Minor fixes

MScalopez · MScalopez · commit 6db871efa14c · 2026-01-15T13:56:05.000-06:00
diff --git a/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/2-understand-rag-patterns-azure-database-postgresql.yml b/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/2-understand-rag-patterns-azure-database-postgresql.yml
@@ -1,8 +1,8 @@
 ### YamlMime:ModuleUnit
 uid: learn.wwl.build-rag-applications-azure-database-postgresql.understand-rag-patterns-azure-database-postgresql
-title: Understand RAG Pattern with Azure Database for PostgreSQL
+title: Understand RAG pattern with Azure Database for PostgreSQL
 metadata:
-  title: Understand RAG Pattern with Azure Database for PostgreSQL
+  title: Understand RAG pattern with Azure Database for PostgreSQL
   description: "Learn what the Retrieval-Augmented Generation (RAG) pattern is and how its stages map to Azure Database for PostgreSQL features and extensions."
   ms.date: 08/08/2025
   author: wwlpublish
diff --git a/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/3-explore-scale-accuracy-problem.yml b/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/3-explore-scale-accuracy-problem.yml
@@ -1,8 +1,8 @@
 ### YamlMime:ModuleUnit
 uid: learn.wwl.build-rag-applications-azure-database-postgresql.explore-scale-accuracy-problem
-title: Explore Information retrieval challenges - scale and accuracy
+title: Explore information retrieval challenges - scale and accuracy
 metadata:
-  title: Explore Information retrieval challenges - scale and accuracy
+  title: Explore information retrieval challenges - scale and accuracy
   description: "Explore how information retrieval balances scale and accuracy, then set a baseline to measure and improve search performance."
   ms.date: 08/08/2025
   author: wwlpublish
diff --git a/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/6-exercise-build-rag-application-postgresql-python.yml b/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/6-exercise-build-rag-application-postgresql-python.yml
@@ -1,8 +1,8 @@
 ### YamlMime:ModuleUnit
 uid: learn.wwl.build-rag-applications-azure-database-postgresql.exercise-build-rag-application-postgresql-python
-title: "Exercise: Build RAG Applications with Azure Database for PostgreSQL and Python"
+title: "Exercise: Build RAG applications with Azure Database for PostgreSQL and Python"
 metadata:
-  title: "Exercise: Build RAG Applications with Azure Database for PostgreSQL and Python"
+  title: "Exercise: Build RAG applications with Azure Database for PostgreSQL and Python"
   description: "Build a RAG application in Azure Database for PostgreSQL."
   ms.date: 08/08/2025
   author: wwlpublish
diff --git a/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/7-improve-accuracy-advanced-rag-architecture.yml b/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/7-improve-accuracy-advanced-rag-architecture.yml
@@ -1,9 +1,9 @@
 ### YamlMime:ModuleUnit
 uid: learn.wwl.build-rag-applications-azure-database-postgresql.improve-accuracy-advanced-rag-architecture
-title: Improve Accuracy with Advanced RAG Architectures
+title: Improve accuracy with advanced RAG architectures
 metadata:
-  title: Improve Accuracy with Advanced RAG Architectures
-  description: "Improve Accuracy with Advanced RAG Architectures."
+  title: Improve accuracy with advanced RAG architectures
+  description: "Improve accuracy with advanced RAG architectures."
   ms.date: 08/08/2025
   author: wwlpublish
   ms.author: calopez
diff --git a/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/8-explore-graphrag.yml b/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/8-explore-graphrag.yml
@@ -10,4 +10,4 @@ metadata:
   ms.topic: unit
 durationInMinutes: 5
 content: |
-  [!include[](includes/8-explore-graph-rag.md)]
+  [!include[](includes/8-explore-graphrag.md)]
diff --git a/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/9-exercise-implement-graphrag.yml b/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/9-exercise-implement-graphrag.yml
@@ -10,4 +10,4 @@ metadata:
   ms.topic: unit
 durationInMinutes: 30
 content: |
-  [!include[](includes/9-exercise-implement-graph-rag.md)]
+  [!include[](includes/9-exercise-implement-graphrag.md)]
diff --git a/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/includes/1-introduction.md b/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/includes/1-introduction.md
@@ -1,6 +1,6 @@
 Retrieval Augmented Generation (RAG) applications combine information retrieval and generative AI to provide accurate, context-aware answers by using large datasets and advanced search techniques.
 
-Imagine you're a solution architect at a mid-sized enterprise tasked with building an internal knowledge assistant for employees to query company policies. The current system struggles with slow response times and often retrieves irrelevant or outdated information, frustrating users. Your team decides to implement a Retrieval Augmented Generation (RAG) application using *Azure Database for PostgreSQL* and *Azure OpenAI*. However, as the dataset grows to millions of rows, challenges like ensuring fast retrieval, maintaining accuracy, and avoiding hallucinated responses arise. Additionally, complex queries involving multiple concepts, such as 'What are the vacation policies for HR in Europe?' require precise filtering and ranking. To address these issues, you need to optimize retrieval speed, improve accuracy through advanced indexing and chunking strategies, and explore innovative solutions like lightweight knowledge graphs. This module guides you through building and refining a scalable RAG pipeline tailored to such real world challenges.
+Imagine you're a solution architect at a mid-sized enterprise tasked with building an internal knowledge assistant for employees to query company policies. The current system struggles with slow response times and often retrieves irrelevant or outdated information, frustrating users. Your team decides to implement a Retrieval Augmented Generation (RAG) application using *Azure Database for PostgreSQL* and *Azure OpenAI*. However, as the dataset grows to millions of rows, challenges like ensuring fast retrieval, maintaining accuracy, and avoiding AI-generated responses that might be incorrect. Additionally, complex queries involving multiple concepts, such as 'What are the vacation policies for HR in Europe?' require precise filtering and ranking. To address these issues, you need to optimize retrieval speed, improve accuracy through advanced indexing and chunking strategies, and explore innovative solutions like lightweight knowledge graphs. This module guides you through building and refining a scalable RAG pipeline tailored to such real world challenges.
 
 After completing this module, you can:
 
diff --git a/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/includes/11-summary.md b/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/includes/11-summary.md
@@ -12,13 +12,13 @@ After completing this module, you learn:
 
 Additional Reading:
 
-1. [Azure Database for PostgreSQL Documentation](/azure/postgresql/)  
-1. [pgvector Extension for PostgreSQL](https://github.com/pgvector/pgvector)  
-1. [LangChain Documentation](https://python.langchain.com/en/latest/)  
-1. [Azure OpenAI Service Overview](/azure/cognitive-services/openai/)
-1. [Tutorial: Create a recommendation system with Azure Database for PostgreSQL and Azure OpenAI](/azure/postgresql/flexible-server/generative-ai-semantic-search)
-1. [Introducing the GraphRAG Solution for Azure Database for PostgreSQL](https://techcommunity.microsoft.com/blog/adforpostgresql/introducing-the-graphrag-solution-for-azure-database-for-postgresql/4299871)
-1. [New Generative AI Features in Azure Database for PostgreSQL](https://techcommunity.microsoft.com/blog/adforpostgresql/new-generative-ai-features-in-azure-database-for-postgresql/4414858)
-1. [AGE extension with Azure Database for PostgreSQL](/azure/postgresql/flexible-server/generative-ai-age-overview)
-1. [Large language model end-to-end evaluation](/azure/architecture/ai-ml/guide/rag/rag-llm-evaluation-phase)
-1. [GraphRAG Solution Accelerator for Azure Database for PostgreSQL](https://github.com/Azure-Samples/graphrag-legalcases-postgres/)
+- [Azure Database for PostgreSQL Documentation](/azure/postgresql/)  
+- [pgvector Extension for PostgreSQL](https://github.com/pgvector/pgvector)  
+- [LangChain Documentation](https://python.langchain.com/en/latest/)  
+- [Azure OpenAI Service Overview](/azure/cognitive-services/openai/)
+- [Tutorial: Create a recommendation system with Azure Database for PostgreSQL and Azure OpenAI](/azure/postgresql/flexible-server/generative-ai-semantic-search)
+- [Introducing the GraphRAG Solution for Azure Database for PostgreSQL](https://techcommunity.microsoft.com/blog/adforpostgresql/introducing-the-graphrag-solution-for-azure-database-for-postgresql/4299871)
+- [New Generative AI Features in Azure Database for PostgreSQL](https://techcommunity.microsoft.com/blog/adforpostgresql/new-generative-ai-features-in-azure-database-for-postgresql/4414858)
+- [AGE extension with Azure Database for PostgreSQL](/azure/postgresql/flexible-server/generative-ai-age-overview)
+- [Large language model end-to-end evaluation](/azure/architecture/ai-ml/guide/rag/rag-llm-evaluation-phase)
+- [GraphRAG Solution Accelerator for Azure Database for PostgreSQL](https://github.com/Azure-Samples/graphrag-legalcases-postgres/)
diff --git a/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/includes/2-understand-rag-patterns-azure-database-postgresql.md b/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/includes/2-understand-rag-patterns-azure-database-postgresql.md
@@ -2,7 +2,7 @@ Imagine you’re building an internal assistant that answers employee questions
 
 ## Why RAG matters
 
-LLMs are trained on vast quantities of text but don’t know your organization’s data or proprietary content. To address this problem, RAG retrieves relevant information from trusted sources you provide, grounding the LLM model’s response. This approach improves accuracy, reduces hallucinations, and ensures answers are based on facts you control.
+LLMs are trained on vast quantities of text but don’t know your organization’s data or proprietary content. To address this problem, RAG retrieves relevant information from trusted sources you provide, grounding the LLM model’s response. This approach improves accuracy, reduces AI-generated responses that might be incorrect, and ensures answers are based on facts you control.
 
 ## How the RAG pipeline works
 
diff --git a/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/includes/3-explore-scale-accuracy-problem.md b/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/includes/3-explore-scale-accuracy-problem.md
@@ -50,6 +50,6 @@ While measuring *accuracy* isn't as straightforward as measuring for *scale*, yo
 
 As your application grows, reevaluate your application's performance and adjust your RAG pipeline as needed. It's important to monitor both retrieval speed and accuracy over time.
 
-## Key Takeaways
+## Key takeaways
 
 For a Postgres RAG path, you keep the retrieval in the database, start with `pgvector`, and move to the appropriate index as the data set and traffic grows. Add a *Semantic Ranking* pass in SQL if ordering needs to be tighter. Bring in *GraphRAG* to add a short graph step when relationships carry meaning. Evaluate regularly and monitor in production so the system stays both fast and correct as data and usage change.
diff --git a/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/includes/5-build-rag-application-postgresql-python.md b/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/includes/5-build-rag-application-postgresql-python.md
@@ -113,9 +113,9 @@ def generate_answer(question, chunks):
 
 > [!NOTE]
 > To use *AzureChatOpenAI*:
->> - Pass messages as a list of (role, content) tuples (or message objects).  
->> - Provide Azure settings via env vars and constructor: `AZURE_OPENAI_API_KEY`, `AZURE_OPENAI_ENDPOINT`, and `azure_deployment`, `api_version`.  
->> - Keep `temperature=0` for factual answers. Larger values increase creativity but might reduce accuracy.
+> - Pass messages as a list of (role, content) tuples (or message objects).  
+> - Provide Azure settings via env vars and constructor: `AZURE_OPENAI_API_KEY`, `AZURE_OPENAI_ENDPOINT`, and `azure_deployment`, `api_version`.  
+> - Keep `temperature=0` for factual answers. Larger values increase creativity but might reduce accuracy.
 
 This function is the core of the RAG application, handling the interaction with the language model. Notice how retrieved chunks are formatted and included in the context. Additionally, the system prompt is designed to ensure the model adheres to the context provided and reduces hallucination. Finally, the messages are processed by the language model to generate a response using the *LangChain* **invoke** method. The `invoke` method is called with the formatted messages, and the model's response is returned as natural language text.
 
diff --git a/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/includes/7-improve-accuracy-advanced-rag-architecture.md b/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/includes/7-improve-accuracy-advanced-rag-architecture.md
@@ -57,6 +57,6 @@ LIMIT 5;
 > [!TIP]
 > You can run separate keyword and vector queries and rerank the merged set (for example, with Reciprocal Rank Fusion) when you need both signals.
 
-## **Key Takeaways**
+## Key takeaways
 
 Accuracy isn’t just about better models, it’s about better retrieval. Start simple, *good chunking + hybrid search* often delivers big wins. *Advanced patterns* (GraphRAG, hierarchical indexing, agentic workflows) unlock new capabilities for complex environments. It's not just about quick searches, it's about finding the right information efficiently and the methods that can help achieve that.
diff --git a/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/includes/8-explore-graphrag.md b/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/includes/8-explore-graphrag.md
@@ -13,6 +13,7 @@ Text similarity is good at "this looks like that" answers, but large sets of sim
 When similarity alone isn't enough, **GraphRAG** is a method from *Microsoft Research* that improves RAG by extracting a knowledge graph from your source data and using that structure to supply better context to the Large Language Model (LLM). It has three main steps:
 
 1. **Graph extraction**
+
    In the company policy database example, use the `company_policies` table as the source. Each row becomes a `Policy` node with `policy_id`, `title`, `policy_text`, `department`, and `category` as properties. Create `Entity` nodes from values you already have (one per department and one per category). Optionally add `Entity` nodes for important terms found in `policy_text` (such as "taxi reimbursement" or "airport"). Connect them with edges:
 
    - `(:Policy)-[:BELONGS_TO]->(:Entity {type: 'Department', name: department})`
@@ -37,6 +38,6 @@ Use *GraphRAG* in *Azure Database for PostgreSQL* to keep data and relationships
 
 You can keep reranking inside the database with SQL semantic operators. The `rank()` operator from the `azure_ai` semantic operators lets you rerank top candidates with state of the art models directly in a query, so you can combine semantic ranking with graph scoring without leaving Postgres. As your data grows, pair the graph step with the right vector index. `DiskANN` in *Azure Database for PostgreSQL* supports high dimensional embeddings (up to 16,000 dimensions), faster index builds, and a Product Quantization that can reduce memory and cost while maintaining high accuracy. Reported gains include up to 10× faster performance and about 4× cost savings versus an *HNSW* (Hierarchical Navigable Small Worlds) index.
 
-## Key Takeaways
+## Key takeaways
 
 A similarity-only search often falls short on real questions. *GraphRAG* with *Apache AGE* adds a knowledge graph so the model gets structure and meaning. You extract relationships into an *AGE* graph, keep short summaries on the nodes, and at question time run vector search and an *openCypher* query to see how the candidates connect. In *Azure Database for PostgreSQL*, this process all stays in one place. Tables hold the content, *Apache AGE* manages the graph, `pgvector` pulls likely passages, and a light rerank with simple fusion tightens the order. The result is a smaller, clearer context that points to the right sources and produces answers you can verify.
diff --git a/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/includes/9-exercise-implement-graphrag.md b/learn-pr/wwl-data-ai/build-rag-applications-azure-database-postgresql/includes/9-exercise-implement-graphrag.md
