Enterprise Deployment Guide

Deployment Architecture

Explore Rescile's enterprise components and layered architecture for scalable hybrid deployments.

Edit this page on GitHub

Deployment Architecture

Rescile’s enterprise architecture is designed around decoupled state and compute for massive horizontal scaling, with a standalone mode for smaller deployments.

Architecture Overview

Rescile Enterprise decouples state generation from compute to allow massive horizontal scaling. This separation enables independent scaling of build pipelines and query servers.

Components

Rescile Portal

The Edge Router and UI hosting layer. It proxies API requests and MCP sessions to internal fleets. It also handles prebuild validation and Server-Sent Events (SSE) streams, allowing users to safely test configurations before they are committed to Git.

Rescile Controller

The core stateless GraphQL and REST API server. It loads the rescile-bundle.tar.gz artifact from S3 into memory and resolves all graph queries, outputs, and impact analysis requests dynamically. Because it is stateless, it scales instantly.

Rescile MCP Server

A specialized fleet serving the Model Context Protocol (MCP) for AI Assistant integration over WebSockets (WSS). It uses the same artifact as the Controller but exposes tools specifically optimized for Large Language Models (LLMs).

Rescile Importer

A CI/CD builder node. It reads the raw Git repository (CSVs, JSONs, TOMLs), processes the graph phases (Assets → Models → Compliance → Output), and packages the final result into an immutable S3 artifact.

Git & S3

  • Git acts as the single source of truth for all human-readable and system-generated configurations.
  • S3 acts as the distribution registry for compiled, machine-readable artifacts.

rescile-ce

The Community Edition binary is a standalone tool that combines the Importer and Controller into a single execution context for local development, testing, and smaller deployments.

Layered Deployment Topology

Rescile’s enterprise architecture is organized in three layers:

1. Public / Edge Layer

Unified Access, UI Hosting & API Routing

  • Handles inbound connections via HTTPS and WSS.
  • Hosts the Rescile Portal Fleet, acting as the Edge Router, serving the Module UI, and running the Prebuild Engine.
  • End Users (Browser) and AI Assistants (LLM/Agent) connect to this layer.
  • Integrates with external Enterprise Self-Service and Execution portals.

2. Private / Internal Layer

Stateless Serving Fleets (Auto-Scaling Groups)

  • Rescile Controller Fleet: Serves the GraphQL & REST API dynamically across the cluster.
  • Rescile MCP Server Fleet: Handles LLM Context Protocol interactions for native AI.
  • Both fleets pull immutable artifacts from shared storage and serve queries entirely from memory.

3. Build & Shared Storage Layer

Immutable Artifact Pipeline & Foundational Configuration

  • Git Repository: Stores all configuration, Asset CSVs, and TOML blueprints.
  • Rescile Importer: Runs as a CI/CD Artifact Builder Node. It pulls source from Git, builds the graph, and publishes the artifact.
  • Shared Storage (S3 Registry): Stores the generated rescile-bundle.tar.gz artifacts.

Standalone Mode (rescile-ce)

For smaller deployments, individual laptops, or air-gapped environments, you can run the entire pipeline in standalone mode using the rescile-ce binary.

This collapses the Importer and Controller into a single execution context, monitoring local directories for changes instead of polling S3.