6 documents into text PGvectorStore

Menyimpan, mengakses dan merubah data pada vector store¶

In [1]:

from langchain_postgres import PGVectorStore, PGEngine
from langchain_community.document_loaders import TextLoader
from langchain_text_splitters import RecursiveCharacterTextSplitter
from langchain_ollama.embeddings import OllamaEmbeddings

from langchain_postgres import PGVectorStore, PGEngine from langchain_community.document_loaders import TextLoader from langchain_text_splitters import RecursiveCharacterTextSplitter from langchain_ollama.embeddings import OllamaEmbeddings

One of the requirements and arguments to establish PostgreSQL as a vector store is a PGEngine object. The PGEngine configures a shared connection pool to your Postgres database. This is an industry best practice to manage number of connections and to reduce latency through cached database connections.

In [2]:

## Membuat koneksi terlebih menggunakan class PGEngine
CONNECTION = "postgresql+psycopg://farras:farras@localhost:6060/farraslang"

## Engine
pg_engine = PGEngine.from_connection_string(CONNECTION)

## Membuat koneksi terlebih menggunakan class PGEngine CONNECTION = "postgresql+psycopg://farras:farras@localhost:6060/farraslang" ## Engine pg_engine = PGEngine.from_connection_string(CONNECTION)

In [3]:

## Embeddings
model_embeddings = OllamaEmbeddings(model="nomic-embed-text")

## Eksisting table name
TABLE_NAME = 'my_pg_vector_store'

## Inisiasi PGVectoreStore menggunakan table baru
# THis only run once
try:
    await pg_engine.ainit_vectorstore_table(
        table_name=TABLE_NAME,
        vector_size=768
    )
except Exception as e:
    print(f"Datase bisa jadi sudah ada {e}")
## Above code akan membuat table dengan nama my_pg_vector_store dengan kolomabs
## * langchain_id
## * content
## * embedding
## * langchain_metadata

## Embeddings model_embeddings = OllamaEmbeddings(model="nomic-embed-text") ## Eksisting table name TABLE_NAME = 'my_pg_vector_store' ## Inisiasi PGVectoreStore menggunakan table baru # THis only run once try: await pg_engine.ainit_vectorstore_table( table_name=TABLE_NAME, vector_size=768 ) except Exception as e: print(f"Datase bisa jadi sudah ada {e}") ## Above code akan membuat table dengan nama my_pg_vector_store dengan kolomabs ## * langchain_id ## * content ## * embedding ## * langchain_metadata

Datase bisa jadi sudah ada (psycopg.errors.DuplicateTable) relation "my_pg_vector_store" already exists
[SQL: CREATE TABLE "public"."my_pg_vector_store"(
            "langchain_id" UUID PRIMARY KEY,
            "content" TEXT NOT NULL,
            "embedding" vector(768) NOT NULL
            ,
"langchain_metadata" JSON
);]
(Background on this error at: https://sqlalche.me/e/20/f405)

In [4]:

model_embeddings = OllamaEmbeddings(model="nomic-embed-text")

class_vector_store = await PGVectorStore.create(
    engine=pg_engine,
    table_name=TABLE_NAME,
    embedding_service=model_embeddings
)

model_embeddings = OllamaEmbeddings(model="nomic-embed-text") class_vector_store = await PGVectorStore.create( engine=pg_engine, table_name=TABLE_NAME, embedding_service=model_embeddings )

In [5]:

root_path = "C:/Users/maruf/Documents/LangChainLabAssets"
text_data_store_vector = TextLoader(f'{root_path}/BigDataForTwo.txt')

## Memeuat loader kedalam bentuk class Documents
docs = text_data_store_vector.load()

## Potong big data kedalam potongan2an
splitter_store_vector = RecursiveCharacterTextSplitter(
    chunk_size = 1000,
    chunk_overlap = 200
)

## Class document baru setelah di potong
new_docs = splitter_store_vector.split_documents(docs)

root_path = "C:/Users/maruf/Documents/LangChainLabAssets" text_data_store_vector = TextLoader(f'{root_path}/BigDataForTwo.txt') ## Memeuat loader kedalam bentuk class Documents docs = text_data_store_vector.load() ## Potong big data kedalam potongan2an splitter_store_vector = RecursiveCharacterTextSplitter( chunk_size = 1000, chunk_overlap = 200 ) ## Class document baru setelah di potong new_docs = splitter_store_vector.split_documents(docs)

Store data ke vector store¶

In [6]:

from langchain_core.documents import Document
from uuid import uuid4
## Membuat documentsa

contents = ['Bismillah, semoga Allah luluskan aku pada kali ini menjadi Aparatur Sipil Negara',
            'Allah maha menerima doa hambanya, Allah selalu mendeger suara hambanya, Allah maha mengabulkan doa hambanya']

documents = [Document(content, metadata={'jenis':f'nasihat ke {angka}','author':'muhammad farras'}) for angka, content in enumerate(contents)]

class_vector_store.add_documents(documents)

from langchain_core.documents import Document from uuid import uuid4 ## Membuat documentsa contents = ['Bismillah, semoga Allah luluskan aku pada kali ini menjadi Aparatur Sipil Negara', 'Allah maha menerima doa hambanya, Allah selalu mendeger suara hambanya, Allah maha mengabulkan doa hambanya'] documents = [Document(content, metadata={'jenis':f'nasihat ke {angka}','author':'muhammad farras'}) for angka, content in enumerate(contents)] class_vector_store.add_documents(documents)

Out[6]:

['dc07ca72-5a18-4dc9-88f1-6f6efcdbc591',
 'a4f5451a-01e7-4ccd-994b-9c7b594604cd']

Data diatas akan membentuk langchain_id UUID, namun kita juga dapat membuatnya secara manual guna melakuakn indexing

In [7]:

from langchain_core.documents import Document
import uuid

contents2 = ['Bismillah, semoga Allah luluskan aku pada kali ini menjadi Aparatur Sipil Negara']

documents = [Document(content, metadata={'jenis':f'nasihat ke 3','author':'muhammad farras'}) for angka, content in enumerate(contents2)]

class_vector_store.add_documents(documents,[str(uuid.uuid4())])

from langchain_core.documents import Document import uuid contents2 = ['Bismillah, semoga Allah luluskan aku pada kali ini menjadi Aparatur Sipil Negara'] documents = [Document(content, metadata={'jenis':f'nasihat ke 3','author':'muhammad farras'}) for angka, content in enumerate(contents2)] class_vector_store.add_documents(documents,[str(uuid.uuid4())])

Out[7]:

['cf3e12bf-41df-4c89-ac35-6356d00517ad']

Delete data using ID index¶

untuk menghapus data kita dapat menggunakan method delete dengan parameter pertama ids diisi dengan list of number UUID.

In [8]:

class_vector_store.delete(ids=['ea619ca2-5fdd-4678-b882-d48dc1db02fd'])

class_vector_store.delete(ids=['ea619ca2-5fdd-4678-b882-d48dc1db02fd'])

Out[8]:

True

Mencari similiary data¶

In [9]:

docs_for_similiary_search = class_vector_store.similarity_search("visi utama", k=1)
docs_for_similiary_search

docs_for_similiary_search = class_vector_store.similarity_search("visi utama", k=1) docs_for_similiary_search

Out[9]:

[Document(id='f30472e6-c2e5-5777-8279-da9bf3aee327', metadata={'tanggal': '2026-02-08 10:10:34.207912', 'di': 'rumah 2'}, page_content='saya sedang belajar langchain juga hari ini.')]

In [10]:

docs_for_similiary_search = class_vector_store.similarity_search("jenenge sopo", k=1)
docs_for_similiary_search

docs_for_similiary_search = class_vector_store.similarity_search("jenenge sopo", k=1) docs_for_similiary_search

Out[10]:

[Document(id='ec2f09a3-f68b-5376-8f30-2a2f163f7398', metadata={'tanggal': '2026-02-08 10:10:34.207796', 'di': 'rst'}, page_content='saya hari ini lagi diruma sakit RST Dompet duafa, saya sendang mengidap penyakit infeksi paru')]

Mencari similiary data dengan score¶

In [11]:

docs_for_similiary_search = class_vector_store.similarity_search_with_relevance_scores ("Semoga tahun ini adalah tahun terkahir di Vensys", k=1)
docs_for_similiary_search[0][1]

docs_for_similiary_search = class_vector_store.similarity_search_with_relevance_scores ("Semoga tahun ini adalah tahun terkahir di Vensys", k=1) docs_for_similiary_search[0][1]

Out[11]:

0.6741104728678319

In [12]:

docs_for_similiary_search = class_vector_store.similarity_search_with_relevance_scores ("Jenenge sopo", k=1)
print(docs_for_similiary_search)

docs_for_similiary_search = class_vector_store.similarity_search_with_relevance_scores ("Jenenge sopo", k=1) print(docs_for_similiary_search)

[(Document(id='70fafa1b-7ec4-4e57-ad75-cac910eeda34', metadata={'jenis': 'nasihat ke 1', 'author': 'muhammad farras'}, page_content='Allah maha menerima doa hambanya, Allah selalu mendeger suara hambanya, Allah maha mengabulkan doa hambanya'), 0.4920965986770105)]

Menggunakan asimilarity_search¶

In [13]:

query = "I'd like a fruit."
docs = await class_vector_store.asimilarity_search_with_relevance_scores(query, k=1)
print(docs)

query = "I'd like a fruit." docs = await class_vector_store.asimilarity_search_with_relevance_scores(query, k=1) print(docs)

[(Document(id='70fafa1b-7ec4-4e57-ad75-cac910eeda34', metadata={'jenis': 'nasihat ke 1', 'author': 'muhammad farras'}, page_content='Allah maha menerima doa hambanya, Allah selalu mendeger suara hambanya, Allah maha mengabulkan doa hambanya'), 0.5015093091748534)]