@openinference/cli

OpenInference Core — a package manager for local AI models.

oi finds, installs, and runs the right open-source model for your computer — the same way apt, brew, or npm manage software. It is hardware-aware: it scans your machine and only offers models that will actually run on it.

Powered by a local inference engine (Ollama today) — but you manage models, not the engine.

Quick start

Install globally, then type oi anytime:

npm install -g @openinference/cli
oi

One line (install + open the terminal):

npm install -g @openinference/cli && oi

On Linux you may need sudo npm install -g @openinference/cli, then run oi.

The default is an interactive shell — chat when you already have a model, or /setup to run the wizard.

The setup wizard (oi start):

1. Ask what you want AI for (coding, chat, PDFs, …)
2. Scan RAM, CPU, GPU, disk, and OS
3. Filter 150+ models → what fits your machine
4. Let you pick and confirm before downloading
5. Install Ollama, pull the model, verify with a quick test
6. On tiny VMs (<4 GB RAM), only micro models are offered; if a model crashes, -y auto-retries the next smallest fit

Skip prompts (power users):

oi start -y
# safest on a 3–4 GB cloud VM:
oi start -y -m smollm2:135m

Commands

Familiar, package-manager-style commands. Older names are kept as aliases.

| Command | Aliases | Description | |---------|---------|-------------| | oi | | Interactive shell (default) — chat + /search, /install, /setup | | oi start | setup | Setup wizard | | oi start -y | | Auto-pick and install (retries on crash) | | oi search <query> | find | Search models — shows installed vs available | | oi info <model> | show | Details: RAM, size, fit, installed state | | oi install <model> | pull, add | Download a model | | oi use [model] | | Switch active model (oi use = pick from installed) | | oi list | models, ls | List installed models | | oi remove <model> | rm, uninstall | Delete a model and free disk space | | oi recommend | | Preview picks for your hardware (no install) | | oi chat | | Chat with active model | | oi storage | | Where models are stored | | oi status | | Current setup |

Use cases

coding · chat · pdfs · writing · image · research

oi recommend --use-case coding
oi --use-case coding -y

Model storage

OpenInference does not store model files. Ollama downloads to:

• Windows: %USERPROFILE%\.ollama\models
• macOS / Linux: ~/.ollama/models

Requirements

• Node.js 18+
• Windows, macOS, or Linux

License

MIT