.env.example

# OwnPilot Configuration
# Copy this file to .env and fill in your values
#
# API keys (OpenAI, Anthropic, etc.) and channel tokens (Telegram)
# are NOT configured here — use the Config Center UI after setup.

# ===========================================
# Server
# ===========================================
PORT=8080
UI_PORT=8199
HOST=127.0.0.1
NODE_ENV=development

# Additional CORS origins (auto-includes http://localhost:UI_PORT)
# CORS_ORIGINS=https://my-domain.com

# Max request body size in bytes (default: 1MB)
# BODY_SIZE_LIMIT=1048576

# ===========================================
# Database (PostgreSQL)
# ===========================================
# These defaults match docker-compose.yml so both paths work out of the box.
#
# Option 1: Full connection URL (overrides individual settings)
# DATABASE_URL=postgresql://ownpilot:ownpilot_secret@localhost:25432/ownpilot

# Option 2: Individual settings
# For "From Source" dev: 127.0.0.1 + exposed port 25432
# For Docker: docker-compose.yml hardcodes host=postgres, port=5432 internally
# Note: use 127.0.0.1 (not "localhost") — on Windows localhost may resolve to ::1
# while Docker publishes the port on IPv4 only, causing silent connection failures.
POSTGRES_HOST=127.0.0.1
POSTGRES_PORT=25432
POSTGRES_USER=ownpilot
POSTGRES_PASSWORD=ownpilot_secret
POSTGRES_DB=ownpilot
# POSTGRES_POOL_SIZE=10
# DB_VERBOSE=false

# ===========================================
# Authentication (DB is primary, env is fallback)
# ===========================================
# Default: api-key. Set AUTH_TYPE=none only for local development.
# AUTH_TYPE=api-key
# API_KEYS=your-secret-key-here
# JWT_SECRET=

# ===========================================
# Rate Limiting (DB is primary, env is fallback)
# ===========================================
# RATE_LIMIT_DISABLED=false
# RATE_LIMIT_WINDOW_MS=60000
# RATE_LIMIT_MAX=500

# ===========================================
# MQTT / Edge Devices (optional)
# ===========================================
# Required only when starting the bundled broker with `--profile mqtt`.
# Generate a strong password, then point the gateway at the Compose service:
# MQTT_BROKER_URL=mqtt://mosquitto:1883
# MQTT_USERNAME=ownpilot
# MQTT_PASSWORD=
# MQTT_PORT=1883
# MQTT_WS_PORT=9001

# ===========================================
# Security & Encryption
# ===========================================
# Required in production for OAuth token encryption (32 bytes hex)
# ENCRYPTION_KEY=

# Data-at-rest encryption key for stored service configs (Config Center
# API keys etc.). Any string works — the AES key is derived via SHA-256.
# If unset, a key is auto-generated at <data>/credentials/data-encryption.key
# on first run. Losing the key (or changing this value) makes existing
# encrypted configs unreadable — they must be re-entered.
# OWNPILOT_ENCRYPTION_KEY=

# Escape hatch: write service configs as plaintext (reads still decrypt)
# OWNPILOT_DISABLE_DATA_ENCRYPTION=0

# Admin API key for debug endpoints (production only)
# ADMIN_API_KEY=

# ===========================================
# Data Storage
# ===========================================
# Override the platform-specific data directory
# Default: ~/.local/share/ownpilot (Linux), ~/Library/Application Support/ownpilot (macOS),
#          %LOCALAPPDATA%/ownpilot (Windows)
# OWNPILOT_DATA_DIR=

# ===========================================
# Logging
# ===========================================
LOG_LEVEL=info

# ===========================================
# Debug (development only)
# ===========================================
# DEBUG_AI_REQUESTS=false
# DEBUG_AGENT=false
# DEBUG_LLM=false
# DEBUG_RAW_RESPONSE=false

# ===========================================
# SECURITY ESCAPE HATCHES
# ===========================================
# Every flag in this section is OFF by default and DISABLES a protection when
# enabled. They exist for legitimate local-development and self-hosting cases,
# but each one widens the attack surface. Enable deliberately, one at a time,
# and prefer narrowing the need instead.
#
# NOTE ON VALUES: most flags below activate on the exact string "true".
# OWNPILOT_ALLOW_LOCAL_EXEC is the exception — it activates on "1" and ignores
# "true". Check the comment on each line.

# --- Network egress (SSRF guards) ---
# Allow LLM base URLs that resolve to private/loopback addresses.
# Needed for local model servers (Ollama, LM Studio, vLLM). Bypasses the SSRF
# guard for the provider request path only.
# OWNPILOT_ALLOW_LOCAL_LLM_URL=true
#
# Same, for the embedding service base URL.
# OWNPILOT_ALLOW_LOCAL_EMBEDDING_URL=true

# --- Code execution ---
# Allow agent code execution to run on the HOST instead of the Docker sandbox.
# Agent-generated code then runs with the gateway process's own privileges.
# Activates on "1" — NOT on "true".
# OWNPILOT_ALLOW_LOCAL_EXEC=1
#
# Drop the hardening applied to the Docker sandbox container.
# DOCKER_SANDBOX_RELAXED_SECURITY=true
#
# Require human approval before each local execution. Default false.
# Turning this ON is a hardening step, not an escape hatch.
# LOCAL_EXEC_REQUIRE_APPROVAL=true
#
# Languages permitted for local execution. Default: javascript,python,shell
# LOCAL_EXEC_LANGUAGES=javascript,python,shell
# LOCAL_EXEC_TIMEOUT=30000            # ms, default 30000
# LOCAL_EXEC_MAX_OUTPUT=1048576       # bytes, default 1 MiB
# EXECUTION_MODE=                     # docker | local | auto

# --- Filesystem reach ---
# Allow agent file tools to read/write the user's home directory.
# ALLOW_HOME_DIR_ACCESS=true
#
# Allow coding agents to operate in ANY directory, not just allowlisted ones.
# OWNPILOT_CODING_AGENT_ANY_DIR=true
#
# Host filesystem mapping for containerised deployments (path translation only).
# OWNPILOT_HOST_FS=
# OWNPILOT_HOST_FS_HOST_PREFIX=

# --- Extension / skill execution ---
# Allow installed skills to run shell scripts.
# OWNPILOT_ENABLE_SKILL_SCRIPTS=true
#
# Allow extension tools to reach host resources from their sandbox.
# OWNPILOT_ENABLE_EXTENSION_HOST_ACCESS=true

# --- Error disclosure ---
# In production, internal error details are suppressed in API responses.
# Setting this returns raw error messages to clients — debugging only.
# EXPOSE_INTERNAL_ERRORS=true
#
# Serve the OpenAPI docs UI in production (always on outside production).
# ENABLE_API_DOCS=true

# --- Proxy / transport trust ---
# Trust X-Forwarded-For / X-Real-IP for client IP and scheme. Enable ONLY when
# the gateway sits behind a proxy you control — otherwise clients can spoof
# their own IP, defeating rate limiting and audit attribution.
# TRUSTED_PROXY=true
#
# Redirect HTTP to HTTPS and enable HSTS. Requires TLS termination in front.
# HTTPS_ONLY=true
# HTTPS_PORT=8443

# ===========================================
# Auth tuning
# ===========================================
# One-time token for first-run UI bootstrap (initial password set).
# BOOTSTRAP_TOKEN=
# JWT_MAX_TOKEN_AGE=7d               # default 7d
# JWT_CLOCK_TOLERANCE_SEC=30         # default 30
# Shared secret for the inbound email webhook. Unset = endpoint fails closed.
# EMAIL_WEBHOOK_SECRET=

# ===========================================
# Operational tuning
# ===========================================
# Database
# DB_STATEMENT_TIMEOUT_MS=30000      # default 30000
# DB_IDLE_TX_TIMEOUT_MS=60000        # default 60000
# DB_BACKUP_TIMEOUT_MS=900000        # default 900000 (15 min)
# DB_KEEPALIVE=true                  # default true; set "false" to disable
#
# Inbound channel flood guard (Telegram/Discord/etc.)
# CHANNEL_INBOUND_RATE_LIMIT_MAX=20          # default 20
# CHANNEL_INBOUND_RATE_LIMIT_WINDOW_MS=60000 # default 60000
#
# Memory hashing salt — changing this invalidates existing derived keys.
# MEMORY_SALT=your-random-salt-here
#
# Extra provider stream logging (very verbose).
# OWNPILOT_DEBUG_STREAM=1
# DEBUG_EXEC_SECURITY=true

# Note: provider API keys (OPENAI_API_KEY, ANTHROPIC_API_KEY, DEEPSEEK_API_KEY,
# COMPOSIO_API_KEY, cloud credentials, …) are intentionally NOT listed here.
# They are managed through the Config Center UI and stored encrypted at rest.
# Env fallbacks exist for headless setups but the UI is the supported path.