LangGraph Checkpoint Greenplum

Greenplum/MPP-optimized checkpoint savers for LangGraph. This package provides drop-in replacements for PostgresSaver and AsyncPostgresSaver that use a CTE-based list query instead of correlated subqueries, avoiding Broadcast Motion in Greenplum's MPP execution engine.

Why use this?

On Greenplum (and similar MPP databases), the default list query in langgraph-checkpoint-postgres uses correlated subqueries. The planner may broadcast the full inner tables to every segment, so list operations can be very slow on large datasets. The Greenplum savers use a CTE + JOIN pattern so the planner uses Redistribute Motion instead, often yielding orders-of-magnitude speedups for list() / alist().

Installation

Install the library and its dependency:

pip install langgraph-checkpoint-greenplum

This will pull in langgraph-checkpoint-postgres as well.

Usage

Use AsyncGreenplumSaver or GreenplumSaver exactly like the Postgres savers. Call .setup() the first time to create tables.

Async

from langgraph.checkpoint.greenplum import AsyncGreenplumSaver

DB_URI = "postgres://user:pass@host:5432/dbname"
async with AsyncGreenplumSaver.from_conn_string(DB_URI) as checkpointer:
    await checkpointer.setup()
    # use as a drop-in replacement for AsyncPostgresSaver

Sync

from langgraph.checkpoint.greenplum import GreenplumSaver

DB_URI = "postgres://user:pass@host:5432/dbname"
with GreenplumSaver.from_conn_string(DB_URI) as checkpointer:
    checkpointer.setup()
    # use as a drop-in replacement for PostgresSaver

Requirements

• Greenplum 6+ (or PostgreSQL 9.3+) for the CTE query (uses CROSS JOIN LATERAL).
• langgraph-checkpoint-postgres (and thus psycopg and langgraph-checkpoint) are installed automatically.

Release to PyPI (GitHub Actions)

This repository includes a publish workflow at .github/workflows/publish-pypi.yml.

One-time setup

1. In PyPI, configure a Trusted Publisher for this project:
   • Owner: StarJourneyMingsing
   • Repository: langgraph-checkpoint-greenplum
   • Workflow name: publish-pypi.yml
   • Environment name: pypi
2. In GitHub, create environment pypi in repository settings (optional protection rules can be added).

Publish a new version

1. Update version in pyproject.toml.
2. Commit and push changes.
3. Create and push a version tag:

git tag v0.1.1
git push origin v0.1.1

The workflow will verify that tag version matches pyproject.toml, build distributions, and publish to PyPI.

License

MIT