Correct usage

[diego-lipinski-de-castro]

Currently we have two options

from langchain_postgres import PGVector
from langgraph.store.postgres import PostgresStore

Are they the made for the same use case?
Which one will be kept and maintained in the future?
What are their differences?

I tested both for vector stores and they both worked the same, besides that, what does one have that the other do not?

Example using PostgresStore

from langchain.embeddings import init_embeddings
from langgraph.store.postgres import PostgresStore, AsyncPostgresStore

with PostgresStore.from_conn_string(
    conn_string=os.getenv("VECTOR_DATABASE_URL"),
    index={
        "dims": 1536,
        "embed": init_embeddings("openai:text-embedding-ada-002"),
        # "fields": ["text"]
    }
) as store:
    store.setup()

    store.put(("docs",), "doc1", {"text": "Python tutorial"})
    store.put(("docs",), "doc2", {"text": "TypeScript guide"})
    store.put(("docs",), "doc3", {"text": "Ancient Mayan architecture and urban planning"})
    store.put(("docs",), "doc4", {"text": "Deep sea bioluminescent organisms"})
    store.put(("docs",), "doc5", {"text": "Traditional Japanese pottery techniques"})
    store.put(("docs",), "doc6", {"text": "History of Renaissance art movements"})
    store.put(("docs",), "doc7", {"text": "Quantum mechanics principles"})
    store.put(("docs",), "doc8", {"text": "Sustainable agriculture methods"})
    store.put(("docs",), "doc9", {"text": "Classical music composition theory"})
    store.put(("docs",), "doc10", {"text": "Volcanic formation processes"})

    results = store.search(("docs",), query="programming tutorial", limit=2)
    
    print(results)

Example using langchain_postgres

from langchain_postgres import PGVector
from langchain_core.documents import Document

embeddings = OpenAIEmbeddings(
    # text-embedding-3-large
    model="text-embedding-ada-002"
)

vector_store = PGVector(
    embeddings=embeddings,
    connection=os.getenv("VECTOR_DATABASE_URL"),
    collection_name="langchain",
    use_jsonb=True,
    create_extension=True,
)

documents = [
    Document(page_content="Python tutorial", metadata={"source": "doc1"}),
    Document(page_content="TypeScript guide", metadata={"source": "doc2"}),
    Document(page_content="Other guide", metadata={"source": "doc3"}),
    Document(page_content="Machine learning fundamentals", metadata={"source": "doc4"}),
    Document(page_content="Web development best practices", metadata={"source": "doc5"}),
    Document(page_content="Data structures and algorithms", metadata={"source": "doc6"}),
    Document(page_content="Cloud computing architecture", metadata={"source": "doc7"}),
    Document(page_content="Software testing methodologies", metadata={"source": "doc8"}),
    Document(page_content="DevOps principles and practices", metadata={"source": "doc9"}),
    Document(page_content="Database optimization techniques", metadata={"source": "doc10"})
]

vector_store.add_documents(documents=documents)

retriever = vector_store.as_retriever()

queries = [
    "software development",
    "ancient history",
    "natural sciences",
    "cultural practices"
]

for query in queries:
    results = retriever.invoke(query)
    print(f"\nQuery: {query}")
    for doc in results:
        print(f"- {doc.page_content}")