Skip to main content

IoWarp-Demos

Explore real-world demonstrations of AI4HPC tools and workflows.

06/2025 - Workflow Orchestration​

Full workflow showcase from scheduling to deployment, data collection, and analysis using IOWarp MCPs with Claude Code.

Phase 1: Environment Discovery​

  • "Explore the filesystem to discover available scientific data files"
  • "Use the slurm MCP to check the current cluster status and available resources of the cluster"
  • "Use the hardware MCP to query the node hardware specifications"
  • "Use the lmod MCP to list all available software modules"
  • "Create initial environment assessment in environment_discovery.md"

Phase 2: Literature Context Gathering​

  • "Search ArXiv for 'Gray-Scott model' papers from 2024-2025"
  • "Filter to top 5 most recent papers"
  • "Extract abstracts and key methodologies"
  • "Write them down on literature_summary.md"
  • "Download the found papers to a 'literature' folder"

Phase 3: Multi-Format Data Processing​

  • "Load and analyze the data/nanoparticles.bp5 ADIOS dataset with the Adios MCP"
  • "Generate a report of this dataset"
  • "Load the data/sensor_data.csv using the Pandas MCP"
  • "Generate a summary of this dataset"
  • "Create a data_inventory_report.md file with the report from both datasets"
  • "Plot the trajectory of a single atom over time..."

Phase 4: Adaptive Resource Management​

  • "Explore optimal resource allocation"
  • "Successfully secure resources via Slurm"
  • "Generate a hostfile 'jarvis.hostfile' with the selected nodes"
  • "Document resource management decisions in resource_allocation.md"

Phase 5: Computational Workflow Execution​

  • "Initialize Jarvis pipeline"
  • "Configure it for the available node allocation with the jarvis.hostfile"
  • "Build the jarvis environment"
  • "Append the Gray Scott application"
  • "Run the job"
  • "Document the results of the phase jarvis_summary.md"

Phase 6: Insight Synthesis & Reporting​

  • "Correlate literature findings with collected data"
  • "Assess data quality across all sources"
  • "Identify key patterns and recommendations"
  • "Generate comprehensive summary in scientific_workflow_summary.md"

07/2025 - Agent Observability​

Showcase of IOWarp's reproducibility visualizer for agentic workflows, enabling full provenance tracking and replay.

Key Prompts​

  • "Launch multiple subagents to explore the code base and generate a folder wiki with full knowledge of the design and architecture of the code with technical details"

Context​

This demo demonstrates IOWarp's comprehensive tracing and reproducibility capabilities by:

  • Multi-Agent Exploration: Deploying multiple subagents to systematically analyze the IOWarp runtime (Chimaera) codebase
  • Architecture Documentation: Generating detailed technical documentation covering the system's design and architecture
  • Provenance Tracking: Recording all interactions and analysis steps for full reproducibility
  • Knowledge Synthesis: Creating a comprehensive wiki documenting the modular runtime system

The demo showcases how IOWarp can trace complex multi-agent workflows and provide complete provenance for scientific computing tasks, enabling researchers to reproduce and understand complex system analyses.

08/2025 - Adios Data & I/O​

Showcase of IOWarp Adios MCP providing full analysis of a BP5 file generated by LAMMPS, demonstrating natural language queries on scientific data.

Example Queries​

  • "What can you tell me about the simulation based on the dataset?"
  • "What was the initial temperature of the system?"
  • "How many gold atoms are in the simulation?"
  • "What are the dimensions of the simulation box?"
  • "What is the total duration of the simulation in picoseconds?"
  • "What is the crystal structure of the gold at the beginning of the simulation?"
  • "Plot the positions of the atoms at the final timestep?"
  • "Plot the trajectory of a single atom over time. The atom of choice should a parameter to the script. The output of the script should a PNG image with the results. Run the script for any single atom"

Context​

This demo demonstrates IOWarp's Adios MCP capabilities for analyzing scientific simulation data by:

  • Scientific Data Interpretation: Analyzing BP5 files generated by LAMMPS molecular dynamics simulations
  • Natural Language Queries: Enabling researchers to ask complex questions about simulation data in plain English
  • Automated Analysis: Extracting key simulation parameters like temperature, atom count, box dimensions, and duration
  • Visualization Generation: Creating plots and visualizations of atomic positions and trajectories
  • Interactive Exploration: Providing an intuitive interface for exploring complex scientific datasets

The demo showcases how IOWarp's Adios MCP can bridge the gap between complex scientific data formats and accessible analysis tools, making it easier for researchers to understand and visualize molecular dynamics simulations of gold melting processes.