langchain-postgres

The langchain-postgres package implementations of core LangChain abstractions using Postgres.

The package is released under the MIT license.

Feel free to use the abstraction as provided or else modify them / extend them as appropriate for your own application.

Requirements

The package currently only supports the psycogp3 driver.

Installation

pip install -U langchain-postgres

Usage

PostgresSaver (LangGraph Checkpointer)

The LangGraph checkpointer can be used to add memory to your LangGraph application.

PostgresSaver is an implementation of the checkpointer saver using Postgres as the backend.

Currently, only the psycopg3 driver is supported.

Sync usage:

from psycopg_pool import ConnectionPool
from langchain_postgres import (
    PostgresSaver, PickleCheckpointSerializer
)

pool = ConnectionPool(
    # Example configuration
    conninfo="postgresql://langchain:langchain@localhost:6024/langchain",
    max_size=20,
)

PostgresSaver.create_tables(pool)

checkpointer = PostgresSaver(
    serializer=PickleCheckpointSerializer(),
    sync_connection=pool,
)

# Set up the langgraph workflow with the checkpointer
workflow = ... # Fill in with your workflow
app = workflow.compile(checkpointer=checkpointer)

# Use with the sync methods of `app` (e.g., `app.stream())

pool.close() # Remember to close the connection pool.

Async usage:

from psycopg_pool import AsyncConnectionPool
from langchain_postgres import (
    PostgresSaver, PickleCheckpointSerializer
)

pool = AsyncConnectionPool(
    # Example configuration
    conninfo="postgresql://langchain:langchain@localhost:6024/langchain",
    max_size=20,
)

# Create the tables in postgres (only needs to be done once)
await PostgresSaver.acreate_tables(pool)

checkpointer = PostgresSaver(
    serializer=PickleCheckpointSerializer(),
    async_connection=pool,
)

# Set up the langgraph workflow with the checkpointer
workflow = ... # Fill in with your workflow
app = workflow.compile(checkpointer=checkpointer)

# Use with the async methods of `app` (e.g., `app.astream()`)

await pool.close() # Remember to close the connection pool.

Testing

If testing with the postgres checkpointer it may be useful to both create and drop the tables before and after the tests.

from psycopg_pool import ConnectionPool
from langchain_postgres import (
    PostgresSaver 
)
with ConnectionPool(
    # Example configuration
    conninfo="postgresql://langchain:langchain@localhost:6024/langchain",
    max_size=20,
) as conn:
    PostgresSaver.create_tables(conn)
    PostgresSaver.drop_tables(conn)
    # Run your unit tests with langgraph

ChatMessageHistory

The chat message history abstraction helps to persist chat message history in a postgres table.

PostgresChatMessageHistory is parameterized using a table_name and a session_id.

The table_name is the name of the table in the database where the chat messages will be stored.

The session_id is a unique identifier for the chat session. It can be assigned by the caller using uuid.uuid4().

import uuid

from langchain_core.messages import SystemMessage, AIMessage, HumanMessage
from langchain_postgres import PostgresChatMessageHistory
import psycopg

# Establish a synchronous connection to the database
# (or use psycopg.AsyncConnection for async)
conn_info = ... # Fill in with your connection info
sync_connection = psycopg.connect(conn_info)

# Create the table schema (only needs to be done once)
table_name = "chat_history"
PostgresChatMessageHistory.create_tables(sync_connection, table_name)

session_id = str(uuid.uuid4())

# Initialize the chat history manager
chat_history = PostgresChatMessageHistory(
    table_name,
    session_id,
    sync_connection=sync_connection
)

# Add messages to the chat history
chat_history.add_messages([
    SystemMessage(content="Meow"),
    AIMessage(content="woof"),
    HumanMessage(content="bark"),
])

print(chat_history.messages)

Vectorstore

See example for the PGVector vectorstore here