rag-implementation

RAG Implementation Patterns

Comprehensive guide to implementing Retrieval-Augmented Generation (RAG) systems including vector database selection, chunking strategies, embedding models, retrieval optimization, and production deployment patterns.

---

Quick Reference

When to use this skill:

• Building RAG/semantic search systems
• Implementing document retrieval pipelines
• Optimizing vector database performance
• Debugging retrieval quality issues
• Choosing between vector database options
• Designing chunking strategies
• Implementing hybrid search

Technologies covered:

• Vector DBs: Qdrant, Pinecone, Chroma, Weaviate, Milvus
• Embeddings: OpenAI, Sentence Transformers, Cohere
• Frameworks: LangChain, LlamaIndex, Haystack

---

Part 1: Vector Database Selection

Database Comparison Matrix

Database	Best For	Deployment	Performance	Cost
Qdrant	Self-hosted, production	Docker/K8s	Excellent (Rust)	Free (self-host)
Pinecone	Managed, rapid prototyping	Cloud	Excellent	Pay-per-use
Chroma	Local development, embedded	In-process	Good (Python)	Free
Weaviate	Complex schemas, GraphQL	Docker/Cloud	Excellent (Go)	Free + Cloud
Milvus	Large-scale, distributed	K8s	Excellent (C++)	Free (self-host)

Qdrant Setup (Recommended for Production)

from qdrant_client import QdrantClient
from qdrant_client.models import Distance, VectorParams, PointStruct

# Initialize client (local or cloud)
client = QdrantClient(url="http://localhost:6333")  # or cloud URL

# Create collection
client.create_collection(
    collection_name="documents",
    vectors_config=VectorParams(
        size=1536,  # OpenAI text-embedding-3-small dimension
        distance=Distance.COSINE  # or DOT, EUCLID
    )
)

# Insert vectors with payload
client.upsert(
    collection_name="documents",
    points=[
        PointStruct(
            id=1,
            vector=[0.1, 0.2, ...],  # 1536 dimensions
            payload={
                "text": "Document content",
                "source": "doc.pdf",
                "page": 1,
                "metadata": {...}
            }
        )
    ]
)

# Search
results = client.search(
    collection_name="documents",
    query_vector=[0.1, 0.2, ...],
    limit=5,
    score_threshold=0.7  # Minimum similarity
)