This is the umbrella repository for the ePIC streaming workflow testbed.
This document contains mandatory critical thinking requirements, coding rules, and architecture understanding that must be followed to avoid common failure patterns.
- Release Notes - What's new in each version
- Installation Guide - Complete setup instructions
- Architecture Overview - System design and components
- Running the Testbed - Starting, stopping, and monitoring services
- Workflow Orchestration - Running and managing workflows
- Monitor Integration - Web interface and API usage
- MCP Integration - Model Context Protocol for LLM interaction
- SSE Real-Time Streaming - Remote workflow event monitoring via HTTPS
- Production Deployment - Complete Apache production deployment guide
This repository includes .mcp.json which automatically configures Claude Code to connect to the testbed's MCP service. When you open this project in Claude Code, the swf-testbed MCP server is available for natural language queries about system state, agents, workflows, and logs.
SSL Setup (required for BNL/SDCC servers): Add to your ~/.bashrc:
export NODE_EXTRA_CA_CERTS=/etc/pki/tls/certs/ca-bundle.crtTo manually add the MCP server:
claude mcp add --transport http swf-testbed https://pandaserver02.sdcc.bnl.gov/swf-monitor/mcp/- Development Guide - Contributing and development workflow
This modular structure is designed to help AI assistants:
- Find specific information quickly without parsing large documents
- Understand focused topics without information overload
- Provide accurate guidance based on targeted documentation sections
- Update documentation incrementally without affecting unrelated sections
Each document focuses on a single concern and cross-references related topics.