Architecture Overview

Computer Agents is built on Google Cloud Platform with enterprise-grade infrastructure designed for reliability, security, and scalability.

Our infrastructure is designed to scale from individual developers to enterprise workloads while maintaining consistent performance and security.

High-Level Architecture


                          Your Application
                                 │
                                 ▼
┌────────────────────────────────────────────────────────────────┐
│                  api.computer-agents.com                        │
│              Global HTTPS Load Balancer                         │
│              • SSL termination                                  │
│              • DDoS protection                                  │
│              • Global anycast                                   │
└────────────────────────────────────────────────────────────────┘
                                 │
                                 ▼
┌────────────────────────────────────────────────────────────────┐
│              Auto-Scaling Compute Instances                     │
│                                                                 │
│    ┌──────────────┐  ┌──────────────┐  ┌──────────────┐       │
│    │  Instance 1  │  │  Instance 2  │  │  Instance N  │       │
│    │              │  │              │  │              │       │
│    │ API Server   │  │ API Server   │  │ API Server   │       │
│    │ + Executor   │  │ + Executor   │  │ + Executor   │       │
│    └──────────────┘  └──────────────┘  └──────────────┘       │
│                                                                 │
│              Scales 1-N instances based on demand               │
└────────────────────────────────────────────────────────────────┘
                    │              │              │
         ┌──────────┴──────────────┼──────────────┴──────────┐
         │                         │                         │
         ▼                         ▼                         ▼
┌─────────────────┐    ┌─────────────────┐    ┌─────────────────┐
│   PostgreSQL    │    │  Cloud Storage  │    │   Container     │
│   Database      │    │  (Workspaces)   │    │   Registry      │
│                 │    │                 │    │                 │
│ Shared state    │    │ File storage    │    │ Execution       │
│ across all      │    │ for all         │    │ images          │
│ instances       │    │ environments    │    │                 │
└─────────────────┘    └─────────────────┘    └─────────────────┘

Key Architectural Principles

1. Stateless API Servers

Every API server is stateless and identical. Your request can be handled by any instance, enabling:

Horizontal scaling - Add instances as demand grows
Zero-downtime deployments - Rolling updates without service interruption
High availability - Automatic failover if an instance fails

2. Shared Data Layer

All instances connect to the same PostgreSQL database and Cloud Storage:

Consistency - Your data is always up-to-date across all instances
Durability - Automatic backups with point-in-time recovery
No data loss - Even if instances restart, your work is preserved

3. Isolated Execution

Each task runs in an isolated Docker container:

Security - Tasks cannot interfere with each other
Clean state - Fresh environment for each execution
Resource limits - CPU and memory limits per container

Architecture Sections

Google Cloud components: load balancing, auto-scaling, database, storage

Infrastructure

Authentication, encryption, isolation, and compliance

Security

Auto-scaling, performance characteristics, and limits

Scalability

Platform Highlights

Feature	Description
99.9% Availability	Global load balancing with health checks
Auto-Scaling	1 to N instances based on demand
Multi-Region	Primary in US with global edge caching
Encryption	TLS 1.3 in transit, AES-256 at rest
Isolated Execution	Docker containers per task
Automatic Backups	Daily database backups, 7-day retention

Request Flow

When you make an API request:

DNS Resolution - api.computer-agents.com resolves to our global load balancer
SSL Termination - HTTPS is terminated at the edge with managed certificates
Load Balancing - Request is routed to a healthy instance
Authentication - API key is validated against the database
Execution - Task runs in an isolated Docker container
Streaming - Real-time events are streamed back via SSE
Persistence - Results are saved to database and cloud storage

Technology Stack

Layer	Technology
Edge	Google Cloud Load Balancing
Compute	GCE Managed Instance Groups
Database	Cloud SQL PostgreSQL 15
Storage	Google Cloud Storage
Containers	Docker with warm pooling
API	Node.js / Express
Execution	OpenAI Codex SDK

Service Level Objectives

Metric	Target
API Availability	99.9%
Request Latency (P50)	< 200ms
Request Latency (P99)	< 2s
Cold Start Time	< 10s
Warm Execution	< 500ms overhead

Learn More

Infrastructure - Deep dive into GCP components
Security - Security model and compliance
Scalability - Performance and scaling behavior