llm-gateway/docs/configuration.md

Configuration

LLM Gateway is configured through environment variables and runtime settings.

The public defaults are intentionally local. Real deployment values should be supplied by your shell, process manager, container platform, or secret manager.

Core Variables

Variable	Default	Purpose
PORT	3100	HTTP port
HOST	localhost	HTTP bind host
LOG_LEVEL	info	Logging level
DATABASE_URL	unset	PostgreSQL connection string
AUTO_SPAWN_BRIDGES	unset	Start detected bridges on boot when set to 1
DISCOVERY_INTERVAL_MS	300000	Subscription discovery interval
WATCHDOG_ENABLED	unset	Enable bridge watchdog when set to 1

Bridge Variables

Variable	Purpose
CLAUDE_BRIDGE_URL	Claude Code bridge URL
CODEX_BRIDGE_URL	Codex bridge URL
CHATGPT_BRIDGE_URL	ChatGPT bridge URL
COPILOT_BRIDGE_URL	GitHub Copilot bridge URL
M365_COPILOT_BRIDGE_URL	Microsoft 365 Copilot bridge URL
GEMINI_BRIDGE_URL	Gemini bridge URL
AIDER_BRIDGE_URL	Aider bridge URL

Bridge URLs usually point to localhost ports.

Safe Local Example

export PORT=3100
export HOST=localhost
export LOG_LEVEL=info
export AUTO_SPAWN_BRIDGES=1

Use your runtime environment for values that grant access.

Database

Database-backed features use DATABASE_URL. Keep credentials outside git.

Recommended pattern:

export DATABASE_URL="$(security-tool read gateway database-url)"

Use whatever secret manager or platform feature is appropriate for your environment.

Gateway Access

Clients should send:

Authorization: Bearer <key>

The exact access mechanism can be adapted per deployment. Avoid hardcoding access values in source code, docs, scripts, tests, or screenshots.

Production Notes

Before production use:

• set explicit CORS origins
• configure TLS at the edge or server
• use a real database if persistence is required
• define backup and retention policy
• decide which providers may receive sensitive data
• disable unused bridges
• run a secret scan before deployment
• verify logs do not expose prompts or account data