AGMind macOS Installer

Русская версия (README.ru.md)

One-command installer for a full local AI/RAG stack on macOS. Deploys Dify, Open WebUI, Ollama, Weaviate/Qdrant, PostgreSQL, Redis, Nginx — with optional Open Notebook and DB-GPT — optimized for Mac Studio and Mac Pro with native Metal GPU acceleration.

git clone https://github.com/botAGI/AGmind-macos.git
cd AGmind-macos
bash install.sh

Dry-run — validate the entire 9-phase flow without sudo, Docker, or any system changes:

bash install.sh --dry-run

Why AGMind?

Running AI locally on macOS means dealing with Docker quirks, Ollama networking, Compose profiles, nginx routing, LaunchAgents instead of systemd, and BSD tools instead of GNU. AGMind handles all of it in one command.

• Metal GPU acceleration — Ollama runs natively via Homebrew, not in Docker
• Zero config — wizard auto-detects your hardware and recommends models
• Optional tools — add Open Notebook or DB-GPT with a single "yes"
• Day-2 ready — agmind CLI for status, doctor, backup, stop/start, uninstall

What It Does

Phase	What happens
1. Diagnostics	Validates macOS 13+, RAM, disk, ports, brew, Docker
2. Wizard	Interactive or non-interactive config (profile, models, vector DB, optional tools)
3. Prerequisites	Docker (Colima or Desktop) + Compose v2
4. Ollama Setup	Native Ollama via Homebrew — Metal GPU acceleration
5. Configuration	Generates .env, nginx.conf, docker-compose.yml, TOML configs, secrets
6. Start	docker compose up + admin credential injection
7. Health	Polls all containers + Ollama API until healthy
8. Models	Pulls LLM + embedding models via Ollama
9. Complete	LaunchAgents, CLI install, final summary

Requirements

	Minimum
macOS	13 Ventura+
Architecture	Apple Silicon (arm64) or Intel (x86_64)
RAM	8 GB (24+ GB recommended)
Disk	30 GB free
Ports	80, 3000 free

Quick Start

Interactive

bash install.sh

The wizard asks 9 questions with smart defaults based on your hardware.

Non-Interactive

NON_INTERACTIVE=1 \
DEPLOY_PROFILE=lan \
LLM_MODEL=qwen2.5:14b \
EMBED_MODEL=nomic-embed-text \
VECTOR_DB=weaviate \
INSTALL_OPEN_NOTEBOOK=1 \
INSTALL_DBGPT=1 \
bash install.sh

Options

Flag	Description
--verbose	Debug-level output
--non-interactive	Skip prompts, read from env vars
--dry-run	Full 9-phase run without system changes
--force-phase N	Re-run phase N even if complete
--help	Show usage

Stack

Component	Runs as	Access
Dify (API + Worker + Web)	Docker	http://<ip>/apps/
Open WebUI	Docker	http://<ip>/
Ollama	Native (brew)	http://localhost:11434
Weaviate or Qdrant	Docker	internal
PostgreSQL + Redis	Docker	internal
Nginx	Docker	port 80
Open Notebook (optional)	Docker	http://<ip>/notebook/
DB-GPT (optional)	Docker	http://<ip>/dbgpt/

Optional Tools

Open Notebook

AI research notebook (lfnovo/open-notebook) — open-source alternative to Google NotebookLM. PDF, video, audio ingestion with vector search and podcast generation. Deploys with SurrealDB.

DB-GPT

AI database assistant (eosphoros-ai/DB-GPT) — Text-to-SQL, RAG over databases, multi-agent orchestration. SQLite mode with Ollama proxy — no GPU needed in Docker.

Both are Docker Compose profile-gated. They connect to native Ollama via host.docker.internal:11434.

Model Recommendations

Auto-selected based on your unified memory:

RAM	LLM
8 GB	gemma3:4b
16 GB	qwen2.5:7b
32 GB	qwen2.5:14b
64 GB	gemma3:27b
96 GB+	qwen2.5:72b

CLI

agmind status      # Service states, Ollama, LAN IP, optional tools
agmind doctor      # Health checks — PASS/WARN/FAIL
agmind start       # Ollama first, then Docker Compose
agmind stop        # Docker Compose down, then Ollama
agmind restart     # Stop + Start
agmind logs [svc]  # Compose logs (tail 50) or follow specific service
agmind backup      # Manual backup with rotation
agmind uninstall   # Full removal with confirmation

Profiles

Profile	Description
lan	Local network, no TLS
offline	Air-gapped, no internet

Project Structure

agmind-macos/
├── install.sh                          # 9-phase orchestrator + dry-run
├── lib/                                # 11 bash modules
│   ├── common.sh                       #   logging, errors, BSD utilities
│   ├── detect.sh                       #   hardware detection, preflight
│   ├── wizard.sh                       #   interactive config (9 questions)
│   ├── docker.sh                       #   Colima / Docker Desktop
│   ├── ollama.sh                       #   native Ollama via brew
│   ├── config.sh                       #   templates, secrets, TOML
│   ├── compose.sh, health.sh           #   deploy + health polling
│   ├── models.sh, openwebui.sh         #   model pull + admin init
│   └── backup.sh                       #   LaunchAgent management
├── scripts/
│   ├── agmind.sh                       # CLI (8 commands)
│   ├── backup.sh                       # scheduled backup + rotation
│   └── health-gen.sh                   # JSON health reports
├── templates/                          # Compose, env, nginx, TOML, plists
└── tests/
    ├── unit/          (230 BATS tests)
    ├── integration-test.sh
    └── helpers/       (mock executables)

Testing

bash install.sh --dry-run               # Full flow, no sudo, 1 second
bats tests/unit/                        # 230 unit tests
bash tests/integration-test.sh          # Sandboxed 9-phase run
shellcheck lib/*.sh scripts/*.sh        # Lint

Security

• Secrets from /dev/urandom (32-char alphanumeric)
• .env and credentials.txt mode 600, umask 077
• Credential key whitelist on credentials.txt read
• sed injection prevention (_sed_escape)
• sudo rm -rf guarded by path validation + .install-state marker
• curl timeouts (5s connect, 10s max) on all requests
• JSON via python3 json.dumps, not string concatenation

macOS-Specific Design

Linux	macOS (AGMind)
apt-get	brew
systemctl / cron	LaunchAgents
/proc/meminfo	sysctl hw.memsize
ss -tlnp	lsof -iTCP
GNU sed	BSD sed -i ''
timeout	counter-based loops
Docker GPU passthrough	Native Ollama (Metal)
bash 5+	bash 3.2.57 (stock)

License

Apache 2.0