Skip to main content

🧙 Sage — Your AI Assistant in the Terminal

Sage is a smart AI helper that lives in your computer's terminal. You type what you want in plain English, and Sage helps you get it done — reading and writing files, running tasks, understanding images, and even creating images.

Think of it as a helpful assistant you can chat with, right in your terminal.

Sage connects to any OpenAI-compatible API endpoint (such as a LiteLLM proxy), so you can point it at whatever models your provider offers.


⚠️ Before You Start

You need three things for Sage to work:

  1. Python installed on your computer (version 3.10 or newer). If you don't have it, download it from https://www.python.org/downloads/ and during installation, tick the box that says "Add python.exe to PATH".
  2. An API endpoint URL — the address of your OpenAI-compatible / LiteLLM server (for example, https://your-litellm-server.com).
  3. An API key for that endpoint.

(If you don't have the endpoint URL or key, ask whoever set up your AI gateway.)


📥 How to Install (One Time)

Sage is installed with pipx, a tool that installs Python command-line apps in their own isolated environment so they don't clash with anything else on your machine.

  1. Install pipx (skip if you already have it) — open a terminal and run:

    python -m pip install --user pipx
    python -m pipx ensurepath
    
  2. Close that terminal and open a new one. (Search for "Command Prompt" or "Terminal" in the Windows Start menu.) This step is needed so your terminal picks up the updated PATH.

  3. Install Sage:

    pipx install litellm-sage
    
  4. Test it — type this and press Enter:

    sage --version
    

    If you see a version number, you're ready! 🎉

    The first time you actually run sage (see below), it will start a short guided setup to collect your endpoint URL and API key.

You only do this once. To upgrade later, run pipx upgrade litellm-sage.


⚙️ Your Settings

The first time you run sage and it can't find a settings file, it automatically starts a short guided setup right there in the terminal:

  1. It creates a .sage folder in your user folder for you.
  2. It asks for your endpoint URL and API key — just answer the two questions.
  3. Everything else (MODEL, IMAGE_MODEL, VERIFY_TLS) is filled in with sensible defaults automatically.

That's it — once setup finishes, Sage starts normally and remembers your answers for next time. If you quit partway through (e.g. with Ctrl + C), no settings file is created, so the same guided setup will simply run again the next time you start sage.

Sage stores these settings in a file at:

C:\Users\<your-name>\.sage\.env

To change your settings later, you can edit this file directly instead of redoing the guided setup:

  1. Open .env in a text editor (make sure it's not saved as .env.txt — in Notepad, choose "All Files" as the type when saving).
  2. Update the values, one per line, like this:
    API_KEY=your-api-key-here
    BASE_URL=https://your-litellm-server.com
    MODEL=claude-opus-4-8
    IMAGE_MODEL=gemini-3-pro-image
    VERIFY_TLS=true
    

The available settings are:

Setting What it is
API_KEY Your API key
BASE_URL Your endpoint URL
MODEL The default AI model to chat with
IMAGE_MODEL The model used for creating images
VERIFY_TLS Set to false if your server uses a private certificate

API_KEY and BASE_URL are required — Sage won't start without them. MODEL, IMAGE_MODEL, and VERIFY_TLS are optional and fall back to sensible defaults if you leave them out.


🚀 How to Use Sage

  1. Open a terminal in the folder you want to work in. (Tip: in File Explorer, type cmd in the address bar and press Enter to open a terminal in that folder.)

  2. Type:

    sage
    
  3. Now just talk to it in plain English. For example:

    › summarize what the files in this folder are about
    › create a to-do list file with 5 example tasks
    › what does this spreadsheet contain?
    
  4. When you're done, type /exit or press Ctrl + C.


🎨 Fun & Useful Things Sage Can Do

Look at an image and describe it

› /attach photo.png
› what's in this picture?

(Replace photo.png with your image's name. It must be in the same folder.)

Create an image from words

› /image a cute cartoon airplane flying over clouds

Sage will create the image, save it in your folder, and open it for you.

Ask it to do a task and check its own work

› create a file called notes.txt with today's date, then show me it worked

🎚️ Modes — Controlling How Much Sage Does on Its Own

Sage asks for your permission before making changes. You can change how cautious it is by typing /mode followed by a name:

Type this What it does
/mode plan Only looks and suggests — changes nothing
/mode normal Asks you before every change (this is the default)
/mode auto-edits Edits files automatically, asks before running tasks
/mode auto Does everything automatically (still safe)

Tip for beginners: stay in normal mode. It always asks first, so nothing happens without your OK.


💬 Handy Commands

Type these at any time:

Command What it does
/help Shows the list of commands
/mode Shows or changes how cautious Sage is
/model Shows the AI models you can pick from
/image <words> Creates a picture from your description
/attach <file> Adds an image so Sage can look at it
/reset Starts a fresh conversation
/exit Closes Sage

🤖 Choosing a Different AI Model

Different AI models are better at different things. To see and switch:

› /model

You'll see a numbered list. To switch, type /model and a number:

› /model 2

Sage remembers your choice for next time.

Note: The models Sage can use depend on what your API endpoint provides. If you pick a model your endpoint doesn't offer, Sage will let you know.


🔒 Is It Safe?

Yes — Sage is built to be careful:

  • It only works inside the folder you open it in. It can't touch the rest of your computer.
  • It asks before making changes (in normal mode).
  • It refuses dangerous commands (like deleting everything).
  • Before editing a file, it shows you exactly what will change.

📤 Sharing Sage With Someone Else

Want to give Sage to a colleague? There's no package to build or file to send — Sage is published on PyPI, so anyone can install it the same way you did.

Just point them at this same guide, or tell them:

  1. Install pipx (if they don't have it):
    python -m pip install --user pipx
    python -m pipx ensurepath
    
  2. Open a new terminal, then install Sage:
    pipx install litellm-sage
    
  3. Run sage — it walks them through entering their own endpoint URL and API key (see "Your Settings" above), then starts right up.

Each person uses their own API key in their own .env file, so nobody's credentials are shared.

(They'll need Python 3.10+ installed, plus their own endpoint URL and API key.)


❓ Something Went Wrong?

"sage is not recognized" Close the terminal and open a brand-new one — pipx needs a fresh terminal to pick up the PATH changes. If it still doesn't work, run python -m pipx ensurepath again, then open another new terminal.

"Connection error" Sage can't reach your endpoint. Check that the BASE_URL in your settings is correct and that you can reach the server (e.g. you're on the right network or VPN if it's a private endpoint).

"[config error] Missing API_KEY" or "Missing BASE_URL" This usually means your settings file exists but is missing a value — edit C:\Users\<your-name>\.sage\.env and fill in API_KEY and BASE_URL (see "Your Settings" above). If the file doesn't exist at all, just run sage to get the guided setup.

"team not allowed to access model" / model error The selected model isn't available on your endpoint. Use /model to pick a different one.

Sage can't see my image Make sure the image is in the same folder you opened Sage in, and that you typed /attach with the correct file name first. Also make sure your chosen model supports image analysis.


📝 Quick Reference Card

Install:           pipx install litellm-sage
First run:         sage   (walks you through setup — endpoint URL + API key)
Start Sage:        sage
Ask something:     just type it and press Enter
Look at an image:  /attach picture.png   then ask about it
Make an image:     /image a sunset over the ocean
Change caution:    /mode plan | normal | auto-edits | auto
Pick a model:      /model   then   /model <number>
Get help:          /help
Quit:              /exit

Enjoy using Sage! 🧙✨

Metadata

Release files for litellm-sage 0.1.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for litellm-sage 0.1.0
File Size Uploaded
litellm_sage-0.1.0.tar.gz 33.3 kB Details

Built distribution (wheel)

Table of built distributions (wheels) for litellm-sage 0.1.0
File Interpreter ABI Platform
litellm_sage-0.1.0-py3-none-any.whl Python 3 none any Details

Total release size: 69.8 kB

Release files / litellm_sage-0.1.0.tar.gz

Download URL litellm_sage-0.1.0.tar.gz
Size 33.3 kB
Tags Source
SHA-256 checksum
How to use checksums
170cd50cf8a86bbc9d930b40488f903175cf00feb0abd12373c4eafdba86bacc
BLAKE2b-256 checksum
How to use checksums
b9a8b91b7b7526223b6ceb13df425784469aff0f727feffd7d11cceac2f7daee
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/6.1.0 CPython/3.13.12

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Jul 18, 2026.

Transparency log

Release files / litellm_sage-0.1.0-py3-none-any.whl

Download URL litellm_sage-0.1.0-py3-none-any.whl
Size 36.5 kB
Tags Python 3
SHA-256 checksum
How to use checksums
0c8f56f9238645559e5a582f294decda9252651c24d2075356342ee9c2c0a298
BLAKE2b-256 checksum
How to use checksums
f21bebfef030f40bed9733bb2b21f34f5833c2c49de0aa881237662405151789
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/6.1.0 CPython/3.13.12

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Jul 18, 2026.

Transparency log

Release history Release notifications | RSS feed

This release

0.1.0 This release

2 release files

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page