tool is:
 4. **Results are injected** - Matching passages are provided to the model.
 5. **Response is generated** - The model answers using retrieved document context.
 
-## Best Practices
+## Best practices
 
 - **Use high-quality source files** - Clean, current documents improve retrieval accuracy
 - **Write focused prompts** - Ask specific questions to reduce ambiguous matches

learn-pr/wwl-data-ai/use-generative-ai-tools/includes/06-function.md

@@ -14,7 +14,7 @@ Key features include:
 - **Multi-turn orchestration** - Return tool output and let the model continue reasoning
 - **Grounded responses** - Answers can include live, system-generated data
 
-## Common Use Cases
+## Common use cases
 
 | Use Case | Example |
 |----------|---------|
@@ -134,7 +134,7 @@ The general process for using the function tool is:
 5. **You return function output** - Send a `function_call_output` item with the result.
 6. **Model completes the answer** - It incorporates tool results into the final response.
 
-## Best Practices
+## Best practices
 
 - **Keep tools focused** - Small, single-purpose functions are easier to control and test
 - **Validate function inputs** - Never trust tool arguments blindly in production systems