Installation

Get Chalie running in minutes — install, pick a model, and go. Works with Ollama, OpenAI, Anthropic, and Gemini.

Install

One command installs Chalie:

curl -fsSL https://chalie.ai/install | bash

It checks your dependencies, downloads the latest release, and installs the chalie command. The whole thing takes about two minutes.

Requirements

  • Python 3.11 or newer, already installed — the installer checks for it, but never installs it for you
  • macOS (Apple Silicon), or Linux (amd64 or arm64) on Debian/Ubuntu (apt) or Fedora/RHEL/CentOS (dnf)
  • About 3 GB of disk, and around 2 GB of memory once Chalie is running
  • Root or sudo on Linux — for the system libraries Chromium needs, and for the chalie command itself. macOS needs neither.

Intel Macs aren’t supported — the ONNX runtime that powers voice and memory search no longer ships Intel builds. On Windows, run from source or use the Docker image instead.

On a Linux distribution outside those two package managers, the installer stops and says so rather than leaving you a half-working instance: Chromium’s system libraries would never be installed and the browser wouldn’t start. The Docker image runs anywhere.

Everything Chalie creates — your account, data, models, and configuration — lives in ~/.chalie/. The only thing written outside it is the chalie command itself, which goes to /usr/local/bin on Linux and ~/.local/bin on macOS.

CLI

Once installed, manage Chalie with the chalie command:

chalie               # Start → http://localhost:31025
chalie --port=9000   # Start on a custom port
chalie --host=127.0.0.1   # Bind to a specific address
chalie stop          # Stop
chalie restart       # Restart
chalie status        # Show whether Chalie is running
chalie update        # Update to the latest release
chalie logs          # Follow the log

Chalie runs in the background and listens on port 31025 by default.

First run

  1. Start Chalie with chalie.
  2. Open http://localhost:31025/on-boarding/ and create your account. Pick a strong password here — it protects your data and conversations.
  3. Log in. You land straight in the chat interface.
  4. Give Chalie a brain: open Brain → Settings → Providers → Add Provider and connect a model (see below).

That’s the whole setup. Chalie generates its own secrets and stores all configuration through the web UI — there’s nothing to edit by hand.

Choose your model

Chalie is model-agnostic. Add your provider in Brain → Settings → Providers → Add Provider — a short wizard that only asks for what each provider needs.

Free, private, and runs entirely on your machine. Nothing leaves your computer. Assumes you already have Ollama running with a model pulled.

In Add Provider, pick Ollama. Host: http://localhost:11434 — no API key needed. Choose any model you’ve already pulled; larger models reason better.

OpenAI

Cloud provider. Requires an API key; usage charges apply.

  1. Get a key at platform.openai.com.
  2. In Add Provider, pick OpenAI and paste your key.

Anthropic

Cloud provider. Requires an API key; usage charges apply.

  1. Get a key at console.anthropic.com.
  2. In Add Provider, pick Anthropic and paste your key.

Google Gemini

Cloud provider. Requires an API key; a free tier is available.

  1. Get a key at aistudio.google.com.
  2. In Add Provider, pick Gemini and paste your key.
Provider Where to get it Notes
Ollama ollama.ai Local, free, private. Recommended. No key.
OpenAI platform.openai.com Cloud. Key required.
Anthropic console.anthropic.com Cloud. Key required.
Gemini aistudio.google.com Cloud. Key required. Free tier available.

Configuration

There’s almost nothing to configure. The only runtime option is the port:

chalie --port=9000   # default: 31025

Voice ships built in — the on-device speech models install alongside Chalie, so there’s nothing to turn on. If a voice install ever misbehaves, reinstalling Chalie is the fix.

Before going public

Chalie is built to run on your own machine. If you want to reach it from outside your network, secure it first:

  • Enable HTTPS. Put a reverse proxy (nginx, Caddy, or Traefik) in front of Chalie to terminate TLS.
  • Use a strong password. Set during onboarding — this is what guards your data and conversations.
  • Restrict CORS. Lock the allowed origin down before exposing Chalie to any network.

Docker

Prefer containers? Pull the published image:

docker run -d -p 31025:31025 -v ~/.chalie/data:/root/.chalie/app/data chalieai/chalie

The image is published to Docker Hub as chalieai/chalie for linux/amd64 and linux/arm64 with every release.

Run from source

git clone https://github.com/chalie-ai/chalie.git && cd chalie
./run.sh             # → http://localhost:31025

run.sh resolves a Python interpreter and syncs dependencies for you before launching. Pass --port or --host just like the CLI.

This gets the app running, but voice models and the browser tool are fetched once, up front, by the full installer (or the Docker build) — not by run.sh. Use the one-command install above for a complete setup, or run the installer script from a source checkout if you need those too.

Update & uninstall

Update to the latest release at any time:

chalie update

This stops Chalie, pulls the newest release, and reinstalls its dependencies. Your data in ~/.chalie/ is never touched.

Each release keeps its own data file, so updating never rewrites what’s already there — your data carries forward automatically the first time you run the new version. Chalie keeps the last three releases’ data files as restore points and cleans up older ones on its own, so rolling back to an earlier release picks up exactly where that version left off.

To remove Chalie completely:

chalie stop
rm -rf ~/.chalie
sudo rm -f /usr/local/bin/chalie   # macOS: rm -f ~/.local/bin/chalie

On macOS the installer adds a PATH line to your ~/.zshrc (or ~/.bash_profile) — remove that too.