nomad3/techopsmind

OpsMind

Company-agnostic SRE operations platform with AI-powered monitoring, plugin architecture for multi-company support, and comprehensive infrastructure observability via Model Context Protocol (MCP).

Architecture: Plugin-based system supporting multiple companies with isolated configurations Current Plugins: Integral FX trading (33 tools, 1,545 servers, 6 datacenters)

🚀 Quick Start

# Clone repository
git clone https://github.com/opsmind/opsmind.git
cd opsmind

# Install dependencies
poetry install

# Configure for your company
cp .env.example .env
# Edit .env with your credentials

# Start OpsMind server
poetry run opsmind-server

Connect to Claude Code

Add to ~/.config/claude-code/mcp_config.json:

{
  "mcpServers": {
    "opsmind": {
      "command": "poetry",
      "args": ["run", "opsmind-server"],
      "cwd": "/path/to/opsmind"
    }
  }
}

Usage Example:

You: "Check health of ppfxidb1"
Assistant: [Uses check_server_health from Integral plugin]

You: "Analyze alerts in Singapore region"
Assistant: [Uses analyze_alerts with AI pattern detection]

🏗️ Architecture

Plugin-Based Design

OpsMind uses a plugin architecture to support multiple companies:

OpsMind Platform
├── Core Framework (company-agnostic)
│   ├── MCP server
│   ├── Plugin system
│   ├── Configuration management
│   └── Base classes
│
└── Company Plugins (company-specific)
    ├── Integral Plugin (33 tools)
    ├── Your Company Plugin
    └── Another Company Plugin

Each company gets:

• Custom hostname parsing
• Custom alert patterns
• Custom monitoring thresholds
• Custom MCP tools
• Isolated configuration

📦 Current Plugins

Integral Plugin (Built-in)

Infrastructure: 1,545 servers | 6 datacenters (LDN, NYC, SG, TYO, UAT, DR) Tools: 33 MCP tools organized in 9 categories

Features:

• AI-powered alert analysis (85% noise detection)
• Vector store (10,874 entries across 6 collections)
• Auto-ingestion (hourly Slack + daily Confluence)
• FX trading infrastructure monitoring
• Hostname parsing (ppfxidb1, spfxiclob5 conventions)

🔧 Creating a Plugin for Your Company

Quick Start (4-8 hours)

1. Copy Template:

cp -r plugins/template plugins/your_company
cd plugins/your_company

2. Edit plugin.yaml:

name: "your_company"
version: "1.0.0"
author: "Your Team"
description: "Your company infrastructure monitoring"
class_name: "YourCompanyPlugin"

3. Customize Hostname Parsing: Edit hostname_parser.py to match your naming conventions:

def parse(self, hostname: str) -> Dict[str, Any]:
    # Your logic: web-prod-us-01 → {environment: "prod", location: "us-east", ...}
    pass

4. Configure Thresholds: Edit monitoring_config.py with your thresholds:

def get_thresholds(self) -> Dict[str, float]:
    return {
        "cpu_warning": 80,  # Your thresholds
        "memory_critical": 95,
    }

5. Enable Plugin:

# In config/your_company.yaml
plugins:
  enabled:
    - your_company

# Start server
poetry run opsmind-server

See: PLUGIN_DEVELOPMENT_GUIDE.md (coming in Batch 5) for full details

🛠️ Components

Core Framework (Company-Agnostic)

Plugin System (core/plugins/):

• Plugin interface (SREPlugin)
• Plugin discovery and loading
• Configuration validation

Configuration (core/config/):

• Pydantic models with validation
• Environment variable substitution
• Multi-environment support

Security (Planned):

• Input validation
• API key authentication
• Audit logging

OpsMind Package

Current State: Contains all implementation (30 modules)

• MCP server (4,346 lines, all 33 tools)
• Monitoring modules (5 types)
• Vector store (ChromaDB, 10,874 entries)
• SSH multi-hop
• Alert intelligence
• Resource monitoring

Future State: Will be split into:

• Generic base classes in core/
• Integral-specific tools in plugins/integral/tools/

📚 Documentation

For Users

• README.md - This file (getting started)
• OPSMIND_REFACTOR_STATUS.md - Current refactor status
• docs/RUNBOOK.md - Operations procedures

For Developers

• REFACTORING_PLAN.md - Complete refactoring plan
• REFACTOR_IMPLEMENTATION_GUIDE.md - Step-by-step implementation
• REFACTOR_FAST_TRACK.md - 1-week fast-track plan
• CLAUDE.md - AI assistant guide (needs updating)

For Plugin Developers

• plugins/integral/ - Reference implementation
• plugins/test_plugin/ - Minimal example
• PLUGIN_DEVELOPMENT_GUIDE.md - Coming soon

🔒 Security

Current (Pragmatic):

• ✅ Input validation in progress
• ✅ Secrets via environment variables
• ✅ Proprietary license

Planned:

• API key authentication (optional)
• Enhanced input sanitization
• Audit logging

Not Planned (internal use):

• ❌ RBAC (not needed for internal)
• ❌ Vault integration (env vars sufficient)
• ❌ Encryption at rest (optional)

📝 Version & Status

Current: 1.0.0 (OpsMind) Previous: 0.7.1 (Integral SRE Assistant) Status: ✅ Phase 1 Complete (Plugin Architecture Proven)

Branch: refactor/company-agnostic Commits: 4 checkpoint commits Time Invested: ~12-16 hours Remaining (if continuing): ~10-14 hours

What Works:

• ✅ OpsMind server starts successfully
• ✅ Plugin system loads Integral plugin
• ✅ All 33 tools available
• ✅ Hostname parsing, alerts, thresholds extracted
• ✅ Memory leaks fixed
• ✅ Resource monitoring active

What's Next:

• Test current state thoroughly
• Optionally continue with Batches 4-5
• Or create second company plugin to validate

🤝 Built With

• Model Context Protocol - AI tool integration
• Claude Code - AI-powered development
• Google Gemini - AI intelligence
• Poetry - Dependency management
• Pydantic - Configuration validation
• ChromaDB - Vector store
• Prometheus - Metrics collection

📄 License

Proprietary - Internal use only

For detailed documentation, see documentation files and OPSMIND_REFACTOR_STATUS.md 📖