Building RAG Applications with LangChain, Pinecone, and OpenAI: A Complete Tutorial

Retrieval Augmented Generation (RAG) has emerged as the most practical approach for building AI applications that can reason over your own data. In this technical deep-dive, we'll build a production-ready RAG system from scratch.

Why RAG Over Fine-Tuning?

Fine-tuning LLMs is expensive, time-consuming, and the model can still hallucinate. RAG provides real-time access to your data, is more cost-effective, and allows you to update your knowledge base without retraining.

Architecture Overview

Our RAG system consists of: 1) Document ingestion pipeline, 2) Vector embedding generation, 3) Pinecone vector storage, 4) Query processing with semantic search, 5) LLM response generation with retrieved context.

Setting Up the Environment

pip install langchain langchain-openai pinecone-client python-dotenv tiktoken

Create your .env file with: OPENAI_API_KEY, PINECONE_API_KEY, and PINECONE_ENVIRONMENT.

Document Loading and Chunking

The quality of your RAG system depends heavily on how you chunk documents. We use recursive character splitting with overlap:

from langchain.text_splitter import RecursiveCharacterTextSplitter

text_splitter = RecursiveCharacterTextSplitter(
    chunk_size=1000,
    chunk_overlap=200,
    length_function=len,
    separators=["\n\n", "\n", " ", ""]
)

Creating Embeddings with OpenAI

We use text-embedding-3-small for cost efficiency or text-embedding-3-large for better accuracy:

from langchain_openai import OpenAIEmbeddings

embeddings = OpenAIEmbeddings(
    model="text-embedding-3-small",
    dimensions=1536
)

Pinecone Vector Store Setup

from pinecone import Pinecone, ServerlessSpec

pc = Pinecone(api_key=os.getenv("PINECONE_API_KEY"))

pc.create_index(
    name="rag-index",
    dimension=1536,
    metric="cosine",
    spec=ServerlessSpec(cloud="aws", region="us-east-1")
)

Building the Retrieval Chain

from langchain.chains import RetrievalQA
from langchain_openai import ChatOpenAI

llm = ChatOpenAI(model="gpt-4-turbo-preview", temperature=0)

qa_chain = RetrievalQA.from_chain_type(
    llm=llm,
    chain_type="stuff",
    retriever=vectorstore.as_retriever(search_kwargs={"k": 5}),
    return_source_documents=True
)

Optimizing Retrieval Quality

Key optimizations include: Hybrid search (combining keyword and semantic), re-ranking retrieved documents, metadata filtering, and query expansion using HyDE (Hypothetical Document Embeddings).

Production Considerations

For production deployments, implement: caching for repeated queries, async processing for document ingestion, monitoring with LangSmith, and rate limiting for API calls.

NyxaLabs RAG Implementation Services

We've built RAG systems processing millions of documents for enterprise clients. Our expertise spans document processing, vector optimization, and LLM orchestration. Contact us for your AI project.