Installation

This page covers everything the quickstart glossed over. The single source of truth for configuration is .env.example in the repository — every variable is documented inline there.

What docker compose up -d starts

Service	Purpose
arta-api	The ARTA engine — FastAPI backend, agents, dispatch pipeline
arta-frontend	The dashboard — Next.js 14
postgres	Primary database
neo4j	Traceability graph (requirement → test → result)
redis	Queues and pub/sub between pipeline stages
chromadb	Vector store for retrieval during generation
zap	OWASP ZAP daemon for security scans

Default host ports: UI on 38088, API on 38087 (the API container’s internal port stays 8000; override the host mapping with ARTA_API_HOST_PORT if 38087 clashes on your machine).

Configure the LLM provider

ARTA defaults to a local model — no key required:

ARTA_LLM_PROVIDER=ollama
ARTA_LLM_MODEL=qwen2.5:32b
OLLAMA_BASE_URL=http://localhost:11434

Supported providers are ollama, anthropic, google_gemini, and openai — set the matching API key variable (for example ANTHROPIC_API_KEY) when using a cloud provider. Providers can also be configured per project in the UI.

ARTA explicitly refuses to silently fall back to a cloud endpoint when a provider is misconfigured — a misconfiguration is an error, not a quiet data-residency violation.

Secrets and first login

• DATABASE_URL is the primary secret (it contains the Postgres password); keep it in sync with POSTGRES_PASSWORD when you rotate.
• ARTA_API_KEY protects admin endpoints — change it from the placeholder.
• Bootstrap the first admin account with ARTA_BOOTSTRAP_ADMIN_EMAIL + ARTA_BOOTSTRAP_ADMIN_PASSWORD (created only when the users table is empty; unset after first login), or use ARTA_DEMO_MODE=1 for a no-database demo (admin@arta.dev / demo1234 — never in production).
• Never commit .env to source control.

Telemetry

ARTA sends an anonymous, bucketed usage ping by default. Every field is documented in docs/TELEMETRY.md, free-form strings are structurally impossible, and:

ARTA_TELEMETRY=0

produces zero network calls — there is a unit test that proves it. This makes air-gapped deployment a supported configuration, not a workaround.

Production self-hosting

For sizing, deployment profiles, and security hardening, read the self-hosting infrastructure spec and the runbook. The repo also ships docker-compose.prod.yml as a production-oriented overlay.