dockette/memzero

dockette/memzero repository overview

⁠Dockette / Memzero

🐳 Mem0⁠ self-hosted memory layer (API + dashboard + pgvector) in a single stack.

⁠Motivation

Mem0 ships a self-hosted server, but no maintained prebuild images (the last mem0/mem0-api-server push is from 2025 and predates auth + dashboard). This repository provides prebuild multiarch images and a ready-to-run stack.

dockette/memzero:server - Mem0 REST API (FastAPI + Alembic migrations)
dockette/memzero:dashboard - Mem0 dashboard (Next.js)

Both images are built from a pinned upstream tag (MEM0_VERSION).

⁠Usage

Create .env file.

# Docker images
DOCKER_SERVER_IMAGE=dockette/memzero:server
DOCKER_DASHBOARD_IMAGE=dockette/memzero:dashboard

# Docker: ports
MEM0_PORT=8888
DASHBOARD_PORT=3000
POSTGRES_PORT=8432

# Docker: mem0
OPENAI_API_KEY=sk-...
JWT_SECRET=change-me
AUTH_DISABLED=false
MEM0_TELEMETRY=false
MEM0_DEFAULT_LLM_MODEL=gpt-5-mini
MEM0_DEFAULT_EMBEDDER_MODEL=text-embedding-3-small

# Docker: dashboard
DASHBOARD_URL=http://localhost:3000
DASHBOARD_API_URL=http://localhost:8888
DASHBOARD_INSTANCE_NAME=Memzero

# Docker: postgres
POSTGRES_DB=postgres
POSTGRES_USER=postgres
POSTGRES_PASSWORD=change-me
POSTGRES_COLLECTION_NAME=memories
APP_DB_NAME=mem0_app

Important
`OPENAI_API_KEY`, `JWT_SECRET` and `POSTGRES_PASSWORD` are required. The API refuses to boot without a key for the default LLM / embedder provider.

Create docker-compose.yml file.

Copy docker-compose.yml⁠ and postgres/init-db.sh⁠ files.

Start Docker services.

docker compose up

Open http://localhost:3000 in your browser and finish the setup wizard.

It creates the first admin account and the first API key.

Use the API.

curl -X POST http://localhost:8888/memories \
	-H "X-API-Key: your-api-key" \
	-H "Content-Type: application/json" \
	-d '{"messages":[{"role":"user","content":"I love hiking"}],"user_id":"alice"}'

OpenAPI docs are on http://localhost:8888/docs.

⁠Services

Service	Image	Port	Description
`mem0`	`dockette/memzero:server`	`8888`	REST API, runs migrations on start
`dashboard`	`dockette/memzero:dashboard`	`3000`	Web UI (memories, entities, API keys)
`postgres`	`pgvector/pgvector:pg17`	`8432`	Vector storage + app database

Data is persisted in .docker/postgres (Postgres) and .docker/history (SQLite history db).

⁠Configuration

Server

Variable	Default	Description
`OPENAI_API_KEY`		LLM + embedder key (`ANTHROPIC_API_KEY`, `GOOGLE_API_KEY` are bundled too)
`MEM0_DEFAULT_LLM_MODEL`	`gpt-5-mini`	Default LLM model
`MEM0_DEFAULT_EMBEDDER_MODEL`	`text-embedding-3-small`	Default embedder model
`JWT_SECRET`		Secret for dashboard sessions, required
`ADMIN_API_KEY`		Optional master key for `X-API-Key`
`AUTH_DISABLED`	`false`	Local development only, never in production
`MEM0_TELEMETRY`	`false`	Upstream anonymous telemetry (on by default upstream)
`WORKERS`	`1`	Uvicorn workers
`HISTORY_DB_PATH`	`/app/history/history.db`	SQLite memory history db
`SKIP_MIGRATIONS`		Set to skip `alembic upgrade head` on start

Postgres

Variable	Default	Description
`POSTGRES_HOST`	`postgres`	Host
`POSTGRES_PORT`	`5432`	Port
`POSTGRES_DB`	`postgres`	Database used by pgvector for memories
`POSTGRES_USER`	`postgres`	User
`POSTGRES_PASSWORD`		Password, required
`POSTGRES_COLLECTION_NAME`	`memories`	Vector collection
`APP_DB_NAME`	`mem0_app`	Database for users / api keys / config

Tip
For more detailed configuration options, please refer to the [Mem0 official documentation](https://docs.mem0.ai/).

⁠Build

make build      # build both images
make push       # push both images
make up         # start the stack
make health     # check the stack
make clean      # stop the stack and drop data

Bump MEM0_VERSION in .env.dist to build a newer upstream release.

⁠Development

See how to contribute⁠ to this package.

This package is currently maintaining by these authors.

⁠

Consider to support⁠ f3l1x. Also thank you for using this package.

Tag summary

Recent tags

Content type

Image

Digest

sha256:c178f5005…

Size

73.3 MB

Last updated

about 3 hours ago

docker pull dockette/memzero:dashboard