llmport-cli

The single entry point to deploy, configure, and manage the llm.port platform — an open-source LLM gateway that routes, secures, and observes traffic across local runtimes and remote providers.

pipx install llmport-cli       # or: uv tool install llmport-cli
llmport deploy                 # installs LLM.Port into ~/llm-port and starts it

The CLI carries the deployment files of its release and runs that release's published images: llmport-cli 0.3.0 deploys LLM.Port 0.3.0. To upgrade, upgrade the CLI and run llmport upgrade. See Installing LLM.Port.

Features

• One-command production deploy — pre-flight checks, .env generation, the release's images, database migrations, and service startup.
• Upgrade, backup and restore — llmport upgrade backs up every database, moves the install to the CLI's release and keeps your settings.
• Auto-tuning — detects host CPU / RAM and computes optimal worker counts, DB pool sizes, and queue channel pools.
• Module management — enable / disable optional services (PII redaction, Auth, Mailer, Docling OCR) via Docker Compose profiles.
• GPU auto-detection — discovers NVIDIA (CUDA), AMD (ROCm), and Intel GPUs; selects the correct vLLM container image automatically.
• Developer workflow — clone all repos, install deps, run infra + migrations, and generate a VS Code workspace in one command.
• Rich terminal UI — tables, progress bars, and themed output powered by Rich. Interactive TUI wizard via Textual.
• Command abbreviation — type llmport st instead of llmport status.

Requirements

Git is only needed to deploy from a source checkout (--build) or for the developer workflow.

Installation

From PyPI

pip install llmport-cli

# or with uv
uv tool install llmport-cli

From source

git clone https://github.com/llm-port/llm-port-core.git
cd llm-port-core/llm_port_cli
uv sync                   # install deps + editable entry point
uv run llmport --help

Verify

llmport version
llmport doctor

Quick start

Production deployment

# Full deploy — pre-flight, env gen, pull this release's images, migrate, start
llmport deploy                      # into ~/llm-port
llmport deploy /opt/llmport         # into a directory of your choice

# Enable optional modules
llmport deploy --modules pii,auth

# From a source checkout: build the images instead of pulling them
llmport deploy --build

# Upgrade to the CLI's release (backs up first), back up, restore
llmport upgrade
llmport backup
llmport restore ~/llm-port/backups/<time>

# Provision node-agent on this host during deploy
llmport deploy /opt/llmport --local-node

# Provision node-agent on remote host over SSH
llmport deploy /opt/llmport --local-node --local-node-host ubuntu@10.0.0.12

Day-to-day operations

llmport up                          # start all services
llmport up --build llm-port-api     # rebuild a single service
llmport down                        # stop all services
llmport down --volumes              # stop + remove volumes
llmport status                      # service table
llmport logs -f                     # tail all logs
llmport logs -f llm-port-backend    # tail one service

Modules

llmport module list                 # show available modules
llmport module enable pii skills    # enable modules
llmport module disable mcp          # disable a module

Configuration

llmport config show                 # print current config
llmport config set dev.branch main  # update a value (dot-notation)
llmport config edit                 # open in $EDITOR
llmport config path                 # print config file location

Auto-tuning

llmport tune                        # detect resources, write .env
llmport tune --profile prod         # production-grade sizing
llmport tune --dry-run              # preview without writing

Admin utilities

llmport admin reset-password --email admin@localhost

Developer workflow

# Bootstrap everything — clone repos, install deps, start infra,
# run migrations, generate VS Code workspace
llmport dev init ~/projects/llm-port
llmport dev init ~/projects/llm-port --ssh        # SSH cloning
llmport dev init ~/projects/llm-port --overwrite   # force pull

# Start / stop / status
llmport dev up
llmport dev up --backend-only
llmport dev up --modules pii,mcp,skills   # also run the optional modules
llmport dev up --local-node
llmport dev up --local-node --local-node-host ubuntu@10.0.0.12
llmport dev down
llmport dev status

# Check dev prerequisites (optionally auto-install missing tools)
llmport dev doctor
llmport dev doctor --install --yes

--modules runs PII (8003), MCP (8007) and skills (8008) on the host beside the backend, and points the gateway and backend at them. Without it, the modules switched on with llmport module enable run; the others are switched off in the gateway and backend. See What the gateway does with a chat request.

--local-node installs a systemd unit by default and requires sudo on Linux. Use --local-node-no-sudo to skip privileged systemd setup.

Command reference

Project structure

llm_port_cli/
├── pyproject.toml
├── src/
│   └── llmport/
│       ├── cli.py                 # Click root group + AliasedGroup
│       ├── commands/
│       │   ├── version.py         # version
│       │   ├── doctor.py          # doctor
│       │   ├── deploy.py          # deploy
│       │   ├── up.py              # up
│       │   ├── down.py            # down
│       │   ├── status.py          # status
│       │   ├── logs_cmd.py        # logs
│       │   ├── config.py          # config group
│       │   ├── module.py          # module group
│       │   ├── tune.py            # tune
│       │   ├── admin.py           # admin group
│       │   └── dev/               # developer workflow
│       │       ├── dev_init.py
│       │       ├── dev_up.py
│       │       ├── dev_status.py
│       │       └── dev_doctor.py
│       ├── core/                  # shared utilities
│       │   ├── compose.py         # Docker Compose wrapper
│       │   ├── detect.py          # OS / GPU / port detection
│       │   ├── settings.py        # YAML config (~/.config/llmport/)
│       │   ├── api_client.py      # Backend REST client (httpx)
│       │   ├── bootstrap.py       # First-admin bootstrap
│       │   ├── env_gen.py         # .env generation (Jinja2)
│       │   ├── git.py             # Git clone / checkout helpers
│       │   ├── install.py         # Cross-platform tool installer
│       │   ├── registry.py        # Central metadata registry
│       │   ├── sysinfo.py         # Resource detection + tuning
│       │   └── console.py         # Rich themed output
│       ├── tui/                   # Textual TUI wizard
│       └── templates/             # Jinja2 .env templates
└── tests/

License

Apache License 2.0 — see LICENSE for details.

Part of the llm.port project.