Voyage AI

# Install the Voyage AI SDK
# pip install voyageai

import voyageai

# Initialize the client
client = voyageai.Client(api_key="your-api-key")

# Generate embeddings for documents
documents = [
    "Voyage AI provides state-of-the-art embedding models.",
    "RAG applications benefit from high-quality embeddings.",
    "Semantic search requires good vector representations."
]

# Embed documents
doc_embeddings = client.embed(
    texts=documents,
    model="voyage-3",
    input_type="document"
)

# Embed a query
query = "What are embedding models used for?"
query_embedding = client.embed(
    texts=[query],
    model="voyage-3",
    input_type="query"
)

# Access the vectors
print(f"Document embeddings shape: {len(doc_embeddings.embeddings)}x{len(doc_embeddings.embeddings[0])}")
print(f"Query embedding shape: 1x{len(query_embedding.embeddings[0])}")
print(f"Total tokens used: {doc_embeddings.total_tokens + query_embedding.total_tokens}")

from langchain_voyageai import VoyageAIEmbeddings
from langchain_community.vectorstores import Chroma
from langchain.text_splitter import RecursiveCharacterTextSplitter

# Initialize Voyage AI embeddings
embeddings = VoyageAIEmbeddings(
    voyage_api_key="your-api-key",
    model="voyage-3"
)

# Split documents
text_splitter = RecursiveCharacterTextSplitter(
    chunk_size=1000,
    chunk_overlap=200
)
chunks = text_splitter.split_documents(documents)

# Create vector store with Voyage embeddings
vectorstore = Chroma.from_documents(
    documents=chunks,
    embedding=embeddings,
    persist_directory="./chroma_db"
)

# Retrieve relevant documents
retriever = vectorstore.as_retriever(
    search_type="similarity",
    search_kwargs={"k": 5}
)

# Query the retriever
docs = retriever.invoke("What is the main topic?")
for doc in docs:
    print(doc.page_content[:200])