Infrastructure & Availability

Dwizi is built on a high-availability, distributed architecture designed to ensure your tools are always ready to run.

Architecture

The Gateway

The Gateway is the entry point for all traffic. It handles:

• Authentication: Verifying tokens and permissions.
• Routing: Directing requests to the correct organization and tool.
• Rate Limiting: Protecting the system and your budget from abuse.

The Gateway is stateless and horizontally scalable. We run multiple instances behind a global load balancer to handle traffic spikes.

The Orchestrator

The Orchestrator manages the lifecycle of the secure containers that execute your code.

• Isolation: Each run happens in a fresh, isolated environment.
• Cleanup: Containers are destroyed immediately after execution to prevent data leakage.
• Pooling: We maintain a pool of "warm" containers to minimize cold-start latency, ensuring your tools respond in milliseconds, not seconds.

Reliability Features

Zero-Downtime Deployments

When you update a tool, the new version is deployed atomically. There is no period where the tool is unavailable.

Timeouts & Limits

To prevent runaway processes, all tools have configurable timeouts (default 60s). This ensures that a buggy script doesn't consume all your resources or hang your AI agent.

Logging & Observability

Audit events are retained by default, and output/log retention is configurable.

• Audit Events: Access decisions, policy changes, and execution metadata.
• Stdout/Stderr (Optional): Captured in real-time when retention is enabled.
• Duration: tracked to help you optimize performance.

We use this data to provide the real-time Runs view in your dashboard, giving you complete visibility into your infrastructure.