# Retrieval-Augmented Generation (RAG)

> Enhances LLM responses by retrieving relevant information from external databases before generating text.

- **Category**: Language Models
- **Subcategory**: general
- **Canonical URL**: https://designpattern.fyi/patterns/rag/

---

## Description
RAG is a pattern where the user query is used to search a vector database or search engine. The retrieved search results are injected into the prompt context, allowing the LLM to generate factual, up-to-date, and context-specific answers.



## Use Cases
Question-answering systems over internal company wikis, technical manuals, or PDF reports.



## Implementation Example

```python
# A simple Python RAG pipeline
class VectorDatabase:
    def similarity_search(self, query, k=1):
        # Simulated vector search
        return ["Design patterns are reusable solutions to common software problems."]

class LLMClient:
    def generate(self, prompt):
        print(f"Sending prompt to LLM:\n---\n{prompt}\n---")
        return "Based on the context, design patterns are reusable solutions..."

class RAGSystem:
    def __init__(self, db, llm):
        self.db = db
        self.llm = llm

    def query(self, user_prompt):
        # 1. Retrieve context
        context_docs = self.db.similarity_search(user_prompt, k=1)
        context = "\n".join(context_docs)
        
        # 2. Augment prompt
        augmented_prompt = (
            f"Use the following context to answer the question.\n"
            f"Context: {context}\n"
            f"Question: {user_prompt}"
        )
        
        # 3. Generate response
        return self.llm.generate(augmented_prompt)

# Usage
rag = RAGSystem(VectorDatabase(), LLMClient())
print(rag.query("What are design patterns?"))

```



## Trade-offs


### Advantages

- Dramatically reduces LLM hallucinations.

- Enables real-time data updates without retraining or fine-tuning.

- Provides source attribution/citations for generated answers.




### Considerations & Drawbacks

- Response latency increases due to the retrieval step.

- Retrieval accuracy directly bounds the generation quality.