Local Deployment

Local Mode is the primary operating environment for Red Station. It allows you to run the entire Control Plane, Agent Runtime, and Vector Database on your local machine, ensuring complete data sovereignty and zero latency.

Architecture

In Local Mode, Red Station functionality is entirely contained within your environment:

• Control Plane: Runs as a local background service managed by the CLI.
• Agents: Execute in your local shell context, inheriting your environment variables and file permissions (subject to policy).
• LLMs: Connect to:
  • Local Models: Ollama running on localhost (default).
  • External APIs: OpenAI, Anthropic, or Azure via direct HTTPS. No proxy server is used.
• Storage: SQLite database located at ~/.pilot/pilot.db.

Installation

Prerequisites

• Red Station CLI installed
• (Optional) Ollama for local inference

Setup

To initialize your Global Environment (Registry, Secrets, Base Policies):

pilot init --global

This creates the ~/.pilot directory:

~/.pilot/
├── config.yaml          # Global configuration
├── pilot.db            # State database (SQLite)
├── secrets.yaml        # Encrypted secrets
├── agents/             # Custom global agents
├── workflows/          # Custom global workflows
└── tools/              # Global tool definitions

Project Setup

To enable Red Station for a specific repository:

cd /path/to/my-repo
pilot init --project

This creates a .pilot directory in the project root. Project configurations override global defaults, allowing you to have different policies or agents for different projects.

Verification

Run the status command to verify all systems are green:

pilot status

Expected output:

[OK] Control Plane Engine
[OK] Agent Runtime
[OK] Policy Engine
[OK] Vector Store