Skip to content

Setup

Arnold reads its settings from three places. For Docker Compose, almost everything goes into one .env file.

Component Configuration In Docker Compose
Backend arnold-backend/arnold-server/config/local.yml; every secret is a ${ENV_VAR} .env (via env_file)
UI arnold-ui/config/*.yaml (node-config), plus a few env vars deploy/ui/development.yaml mounted as production.yaml
Runner Environment variables only .env (via env_file)

.env is created from deploy/.env.example and is git-ignored. Never commit real values.

Setup checklist

Step Required Page
Run the stack yes Quick start
Connect at least one forge yes GitHub App, GitLab
Configure a review model yes, for AI features LLM and harnesses
Choose where reviews run default LOCAL works Runners
Real sign-in before anyone else uses it Sign-in
Elasticsearch for insights optional Storage and analytics
Webhooks for instant updates optional Webhooks and sync
Hardening, TLS, scaling for shared deployments Production deployment

Choosing options

GitHub or GitLab? Both can be enabled together. GitHub uses a GitHub App, so Arnold can mint short-lived, repository-scoped tokens for runners and act as the user through user-to-server tokens. GitLab uses an OAuth application per user plus a token for the runner.

OpenCode or Copilot? Use OpenCode with your own OpenAI-compatible endpoint. Copilot is opt-in and has licensing caveats; see LLM and harnesses.

Which runner mode? LOCAL (workers next to the backend) is the simplest and the default. Use GITLAB or GITHUB to run reviews on CI capacity you already have; see Runners.

Applying changes

After editing configuration, recreate what reads it:

You changed Run
.env (backend or runner settings) docker compose up -d --force-recreate server runner
deploy/ui/*.yaml docker compose up -d --force-recreate ui
local.yml mvn -f arnold-backend/pom.xml -DskipTests package && docker compose up -d --build server

The backend logs every configuration validation error at boot (docker compose logs server).