Architecture

D.A.M.I uses a multi-agent architecture where specialized AI agents collaborate to handle complex DevOps tasks.

System Overview

Core Components

Orchestrator

The central brain of D.A.M.I. Built with FastAPI and deployed on Cloud Run.

Receives user requests and webhook events
Routes to appropriate AI agents using Gemini-powered classification
Manages tool execution with risk-based approval gates
Provides real-time SSE streaming for dashboard updates

Agent System

D.A.M.I has 29 specialized agents organized in tiers:

Tier	Agents	Purpose
Core	SRE, Architect, Code Review, Release Manager	Essential DevOps functions
Security	SecOps, Compliance, Audit	Security scanning and compliance
Operations	Monitoring, FinOps, Capacity Planner	Observability and cost optimization
Advanced	War Room, Chaos Engineering, ML/AI Ops	Complex scenarios and testing

MCP Protocol (Model Context Protocol)

D.A.M.I uses the MCP protocol to communicate with external tools:

Standardized tool discovery and execution
Risk classification (READ / WRITE / DESTRUCTIVE)
Approval gates for dangerous operations
Circuit breakers for fault tolerance

Webhook Engine

Real-time event ingestion from 5+ sources:

GitHub — Push, PR, workflow events (HMAC-SHA256 verified)
Jenkins — Build status notifications
Jira — Issue creation and updates
PagerDuty — Incident alerts with urgency routing
Datadog — Metric alerts and anomalies

Security Model

Firebase Authentication with JWT verification
RBAC with 4 roles: Owner, Admin, Operator, Viewer
Multi-tenancy with complete data isolation
AES-256-GCM encryption for all stored secrets
HMAC signature verification for all webhooks

Data Flow

Event arrives (webhook, user chat, API call)
Auth middleware verifies identity and permissions
Orchestrator classifies the request
Agent Router selects the best agent(s)
Agent reasons about the task using Gemini
Tool Executor runs actions via MCP (with approval if needed)
Results streamed back to user via SSE

Getting Started Overview