.env.example

# Database Configuration
DATABASE_HOST=postgres
DATABASE_PORT=5432
DATABASE_USERNAME=omni
DATABASE_PASSWORD=omni
DATABASE_NAME=omni
DATABASE_SSL=false
DB_MAX_CONNECTIONS=10
DB_ACQUIRE_TIMEOUT_SECONDS=3

# Redis Configuration
REDIS_URL=redis://redis:6379

# Core Service Ports
WEB_PORT=3000
SEARCHER_PORT=3001
INDEXER_PORT=3002
AI_SERVICE_PORT=3003
CONNECTOR_MANAGER_PORT=3004

# Connector Service Ports
GOOGLE_CONNECTOR_PORT=4001
SLACK_CONNECTOR_PORT=4002
ATLASSIAN_CONNECTOR_PORT=4003
WEB_CONNECTOR_PORT=4004
GITHUB_CONNECTOR_PORT=4005
HUBSPOT_CONNECTOR_PORT=4006
MICROSOFT_CONNECTOR_PORT=4007
NOTION_CONNECTOR_PORT=4008
FIREFLIES_CONNECTOR_PORT=4009
IMAP_CONNECTOR_PORT=4010
CLICKUP_CONNECTOR_PORT=4011
LINEAR_CONNECTOR_PORT=4012
FILESYSTEM_CONNECTOR_PORT=4013
NEXTCLOUD_CONNECTOR_PORT=4014
PAPERLESS_CONNECTOR_PORT=4015
GOOGLE_ADS_CONNECTOR_PORT=4016
DARWINBOX_CONNECTOR_PORT=4017
WINDSHIFT_CONNECTOR_PORT=4018

# Sandbox Port
SANDBOX_PORT=8090

# Optional Service Ports
LOCAL_EMBEDDINGS_PORT=8001 # For local embedding models via TEI
DOCLING_PORT=8003 # For AI-based document conversion via Docling
LOCAL_INFERENCE_MODEL_PORT=8000 # For local LLM via llama.cpp (see docker-compose.local-inference.yml)

# =============================================================================
# Resource limits
# =============================================================================
# CPU limits are expressed in cores. Defaults are conservative for a 4-core host
# so the Omni stack does not make the machine unresponsive under load.
# Increase these values on larger machines.
OMNI_POSTGRES_CPUS=1.0
OMNI_REDIS_CPUS=0.2
OMNI_MIGRATOR_CPUS=0.25
OMNI_SEARCHER_CPUS=0.5
OMNI_INDEXER_CPUS=0.25
OMNI_AI_CPUS=0.75
OMNI_SANDBOX_CPUS=0.5
OMNI_CONNECTOR_MANAGER_CPUS=0.5
OMNI_WEB_CPUS=0.5
OMNI_CADDY_CPUS=0.2
OMNI_CONNECTOR_CPUS=0.2
OMNI_EMBEDDINGS_CPUS=0.75
DOCLING_CPUS=0.75

# Relative CPU weights used when containers contend for CPU.
OMNI_CPU_SHARES_CRITICAL=2048
OMNI_CPU_SHARES_HIGH=1024
OMNI_CPU_SHARES_NORMAL=512
OMNI_CPU_SHARES_BACKGROUND=256
OMNI_CPU_SHARES_CONNECTOR=128

OMNI_POSTGRES_MEMORY=3g
OMNI_POSTGRES_SHM_SIZE=2g
OMNI_REDIS_MEMORY=256m
OMNI_MIGRATOR_MEMORY=256m
OMNI_SEARCHER_MEMORY=512m
OMNI_INDEXER_MEMORY=512m
OMNI_AI_MEMORY=1g
OMNI_SANDBOX_MEMORY=512m
OMNI_CONNECTOR_MANAGER_MEMORY=3g
OMNI_WEB_MEMORY=2g
OMNI_CADDY_MEMORY=128m
OMNI_CONNECTOR_MEMORY=384m
OMNI_EMBEDDINGS_MEMORY=2g
DOCLING_MEMORY=2g

# =============================================================================
# Enable required connectors
#
# Enable connectors you want to run by adding their profile to ENABLED_CONNECTORS (comma-separated).
# Available connector names:
# 	google, google_ads, slack, atlassian, web, github, notion, hubspot, fireflies, microsoft, filesystem, imap, linear, clickup, nextcloud, paperless, darwinbox, windshift
#
# Example: ENABLED_CONNECTORS=google,slack
#
# By default, as configured below, only the web connector will run.
# =============================================================================
ENABLED_CONNECTORS=web
COMPOSE_PROFILES=${ENABLED_CONNECTORS}

# Service URLs (for inter-service communication)
SEARCHER_URL=http://searcher:${SEARCHER_PORT}
INDEXER_URL=http://indexer:${INDEXER_PORT}
AI_SERVICE_URL=http://ai:${AI_SERVICE_PORT}
CONNECTOR_MANAGER_URL=http://connector-manager:${CONNECTOR_MANAGER_PORT}

# Sandbox URL
SANDBOX_URL=http://sandbox:${SANDBOX_PORT}

# Memory (in-process in the AI service)
MEMORY_ENABLED=false
# Backend implementation. Options: mem0
MEMORY_PROVIDER=mem0

# Optional service URLs
LOCAL_EMBEDDINGS_URL=http://embeddings:${LOCAL_EMBEDDINGS_PORT}/v1
DOCLING_URL=http://docling:${DOCLING_PORT}

# Docling Configuration (AI-based document conversion)
#
# Opt-in deployment:
#   1. Set DOCLING_ENABLED=true (controls whether the Document Conversion admin
#      settings page is shown).
#   2. Activate the `docling` compose profile when starting the stack, e.g.:
#        docker compose -f docker/docker-compose.yml --profile docling up -d
#      (or add `docling` to COMPOSE_PROFILES above).
DOCLING_ENABLED=false
DOCLING_DEVICE=""  			# Leave empty to pull the default, CPU-only image. Set to "cuda" to pull the CUDA-enabled image
DOCLING_MAX_CONCURRENT_CONVERSIONS=1

# Connector Manager Configuration
MAX_CONCURRENT_SYNCS=10
MAX_CONCURRENT_SYNCS_PER_TYPE=3
EXTRACTION_CONCURRENCY=2 # Max concurrent document extraction requests handled by connector-manager
EXTRACTION_RETRY_AFTER_SECONDS=30
SCHEDULER_POLL_INTERVAL_SECONDS=60
STALE_SYNC_TIMEOUT_MINUTES=60

# Session Configuration
SESSION_COOKIE_NAME=auth-session
SESSION_DURATION_DAYS=7

# Application Configuration
OMNI_VERSION=latest
APP_URL=http://localhost:3000
OMNI_DOMAIN=localhost
ACME_EMAIL=admin@yourcompany.com

# Feature Flags
AI_ANSWER_ENABLED=false # AI generated summary on search results page
AGENTS_ENABLED=false # Background/scheduled agents feature

# AI Service Configuration
AI_WORKERS=2 # The number of workers to spawn in the omni-ai service
MODEL_PATH=/models
AGENT_MAX_ITERATIONS=15
APPROVAL_TIMEOUT_SECONDS=600

# Searcher Service Configuration
RAG_CONTEXT_WINDOW=2 # The number of surrounding chunks to fetch in RAG search
SEMANTIC_SEARCH_TIMEOUT_MS=1000 # Semantic search (vector search) will timeout if it takes longer than this

# Google Workspace Connector
WEBHOOK_RENEWAL_CHECK_INTERVAL_SECONDS=3600
GOOGLE_MAX_AGE_DAYS=712 # Documents older than this will not be indexed

# Web Connector Configuration
WEB_SYNC_INTERVAL_SECONDS=86400  # Daily recrawl (24 hours)

# Optional: Spider Cloud (https://spider.cloud) for higher crawl success rates.
# Sites behind Cloudflare, a WAF, or IP/geo blocking return 403/429/503 to a
# direct fetch and are dropped from the index. Spider Cloud routes crawls
# through rotating proxies and anti-bot bypass instead.
# Leave SPIDER_CLOUD_API_KEY unset to keep the current direct-fetch behavior.
# SPIDER_CLOUD_API_KEY=sk-...
# Mode: smart (default) | proxy | unblocker | api | fallback
#   smart     - proxy everything, auto-escalate to the unblocker on bot walls
#   proxy     - proxy transport only, cheapest
#   unblocker - always use the unblocker API
#   api       - fetch each page through the crawl API
#   fallback  - direct fetch first, cloud only after a failure
# SPIDER_CLOUD_MODE=smart
# SPIDER_CLOUD_API_URL=https://api.spider.cloud

# Log level for all rust services
RUST_LOG=info
RUST_BACKTRACE=

# =============================================================================
# Embedding Configuration
# =============================================================================
# Embedding provider settings (provider, model, dimensions, API key/URL) are
# managed in the database via the UI.
EMBEDDING_MAX_MODEL_LEN=8192

# AWS configuration (for online bedrock embedding/LLM provider)
AWS_REGION=

# Storage Backend. One of: s3, postgres
STORAGE_BACKEND=postgres

# Only required if STORAGE_BACKEND=s3
S3_BUCKET=your-s3-bucket
S3_REGION=your-s3-region

# Encryption Configuration. Used to encrypt sensitive keys to the connectors.
# Service account credentials, API keys/tokens, etc.
ENCRYPTION_KEY=your-encryption-key-must-be-at-least-32-characters-long
ENCRYPTION_SALT=your-salt-16-chars

# OpenTelemetry Configuration
# Leave OTEL_EXPORTER_OTLP_ENDPOINT empty for local-only telemetry
OTEL_EXPORTER_OTLP_ENDPOINT=
OTEL_DEPLOYMENT_ID=
OTEL_DEPLOYMENT_ENVIRONMENT=production
SERVICE_VERSION=0.1.0