Configuration Reference

Fabric is configured via CLI flags, environment variables, or a TOML config file. Environment variables take precedence over config file values; CLI flags take precedence over both.

Resolution order (lowest to highest):

1. Built-in defaults
2. fabric.toml config file
3. Environment variables / .env file
4. CLI flags

Control Plane (fabric serve)

Server

Flag	Env Var	TOML Path	Default	Description
--host	HOST	server.host	127.0.0.1	HTTP bind address
--port	PORT	server.port	3001	HTTP port
--grpc-port	GRPC_PORT	server.grpc_port	3002	gRPC port
--database-url	DATABASE_URL	database.url	—	Postgres connection string (required)
—	DATABASE_DIRECT_URL	database.direct_url	—	Direct Postgres URL (bypasses pooler). Used for LISTEN/NOTIFY and migrations. Falls back to DATABASE_URL.
—	DATABASE_MIGRATION_URL	database.migration_url	—	Separate connection string for migrations (overrides DATABASE_DIRECT_URL)
—	DATABASE_POOL_MAX	database.pool_max	15	Maximum connections in the pool
—	DATABASE_POOL_MIN	database.pool_min	2	Minimum idle connections in the pool
—	RUST_LOG	—	info	Log level filter (tracing)

Runtime

Flag	Env Var	TOML Path	Default	Description
--fabric-env	FABRIC_ENV	runtime.env	development	Set to production to disable dev-header auth
—	FABRIC_MODE	runtime.mode	server	server or request-only (stateless/serverless)
—	FABRIC_DEV_MODE	runtime.dev_mode	—	1 or true to force dev-header auth
—	FABRIC_API_KEY	runtime.api_key	—	Admin API key for CLI subcommands
—	FABRIC_WORKFLOW_DIRS	runtime.workflow_dirs	—	CSV of extra workflow YAML directories
—	FABRIC_CONFIG	—	—	Explicit path to config file

Authentication (Native)

Flag	Env Var	TOML Path	Default	Description
—	OAUTH_SIGNING_SECRET	auth.oauth_signing_secret	—	Secret for signing native OAuth tokens
—	JWT_SECRET	auth.jwt_secret	—	JWT HS256 secret for token validation

Social Login (Optional)

Env Var	Description
GOOGLE_CLIENT_ID	Google OAuth client ID
GOOGLE_CLIENT_SECRET	Google OAuth client secret
GITHUB_CLIENT_ID	GitHub OAuth client ID
GITHUB_CLIENT_SECRET	GitHub OAuth client secret

Frontend URLs

Env Var	Description
PUBLIC_BASE_URL	Public API base URL (used in email links)
AUTH_SITE_URL	Frontend URL for auth redirects
AUTH_CALLBACK_REDIRECT_URL	OAuth callback redirect URL

AI Providers (Remote)

Flag	Env Var	TOML Path	Default	Description
--openai-api-key	OPENAI_API_KEY	providers.openai_api_key	—	OpenAI (text, image, embedding, streaming)
—	OPENAI_BASE_URL	—	https://api.openai.com/v1	Custom OpenAI-compatible endpoint
--anthropic-api-key	ANTHROPIC_API_KEY	providers.anthropic_api_key	—	Anthropic Claude (text)
—	GOOGLE_API_KEY	providers.google_api_key	—	Google Gemini (text, image, audio)
—	FAL_API_KEY	providers.fal_api_key	—	FAL (image, video generation)
—	ELEVENLABS_API_KEY	—	—	ElevenLabs text-to-speech
—	REPLICATE_API_TOKEN	—	—	Replicate model API
—	PEXELS_API_KEY	—	—	Pexels stock footage (b-roll)
—	FABRIC_LOCAL_ONLY	providers.local_only	false	Disable all remote providers
—	FABRIC_ROUTING_STRATEGY	providers.routing_strategy	cheapest_qualified	Default routing strategy (local_first, fastest, best_quality, balanced)

AI Providers (Local)

Flag	Env Var	TOML Path	Default	Description
--ollama-enabled	OLLAMA_ENABLED	providers.ollama_enabled	false	Enable Ollama LLM provider
--ollama-url	OLLAMA_URL	providers.ollama_url	http://localhost:11434	Ollama server URL
—	WHISPER_URL	providers.whisper_url	—	Whisper transcription server URL
—	WHISPER_MODEL	providers.whisper_model	—	Whisper model selection override
—	PIPER_URL	providers.piper_url	—	Piper TTS server URL
—	CHATTERBOX_URL	providers.chatterbox_url	—	Chatterbox voice cloning server URL
--comfyui-enabled	COMFYUI_ENABLED	providers.comfyui_enabled	false	Enable ComfyUI image generation
--comfyui-url	COMFYUI_URL	providers.comfyui_url	http://localhost:8188	ComfyUI server URL
--onnx-model-dir	ONNX_MODEL_DIR	providers.onnx_model_dir	—	ONNX model directory
--candle-enabled	CANDLE_ENABLED	providers.candle_enabled	true	Enable Candle ML inference
—	LLAMA_CPP_MODEL	providers.llama_cpp_model	—	Path to GGUF model file (requires llama-cpp feature)

Custom OpenAI-Compatible Endpoints

Configured via fabric.toml only (not env vars):

[[providers.custom_openai_endpoints]]
name = "localai"           # Provider ID for routing
url = "http://localhost:8080/v1"
api_key = "no-key"         # Use "no-key" if server doesn't require auth
model = "mistral-7b"       # Default model to advertise
tier = "basic"             # "basic", "balanced", or "premium"

Email & Notifications

Env Var	TOML Path	Default	Description
RESEND_API_KEY	email.resend_api_key	—	Resend API key for transactional email
FABRIC_EMAIL_FROM	email.from_address	noreply@fabric.local	Default sender email address
FABRIC_INVITE_TOKEN_KEY	—	auto-generated	64-char hex (32-byte BLAKE3 key) for invitation tokens
FABRIC_INVITE_REDIRECT_URL	—	—	URL to redirect to after invitation accept

Asset Storage

Env Var	TOML Path	Default	Description
ASSET_STORE_URL	assets.store_url	—	file:///path, s3://bucket, gs://bucket, memory://
ASSET_TTL_HOURS	assets.generated_asset_ttl_hours	72	Auto-expire workflow assets (0 = never)
ASSET_REAPER_INTERVAL_SECS	assets.reaper_poll_interval_secs	300	Cleanup poll interval in seconds

For S3/GCS backends, also set:

Env Var	Description
AWS_ACCESS_KEY_ID	S3/GCS access key
AWS_SECRET_ACCESS_KEY	S3/GCS secret key
AWS_REGION	AWS region or local for Supabase Storage
AWS_ENDPOINT	Custom S3 endpoint (for Supabase Storage, MinIO)
AWS_ALLOW_HTTP	true for non-TLS endpoints (local dev)

CORS

Env Var	TOML Path	Default	Description
CORS_ALLOWED_ORIGINS	cors.allowed_origins	—	CSV of allowed origins

Rate Limiting

Env Var	Default	Description
RATE_LIMIT_MAX_REQUESTS	200	Max requests per window
RATE_LIMIT_WINDOW_SECS	60	Window in seconds
RATE_LIMIT_BACKEND	memory	memory or redis

Workers (Sayiir)

Env Var	TOML Path	Default	Description
FABRIC_WORKERS	workers.count	1	Number of worker threads
FABRIC_WORKER_ID_PREFIX	workers.id_prefix	fabric-worker	Worker ID prefix
FABRIC_WORKER_POLL_INTERVAL	workers.poll_interval	2.0	Poll interval in seconds
FABRIC_WORKER_TAGS	workers.tags	—	CSV of worker capability tags

Scheduler

Env Var	TOML Path	Default	Description
CRON_POLL_INTERVAL_SECS	cron.poll_interval_secs	30	Cron job poll interval in seconds

Execution Plane (fabric executor)

Flag	Env Var	Default	Description
--worker-id	WORKER_ID	random UUID	Unique worker identity
--concurrency	WORKER_CONCURRENCY	4	Max steps to claim per poll cycle
--poll-interval	WORKER_POLL_INTERVAL	500	Poll interval in milliseconds
--lease-seconds	WORKER_LEASE_SECONDS	30	Lease TTL before auto-expiry
--max-retries	WORKER_MAX_RETRIES	3	Max retry attempts per step
--database-url	DATABASE_URL	—	Postgres connection string
—	RUST_LOG	info	Log level filter

Infrastructure

Docker Compose Profiles

Profile	Services
default	Postgres
ai	Postgres + Ollama + Whisper + ComfyUI
full	Everything

Just Commands

Command	Description
just infra-up	Full stack (Postgres + Ollama + Whisper + ComfyUI)
just infra-up-db	Postgres only
just infra-up-ai	Postgres + AI services
just infra-down	Stop everything
just infra-reset	Fresh start (delete volumes)
just infra-psql	Open psql shell
just infra-migrate	Re-apply migrations
just infra-smoke-test	Verify migrations on clean DB
just ollama-pull <model>	Pull an Ollama model
just ollama-list	List available Ollama models

Development

Command	Description
just dev	Start control plane + executor
just run	Start control plane only
just fmt	Format all code
just clippy	Lint with -D warnings
just test	Run all tests
just doc	Build documentation
just ci	All of the above

Example .env File

# === Required ===
DATABASE_URL=postgres://fabric:fabric@localhost:5432/fabric
RUST_LOG=info

# === Auth ===
OAUTH_SIGNING_SECRET=your-32-char-secret-here-generate-one

# === Local AI ===
OLLAMA_ENABLED=true
OLLAMA_URL=http://localhost:11434
WHISPER_URL=http://localhost:8080
FABRIC_ROUTING_STRATEGY=local_first

# === Remote AI (add keys as needed) ===
# OPENAI_API_KEY=sk-...
# ANTHROPIC_API_KEY=sk-ant-...
# GOOGLE_API_KEY=...
# FAL_API_KEY=...

# === Assets ===
ASSET_STORE_URL=file:///tmp/fabric/assets

# === CORS (for frontend dev) ===
# CORS_ALLOWED_ORIGINS=http://localhost:3000,http://localhost:5173

Tip

See the .env.example file in the repository root for the full annotated template with all available options.