Configuration

v1 preview

The complete repository and supported production Compose release are not public during beta. This reference describes the intended v1 configuration surface and may change before release. It is not a current deployment guide.

The planned self-hosted release uses environment variables for application, database, cache, search, AI, mail, and administration settings. Final required values, defaults, secret handling, and service versions will be published with v1.

Compose File

The planned compose.prod.yml file defines the service topology and reads its configuration from environment variables or a Docker secrets workflow. The public v1 release will identify which values are required and provide the supported startup procedure.

Environment Variables

Application

Variable	Description	Required	Default
APP_URL	Public URL of the application	Yes	—
WS_URL	Public WebSocket URL. Set this when the WebSocket server runs on a different host or port than the app	No	Same as APP_URL
PORT	Application port	No	3000
SESSION_SECRET	Express session secret	Yes	—
JWT_SECRET	JWT signing secret	Yes	—
NODE_ENV	Environment mode	No	production
SERVER_MODE	Run mode: omit for full app, ws for WebSocket-only, scheduler for background jobs	No	—
SOCKET_IO_ADAPTER	Socket.IO cluster transport: mongodb or memory	No	mongodb
SOCKET_IO_MONGO_URL	Optional MongoDB connection override for the MongoDB adapter	No	MONGO_URI

The MongoDB Socket.IO adapter stores events in the socketio collection and uses change streams, so it requires a MongoDB replica set or sharded cluster.

Database

Variable	Description	Required	Default
MONGO_URI	MongoDB connection string	No	mongodb://mongo:27017/managani

Cache

Variable	Description	Required	Default
MEMCACHED_SERVERS	Comma-separated Memcached endpoints	No	memcached:11211 in Docker Compose

Search (Typesense)

Variable	Description	Required	Default
TYPESENSE_HOST	Typesense server host	No	typesense
TYPESENSE_PORT	Typesense server port	No	8108
TYPESENSE_PROTOCOL	Protocol (http or https)	No	http
TYPESENSE_API_KEY	Typesense API key	Yes	—
TYPESENSE_NODES	Multi-node cluster config as JSON. Example: [{"host":"ts1","port":8108,"protocol":"http"},{"host":"ts2","port":8108,"protocol":"http"}]	No	—
TYPESENSE_COLLECTION_PREFIX	Prefix added to all collection names (mg_urls_<id>), so multiple products can share one Typesense cluster. Set to an empty string to keep unprefixed names on a dedicated cluster.	No	mg

When TYPESENSE_NODES is set, the individual host/port/protocol variables are ignored and the cluster config is used instead.

AI / LLM

Managani uses plan-aware model tiers for AI tasks:

Variable	Description	Required	Default
LLM_BASIC_MODEL	Lower-cost model for basic plans, parsing, classification, and internal analysis	No	gpt-5.4-mini
LLM_BASIC_MODEL_PROVIDER	Provider for the basic model: google or openai	No	openai
LLM_TOP_MODEL	Top model for Pro/customer-facing replies and generated content	No	gpt-5.6-terra
LLM_TOP_MODEL_PROVIDER	Provider for the top model: google or openai	No	openai
CHAT_AI_MODEL	Legacy alias for LLM_TOP_MODEL when LLM_TOP_MODEL is unset	No	gpt-5.6-terra
CHAT_AI_MODEL_PROVIDER	Legacy alias for LLM_TOP_MODEL_PROVIDER when LLM_TOP_MODEL_PROVIDER is unset	No	openai
NL_SEARCH_MODEL	Optional override for intent classification	No	gpt-5.4-mini
NL_SEARCH_MODEL_PROVIDER	Optional provider override for intent classification	No	openai
TS_CONVERSATION_MODEL	Optional override for Typesense conversation search	No	gpt-5.4-mini
TS_CONVERSATION_MODEL_PROVIDER	Optional provider override for Typesense conversation search	No	openai
GOOGLE_API_KEY	Google AI API key (required when using Google models)	Conditional	—
OPENAI_API_KEY	OpenAI API key (required when using OpenAI models)	Conditional	—
WIDGET_CHAT_EMAIL_FALLBACK_ENABLED	Global emergency switch for unread chat fallback email	No	true
WIDGET_CHAT_EMAIL_FALLBACK_CRON	Scheduler cadence for due fallback email	No	* * * * *
WIDGET_CHAT_EMAIL_FALLBACK_DELAY_MINUTES	Unread delay reset after each team or AI reply	No	3
WIDGET_CHAT_EMAIL_FALLBACK_MAX_PER_RUN	Maximum fallback deliveries claimed per scheduler run	No	50

At least one API key is required for AI chat to function. Hosted Free and Starter tenants are downgraded to the basic model for top-tier requests; hosted Pro tenants use the top model.

Email (SMTP)

Variable	Description	Required	Default
SMTP_HOST	SMTP server host	Yes	—
SMTP_PORT	SMTP server port	No	587
SMTP_USER	SMTP username	Yes	—
SMTP_PASS	SMTP password	Yes	—
SMTP_FROM	Sender email address	Yes	—

Admin

Variable	Description	Required	Default
SYSADMIN_EMAIL	System admin email. If set, a sysadmin account is provisioned on startup	No	—
SYSADMIN_PASSWORD	System admin password	No	—

Analytics (OpenPanel)

Variable	Description	Required	Default
ENABLE_OPENPANEL	Enable OpenPanel analytics	No	false
OPENPANEL_CLIENT_ID	OpenPanel client ID	No	—
OPENPANEL_CLIENT_SECRET	OpenPanel client secret	No	—
OPENPANEL_API_URL	OpenPanel API URL	No	—

Reverse Proxy

The planned production deployment will sit behind a reverse proxy such as nginx or Caddy. The final v1 guide will require operators to:

1. Set APP_URL to your public domain
2. Proxy WebSocket connections to port 3001
3. Proxy MCP connections to port 3002
4. Set X-Forwarded-For and X-Forwarded-Proto headers

Example nginx configuration:

server {
    listen 443 ssl;
    server_name your-instance.com;

    location / {
        proxy_pass http://localhost:3000;
        proxy_set_header Host $host;
        proxy_set_header X-Real-IP $remote_addr;
        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
        proxy_set_header X-Forwarded-Proto $scheme;
    }

    location /ws {
        proxy_pass http://localhost:3001;
        proxy_http_version 1.1;
        proxy_set_header Upgrade $http_upgrade;
        proxy_set_header Connection "upgrade";
    }

    location /mcp {
        proxy_pass http://localhost:3002;
        proxy_http_version 1.1;
        proxy_set_header Upgrade $http_upgrade;
        proxy_set_header Connection "upgrade";
    }
}