.. _uses:

Use Cases
+++++++++

Examples By Category
====================

.. raw:: html

    <div class="tiles">
        <a href="cbook/multiprocessing.html" class="tile_item">
            <h3>Multiprocessing</h3>
            <p>Distributed Python examples for queues, pools, pipelines, and shared state</p>
        </a>
        <a href="cbook/data.html" class="tile_item">
            <h3>Data</h3>
            <p>Distributed data sharing with a DDict-focused walkthrough</p>
        </a>
        <a href="cbook/telemetry.html" class="tile_item">
            <h3>Telemetry</h3>
            <p>Collect and inspect Dragon runtime and application metrics</p>
        </a>
        <a href="cbook/workflows.html" class="tile_item">
            <h3>Workflow</h3>
            <p>Workflow patterns for AI loops, MPI orchestration, Parsl, and Batch DAGs</p>
        </a>
        <a href="cbook/ai.html" class="tile_item">
            <h3>AI</h3>
            <p>AI examples for data loading, agents, and inference-oriented workflows</p>
        </a>
    </div>

.. toctree::
    :maxdepth: 2
    :hidden:

    cbook/multiprocessing.rst
    cbook/data.rst
    cbook/telemetry.rst
    cbook/workflows.rst
    cbook/ai.rst
    cbook/core.rst

|

Tutorials
=========

.. raw:: html

    <div class="tiles">
        <a href="uses/data_processing.html" class="tile_item">
            <h3>Data processing</h3>
            <p>Process a large dataset in parallel</p>
        </a>
        <a href="uses/orchestrate_procs.html" class="tile_item">
            <h3>Orchestrating processes</h3>
            <p>Orchestrate executing a serial executable with many different arguments</p>
        </a>
        <a href="uses/orchestrate_mpi.html" class="tile_item">
            <h3>Orchestrating MPI applications</h3>
            <p>Orchestrate a parameter sweep for an MPI application</p>
        </a>
        <a href="uses/distributed_training.html" class="tile_item">
            <h3>Distributed PyTorch</h3>
            <p>Set up PyTorch distributed launch, rank info, and data loading on GPUs</p>
        </a>
        <a href="uses/ddict_checkpoint.html" class="tile_item">
            <h3>Resiliency with DDict Checkpointing</h3>
            <p>Track application state to enable fault tolerance and recovery</p>
        </a>
    </div>
    <div class="tiles">
        <a href="uses/multinode.html" class="tile_item">
            <h3>Running across nodes</h3>
            <p>Run on a collection of servers or a supercomputer</p>
        </a>
        <a href="uses/jupyter.html" class="tile_item">
            <h3>Jupyter</h3>
            <p>Use a Jupyter notebook with Dragon</p>
        </a>
        <a href="uses/gpus.html" class="tile_item">
            <h3>Controlling GPU Affinity</h3>
            <p>Running functions and processes that use specific GPUs</p>
        </a>
        <a href="uses/grafana.html" class="tile_item">
            <h3>Telemetry with Grafana</h3>
            <p>Visualize and analyze system and custom metrics</p>
        </a>
        <a href="uses/workflow.html" class="tile_item">
            <h3>Workflows</h3>
            <p>Build a simple end-to-end workflow with Batch, DDict, and external jobs</p>
        </a>
    </div>
    <div class="tiles">
        <a href="uses/proxy.html" class="tile_item">
            <h3>Running multi-system workflows</h3>
            <p>Connect and run workflows across multiple systems</p>
        </a>
        <a href="uses/agent.html" class="tile_item">
            <h3>Building AI Agents</h3>
            <p>Multi-agent LLM pipelines with tools, HITL, and tracing</p>
        </a>
        <a href="uses/inference.html" class="tile_item">
            <h3>Inference Service</h3>
            <p>Serve shared vLLM backends with Dragon queues and GPUs</p>
        </a>
        <a href="uses/batch.html" class="tile_item">
            <h3>Batch</h3>
            <p>Schedule parallel tasks and DAG-shaped dependencies with the Batch API</p>
        </a>
        <a href="uses/debugging.html" class="tile_item">
            <h3>Debugging</h3>
            <p>Use Dragon logging to inspect runtime behavior and application issues</p>
        </a>
        <a href="uses/serializables.html" class="tile_item">
            <h3>Python/C++ Integration</h3>
            <p>Write Code in C++ that shares Data and Objects with Python</p>
        </a>
    </div>

.. toctree::
    :maxdepth: 2
    :hidden:

    uses/data_processing.rst
    uses/orchestrate_procs.rst
    uses/orchestrate_mpi.rst
    uses/distributed_training.rst
    uses/ddict_checkpoint.rst
    uses/batch.rst
    uses/multinode.rst
    uses/jupyter.rst
    uses/gpus.rst
    uses/grafana.rst
    uses/workflow.rst
    uses/proxy.rst
    uses/agent.rst
    uses/inference.rst
    uses/debugging.rst