CloseAI Support

Help, requirements, and frequently asked questions

Server Requirements

CloseAI connects to a server you own. The server must meet these requirements before setup:

Requirement	Value
Operating System	Ubuntu 24.04 LTS (x86_64)
RAM	8 GB minimum (16 GB recommended)
Free Disk Space	30 GB minimum
SSH Access	A user with sudo privileges
Outbound Internet	Required during install for model download (~2–5 GB)
Inbound Port	443 (HTTPS) — opened automatically by the installer

Any standard VPS from providers such as DigitalOcean, Linode, Hetzner, or Vultr will work. Bare-metal servers and home servers are also supported.

Getting Started

Choose one of two setup methods:

Option A — Password Setup

1. Tap Add Server and choose Password Setup.
2. Enter your server’s IP address or hostname, SSH port (default: 22), username, and password.
3. Tap Connect. Verify the server’s host key fingerprint, then tap Trust & Continue.
4. The app runs a preflight check. All items must pass before install can begin.
5. Select an initial AI model and tap Install.
6. After installation, the app automatically sets up SSH key authentication so your server password is no longer needed.

Option B — Key Setup (no password needed)

1. Tap Add Server and choose Key Setup.
2. The app generates an SSH key pair on your device and displays the public key.
3. Copy the public key and add it to your server (paste into your VPS provider’s dashboard or append to ~/.ssh/authorized_keys).
4. Enter your server’s hostname and username. The app connects using the key.
5. Verify the fingerprint, pass preflight checks, pick a model, and install.

If CloseAI is already installed on your server, the app detects this and reconnects automatically without reinstalling.

AI Models

Five models are available from the app’s Server tab:

• Llama 3.2 (3B) — ~2 GB. Fast and versatile. Best starting point for most users.
• Phi-4 Mini — ~2.5 GB. Strong reasoning in a compact size (Microsoft Research).
• Gemma 3 (4B) — ~3 GB. Highest output quality (Google DeepMind).
• Qwen 2.5 Coder (7B) — ~4.7 GB. Specialized for code generation and debugging (Alibaba).
• DeepSeek-R1 (7B) — ~4.7 GB. Chain-of-thought reasoning, strong at math and logic.

To switch the active model, tap the model name in the chat toolbar. You can switch models mid-conversation. Install or remove models from the Server tab.

Advanced users can connect any Ollama-compatible model using the direct Ollama URL shown in the Server tab (http://<your-server>:11434).

Frequently Asked Questions

The preflight check fails with “Insufficient RAM” or “Insufficient disk”.

Your server does not meet the minimum requirements. Upgrade your VPS plan or free up disk space before retrying.

The app gets stuck on “Waiting for service to start”.

The install is running on your server and may take a few minutes depending on your server’s internet speed. The install continues even if you lock your phone. Open the app again after a few minutes to check progress. If it is still stuck after 10 minutes, check that port 443 is not blocked by a firewall rule in your VPS control panel.

Chat returns an error or shows “Connection failed”.

Open the Server tab and check the health indicator. If the server is shown as offline, make sure the server is running and that port 443 is reachable. If the status shows a red API key error, the app will attempt automatic credential recovery — tap Retry if prompted.

I locked my phone during install and the app shows a blank progress screen when I return.

The install continues on the server in the background even when the app is suspended. Pull down to refresh the progress screen or tap the back button and re-enter the setup flow to reconnect to the running install.

Can I use the AI from a computer or another app?

Yes. Ollama’s API runs on port 11434 of your server with no authentication. Tools like Open WebUI and Continue.dev can connect to http://<your-server>:11434 directly. The URL is shown in the Server tab.

How do I completely remove CloseAI from my server?

Open the Server tab, scroll to the bottom, and tap Uninstall Server. This removes the CloseAI backend, Ollama, all downloaded models, and every file written during install. Your server is returned to its original state.

Key setup fails — the app says it can’t connect.

Make sure the public key was pasted correctly into your VPS provider’s dashboard or ~/.ssh/authorized_keys. The key must be on a single line with no extra spaces or line breaks. Also verify the username matches the account the key was added to.

How do I search my conversation history?

Type in the search bar at the top of the conversations list. The app searches across all conversations and all messages. Tap any result to jump directly to that message in its conversation.

How do I upgrade the server stack?

When a new version is available, an “Upgrade Available” banner appears in the Server tab. Tap Upgrade to update with one tap. The upgrade streams progress in real time. If it fails, your server stays on the current version — nothing is broken.

The upgrade asks for a sudo password. Why?

Server upgrades require root privileges. If you set up with a password, the app uses it automatically. If you use key-based auth, the app will prompt for a sudo password. You can optionally save it to Keychain with Face ID / Touch ID protection so you won’t be asked again.

Is my data backed up to the cloud?

No. Chat history and server settings are stored locally on your iPhone. Your SSH password, SSH keys, and API key are stored in the iOS Keychain and are not synced to iCloud. See our Privacy Policy for full details.

Which iOS version is required?

CloseAI requires iOS 17 or later.