How skysingh04/dreamops works
Project type and purpose
• From the file tree (“backend/”, “frontend/”, “infrastructure/”) and README, we see this is a full-stack SaaS platform (“DreamOps” / “oncall-agent”) combining:
• A Python FastAPI backend AI-powered incident response agent
• A Next.js + React frontend dashboard
• Terraform & EKS/ECS infrastructure for AWS deployment
• External MCP servers (GitHub-MCP, Grafana-MCP)
File structure
• Top-level:
• backend/ – Python AI agent with FastAPI, CLI, examples, tests, Dockerfile
• frontend/ – Next.js app router, React components, Drizzle ORM, API routes, Dockerfile
• infrastructure/ – Terraform modules for ECS, CloudFront, EKS test cluster
• github-mcp-server/ & mcp-grafana/ – Go-based Model Context Protocol servers
• Patterns & config:
• Backend uses src/oncall_agent package, MVC-style routers under api/routers
• Frontend uses app/ and components/, plus client-side calls to backend via API routes and WebSocket
• CI/CD pipelines in .github/workflows for backend, frontend, security
• Docker-Compose for local dev and multiple microservices
README clues about architecture
• MCP integration block (Kubernetes, GitHub, Grafana, Notion)
• AWS infrastructure: VPC, ECS Fargate for backend, S3+CloudFront for frontend, ALB, ECR, CloudWatch, Secrets Manager
• Real-time flow via REST API, WebSockets, frontend dashboard
Main components
a. Alert Source layer
• PagerDuty SNS webhooks
• (Test scripts, “fuck_kubernetes.sh”)
b. Oncall AI Agent (Backend)
• FastAPI REST API & WebSocket server
• Core agent logic (oncall_agent.agent, remediation_pipeline)
• MCP plugin loader (mcp_integrations/*)
c. External AI Service
• Claude AI API
d. MCP Integration Services
• Kubernetes cluster (EKS or local kind)
• GitHub-MCP server
• Grafana-MCP server
• Optional: Notion, PagerDuty, Slack
e. Frontend SaaS UI
• Next.js app with API routes and real-time updates
• Drizzle ORM + SQLite/Postgres for session & dashboard data
f. AWS Infrastructure
• ECS Fargate backend behind ALB
• S3 + CloudFront hosting frontend
• EKS cluster for testing & debug integration
• CloudWatch + SNS → PagerDuty
• Terraform modules (networking, monitoring, backend, frontend)
Architectural patterns & design principles
• Microservices / plugin architecture for MCP integrations
• Async event-driven processing (async/await in Python, background tasks)
• Clear separation:
• API layer (FastAPI)
• Core logic & strategies
• Integrations as interchangeable plugins
• Frontend SSR/CSR hybrid (Next.js / App Router) with real-time WebSocket feeds
• Infrastructure as code (Terraform) with separate modules for networking, compute, monitoring
Technologies & frameworks
• Backend: Python 3.11, FastAPI, Uvicorn, Pydantic, pytest, mypy, ruff
• Frontend: Next.js 13+, React, TypeScript, Drizzle ORM, Stripe, WebSocket
• External MCP: Go, Gorilla Mux, Go-GitHub-SDK, EasyJSON
• Infra: Terraform, AWS (ECS Fargate, EKS, CloudFront, S3, ALB, CloudWatch, SNS, Secrets Manager)
• CI/CD: GitHub Actions, Docker, Docker-Compose, Amplify
Be detailed but not overwhelming
• Every essential moving piece: alert ingress, AI call, integration calls, front-end updates, infra components
• Avoid unnecessary low-level file paths—focus on logical components