What is RAG and Why You Need It
LLMs do not know your internal documents. RAG (Retrieval-Augmented Generation) retrieves relevant content from YOUR documents at query time and gives it to the LLM as context. The LLM answers based on YOUR data, not just its training.
Install Dependencies
bash
pip install anthropic langchain langchain-community chromadb pypdf sentence-transformers
Complete RAG System
python
import anthropic
from langchain_community.document_loaders import PyPDFLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain_community.vectorstores import Chroma
client = anthropic.Anthropic()
def load_documents(paths):
all_docs = []
for p in paths:
all_docs.extend(PyPDFLoader(p).load())
chunks = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=50).split_documents(all_docs)
embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
return Chroma.from_documents(chunks, embeddings)
def answer(vectorstore, question):
docs = vectorstore.similarity_search(question, k=4)
context = "\n\n".join([d.page_content for d in docs])
r = client.messages.create(
model="claude-haiku-4-5-20251001",
max_tokens=800,
system="Answer ONLY from the provided context. Say so if the answer is not there.",
messages=[{"role": "user", "content": f"Context:\n{context}\n\nQuestion: {question}"}]
)
return r.content[0].text
vs = load_documents(["company_handbook.pdf"])
print(answer(vs, "What is the refund policy?"))
Add Streamlit UI
python
import streamlit as st
st.title("Document Q&A")
uploaded = st.file_uploader("Upload PDFs", type="pdf", accept_multiple_files=True)
if uploaded:
# save to temp files, load_documents(), show chat interface
pass
What Makes a Production RAG Better
This is a basic implementation. Production systems add: hybrid search (BM25 + vectors), reranking, query rewriting, and evaluation pipelines. See the RAG section in the AI Academy for advanced patterns.
