[FEAT] Add checkpoint for long-running GraphWorkflows

[adichaudhary]

Description

This PR adds checkpoint-based fault tolerance to GraphWorkflow. Long-running pipelines that crash or time out mid-execution previously required a full restart from scratch, re-running every agent and paying for every LLM call again. Now, a checkpoint_dir parameter can be passed at construction time — after each layer completes its outputs are persisted to disk, and on the next run() call with the same task string any already-completed layers are loaded from the checkpoint files and skipped entirely. A clear_checkpoints(task) method is also provided to clean up after a confirmed successful run.

Files Changed

• swarms/structs/graph_workflow.py — added checkpoint_dir parameter to __init__, checkpoint save/resume logic in the layer execution loop of run(), and the clear_checkpoints() method
• examples/multi_agent/graphworkflow_examples/graph_workflow_checkpointing.py

Issue

#1484

Dependencies

No extra dependencies required.

Maintainer

@kyegomez

Twitter

@akc__2025

📚 Documentation preview 📚: https://swarms--1451.org.readthedocs.build/en/1451/

[Copilot]

Pull request overview

Adds checkpoint-based fault tolerance to GraphWorkflow so long-running, multi-layer executions can resume without re-running completed layers.

Changes:

• Add checkpoint_dir parameter to GraphWorkflow.__init__ and implement per-layer checkpoint save/resume in run().
• Add clear_checkpoints(task) helper to remove checkpoint files for a given task.
• Add an example script demonstrating checkpointing and a simulated crash/re-run flow.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 5 comments.

File	Description
swarms/structs/graph_workflow.py	Introduces checkpoint_dir, checkpoint save/resume logic in run(), and clear_checkpoints() cleanup method.
examples/multi_agent/graphworkflow_examples/graph_workflow_checkpointing.py	New example showcasing checkpoint usage and a simulated crash scenario.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.