3.3.2. Retrieval and Grounding

The retrieval step finds relevant documents; grounding ensures the model uses that context rather than hallucinating.

RAG Pattern:

# 1. Generate query embedding
embedding = client.embeddings.create(model="text-embedding-ada-002", input=query).data[0].embedding

# 2. Search for relevant documents
results = search_client.search(search_text=query, vector_queries=[VectorizedQuery(vector=embedding, fields="contentVector")])

# 3. Build prompt with context
context = "\n".join([doc["content"] for doc in results])
messages = [
    {"role": "system", "content": f"Answer based on this context:\n{context}"},
    {"role": "user", "content": query}
]

# 4. Generate grounded response
response = client.chat.completions.create(model="gpt-4o", messages=messages)

Azure OpenAI "On Your Data" (Built-in RAG):

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "What are our vacation policies?"}],
    extra_body={
        "data_sources": [{
            "type": "azure_search",
            "parameters": {
                "endpoint": search_endpoint,
                "index_name": index_name,
                "authentication": {"type": "api_key", "key": search_key}
            }
        }]
    }
)

Error Handling Pattern:

from openai import AzureOpenAI, APIError
from azure.core.exceptions import HttpResponseError

def rag_query_with_fallback(query: str) -> str:
    """RAG query with graceful degradation."""
    try:
        # Step 1: Generate embedding
        embedding = client.embeddings.create(
            model="text-embedding-ada-002", 
            input=query
        ).data[0].embedding
    except APIError as e:
        logging.error(f"Embedding generation failed: {e}")
        # Fallback to keyword search only
        embedding = None

    try:
        # Step 2: Search for context
        if embedding:
            results = search_client.search(
                search_text=query,
                vector_queries=[VectorizedQuery(vector=embedding, fields="contentVector")]
            )
        else:
            # Fallback: keyword search without vector
            results = search_client.search(search_text=query)
        
        context = "\n".join([doc["content"] for doc in results])
    except HttpResponseError as e:
        logging.error(f"Search failed: {e}")
        context = ""  # Proceed without context (acknowledge limitation to user)

    try:
        # Step 3: Generate response
        messages = [
            {"role": "system", "content": f"Answer based on: {context}" if context else "Answer to the best of your ability."},
            {"role": "user", "content": query}
        ]
        response = client.chat.completions.create(model="gpt-4o", messages=messages)
        return response.choices[0].message.content
    except APIError as e:
        logging.error(f"Generation failed: {e}")
        return "I'm sorry, I couldn't process your request. Please try again."

⚠️ Exam Trap: RAG implementations should handle failures at each stage (embedding, search, generation) independently with appropriate fallbacks.