mohamedamgad2002/Simple-RAG

1# import basics
2import os
3from dotenv import load_dotenv
4
5# import pinecone
6from pinecone import Pinecone, ServerlessSpec
7
8# import langchain
9from langchain_pinecone import PineconeVectorStore
10from langchain_google_genai import GoogleGenerativeAIEmbeddings
11from langchain_core.documents import Document
12
13load_dotenv()
14
15# initialize pinecone database
16pc = Pinecone(api_key=os.environ.get("PINECONE_API_KEY"))
17
18# set the pinecone index
19
20index_name = os.environ.get("PINECONE_INDEX_NAME") 
21index = pc.Index(index_name)
22
23# initialize embeddings model + vector store
24
25embeddings = GoogleGenerativeAIEmbeddings(model="models/gemini-embedding-001")
26
27vector_store = PineconeVectorStore(index=index, embedding=embeddings)
28
29# retrieval
30retriever = vector_store.as_retriever(
31    search_type="similarity_score_threshold",
32    search_kwargs={"k": 5, "score_threshold": 0.5},
33)
34results = retriever.invoke("what is retrieval augmented generation?")
35
36# show results
37print("RESULTS:")
38
39for res in results:
40    print(f"* {res.page_content} [{res.metadata}]")