Deploy and Host Flint AI Switch

Deploy a production-ready instance of Flint AI Switch (v0.25.0) on Railway, fully equipped with a managed PostgreSQL database, persistent volume storage, and pre-enabled Gateway and Collab (Mattermost) profiles.

About Hosting

Hosting Flint AI Switch on Railway provides an isolated, self-hosted environment for managing and orchestrating AI agents. By utilizing Railway's container orchestration and private networking, this deployment couples the ghcr.io/sandbox-quantum/standalone-compose:0.25.0 image with a dedicated PostgreSQL database (mattermost). Railway automatically handles container lifecycle management, automatic domain assignment, SSL termination, and internal networking.

Why Deploy

Centralized AI Orchestration: Unify external AI agents, team communication channels, and administrative controls within a single self-hosted switch.
Turnkey Setup: Provisions the database, volume mounts, network bridges, and automatic secret generation (AGENT_REGISTRATION_TOKEN) out of the box.
Persistent Cryptographic Identity: Guarantees that server signing keys and Matrix state are preserved across redeployments using Railway persistent volumes.
Integrated Collaboration: Pre-configured support for Mattermost (COMPOSE_PROFILES=collab) enables direct interaction between AI agents and team chat channels.

Common Use Cases

Multi-Agent Management: Securely route, register, and manage autonomous AI agents connecting via API on port 8000.
Self-Hosted AI Chat Hubs: Run team collaboration workflows powered by background AI services integrated into Mattermost.
Enterprise AI Gateways: Maintain an internal administrative control plane on port 3000 to monitor agent connections and system health.

Dependencies for Deployment

PostgreSQL Database: Required for storing application state and Mattermost collaboration data.
Persistent Volume (tuwuneldata): Storage volume mounted at /var/lib/tuwunel to retain cryptographic signing keys and state.
Network Ports:
3000: Web Gateway UI.
8000: Agent API endpoint.

⚙️ Environment Variables

Variable	Description	Default / Railway Value
`AGENT_REGISTRATION_TOKEN`	Security token for registering new AI agents	Auto-generated via `${{ secret(32) }}`
`DATABASE_URL`	PostgreSQL connection string	Linked internally to the `mattermost` database
`TUWUNEL_SERVER_NAME`	Public identity name of the server	`${{ RAILWAY_PUBLIC_DOMAIN }}`
`MATRIX_SERVER_NAME`	Server name for routing and federation	`${{ RAILWAY_PUBLIC_DOMAIN }}`
`SWITCH_BIND_ADDR`	Network listen address	`0.0.0.0`
`COMPOSE_PROFILES`	Active feature profiles	`gateway,collab`

⚠️ Critical Security & Persistence Notes

Data Persistence: The tuwuneldata volume must never be deleted or unmounted. It holds the private signing key. Losing this volume will irrecoverably orphan all server accounts and stored data.
Domain Stability: Do not modify TUWUNEL_SERVER_NAME or MATRIX_SERVER_NAME after the initial boot to avoid breaking the cryptographic identity of the deployment.

🛠️ Post-Deployment Setup

Access the Gateway: Navigate to the public URL generated for port 3000 in your browser to view the Gateway dashboard.
Connect Agents: Copy the auto-generated AGENT_REGISTRATION_TOKEN from your Railway Variables tab to authenticate external AI agents connecting on port 8000 (API).