All-in-one Docker Compose Deployment

← Deployment · 简体中文

Bring up every service with one command from the docker/ directory at the repository root: backend (aideepin-api), Redis, the user web, the admin web, and nginx. PostgreSQL is not included in the compose stack — install and initialize it first.

1. Initialize the Database

Install PostgreSQL and Extensions

Storage falls into three parts; the vector store and the graph store can each run either as a PostgreSQL extension or on Neo4j:

Purpose	PostgreSQL option (default)	Alternative	Config key
Primary database	PostgreSQL (required)	—	—
Vector search	pgvector extension	Neo4j	adi.vector-database: pgvector / neo4j
Knowledge graph	Apache AGE extension	Neo4j	adi.graph-database: apache-age / neo4j

PostgreSQL itself is required in every setup — only the extensions are optional. On Ubuntu, add the PGDG repository first, then install PostgreSQL and the extensions in one command:

sudo apt install -y postgresql-16 postgresql-16-pgvector postgresql-16-age

NOTE

Install the extensions only if you use them: for any store running on Neo4j, skip the corresponding apt package and CREATE EXTENSION statement.

Enable the extensions and create the database:

CREATE EXTENSION IF NOT EXISTS vector;
CREATE EXTENSION IF NOT EXISTS age;
CREATE DATABASE aideepin;

Keeping the defaults (pgvector + Apache AGE)

No extra configuration needed.

Switching the vector store or graph store to Neo4j

The two are independent — switch either one or both. For Docker Compose deployments:

1. Append the matching flag to JAVA_OPTS in .env: -Dadi.vector-database=neo4j for the vector store, -Dadi.graph-database=neo4j for the graph store
2. Uncomment the ADI_NEO4J_* entries in docker-compose.yml and fill in the connection details in .env (required if either one is switched)

For source deployments, edit the matching application-*.yml directly.

Run the initialization scripts

The scripts live in server/db_migration/; run them in order:

1. all_ddl.sql — creates all tables, triggers, and comments
2. all_dml.sql — inserts base data (admin account, system config, model platforms, AI models)
3. all_dml_cn.sql or all_dml_en.sql — pick one; inserts language-specific display data (preset characters, workflow components, MCP services)

NOTE

The default SCHEMA in the DDL is public; adjust it if yours differs. Naming conventions and details are in the db_migration README.

2. Configure and Start

cd docker

# Edit .env (compose reads .env by default; use .env.prod for production)
# ⚠️ Must update: ADI_DB_HOST, ADI_DB_USERNAME, ADI_DB_PASSWORD,
#    ADI_MAIL_HOST, ADI_MAIL_USERNAME, ADI_MAIL_PASSWORD, ADI_ENCRYPT_AES_KEY

# Servers with ≥ 8 GB RAM: build everything in one step
docker compose up -d --build

# Servers with < 8 GB RAM: build one image at a time to avoid OOM
docker compose build aideepin-admin-web aideepin-user-web
docker compose build aideepin-api
docker compose up -d

Common commands: docker compose logs -f to follow logs; docker compose down to stop all services.

3. Architecture

Browser → nginx:80
           ├── /api/      → aideepin-api:9999 (backend)
           ├── /admin/    → aideepin-admin-web:80 (admin web)
           └── /          → aideepin-user-web:80 (user web)

4. First-time Initialization

Enable at least one model platform after startup — otherwise regular users won't be able to use the app. The admin account is initialized by all_dml.sql; after signing in to the admin console:

• Recommended: fill in the platform API keys and enable platforms under "Platform Configuration" (see the Admin Guide);
• or configure directly via SQL (examples):

-- DeepSeek
UPDATE adi_model_platform SET api_key = 'my_deepseek_secret_key' WHERE name = 'deepseek';

-- OpenAI
UPDATE adi_model_platform SET api_key = 'my_openai_secret_key' WHERE name = 'openai';

-- DashScope
UPDATE adi_model_platform SET api_key = 'my_dashscope_api_key' WHERE name = 'dashscope';

-- SiliconFlow
UPDATE adi_model_platform SET api_key = 'my_siliconflow_api_key' WHERE name = 'siliconflow_setting';

-- Ollama
UPDATE adi_model_platform SET base_url = 'my_ollama_base_url' WHERE name = 'ollama';

Then enable the models you need:

UPDATE adi_ai_model SET is_enable = true WHERE name = 'deepseek-flash';

Configuration Reference (.env)

Variable	Description	Default
TZ	Timezone	Asia/Shanghai
IMAGE_VERSION	Backend image version	v1.1.0-dev
APP_VERSION	Application version	1.0.0-SNAPSHOT
APP_PROFILE	Spring profile (dev/prod)	dev
JAVA_OPTS	JVM options	-Xms1024m -Xmx2048m ...
ADI_DB_HOST	PostgreSQL host	localhost
ADI_DB_PORT	PostgreSQL port	5432
ADI_DB_NAME	PostgreSQL database name	aideepin
ADI_DB_USERNAME	PostgreSQL user	your-db-account
ADI_DB_PASSWORD	PostgreSQL password	your-db-password
ADI_REDIS_HOST	Redis host	redis
ADI_REDIS_PORT	Redis port	6379
ADI_MAIL_HOST	SMTP mail host	your-email-host
ADI_MAIL_USERNAME	SMTP mail username	your-email-username
ADI_MAIL_PASSWORD	SMTP mail password	your-email-password
ADI_ENCRYPT_AES_KEY	AES encryption key (16 chars)	your-16-char-key
ADI_NEO4J_HOST	Neo4j host (optional)	localhost
ADI_NEO4J_PORT	Neo4j port (optional)	7687
ADI_NEO4J_USERNAME	Neo4j username (optional)	neo4j
ADI_NEO4J_PASSWORD	Neo4j password (optional)	your-neo4j-password
ADI_NEO4J_DATABASE	Neo4j database (optional)	neo4j

TIP

For production, set APP_PROFILE to prod and manage configuration via .env.prod.

---

Previous: Deployment · Next: Per-service Docker deployment