Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/TrinaxCode/TrinaxAI/llms.txt

Use this file to discover all available pages before exploring further.

TrinaxAI runs entirely on your Mac — no cloud, no subscriptions, no data leaving your network. The installer detects whether you’re on Apple Silicon (M1/M2/M3/M4) or Intel, auto-sizes a hardware profile based on your RAM, writes .env, installs dependencies via Homebrew, builds the PWA frontend, and registers a LaunchAgent in ~/Library/LaunchAgents/ for autostart. Apple Silicon gets Metal-accelerated inference through Ollama automatically; Intel Macs fall back to CPU. When the install completes you’ll have:
  • Ollama running at http://localhost:11434
  • RAG API (FastAPI) at https://localhost:3333
  • PWA chat interface at https://localhost:3334

​
One-command install

Paste this into Terminal. The script clones the repo to ~/trinaxai if it doesn’t already exist, then runs the guided installer: 
curl -fsSL https://raw.githubusercontent.com/TrinaxCode/TrinaxAI/main/install.sh | bash
Security tip: Review the script before running it.
curl -fsSL https://raw.githubusercontent.com/TrinaxCode/TrinaxAI/main/install.sh | less
macOS may also block scripts downloaded from the internet via Gatekeeper. Cloning the repository and running locally avoids any quarantine issues — see the manual steps below.

​
Manual install

1

Install Xcode Command Line Tools

Required by Homebrew and for building native Python/Node modules:
xcode-select --install
A system dialog will appear. Click Install and wait for it to finish.
2

Install Homebrew

If you don’t have Homebrew yet:
/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"
On Apple Silicon, follow the post-install instructions to add Homebrew to your PATH (usually eval "$(/opt/homebrew/bin/brew shellenv)").
3

Install dependencies

brew install python@3.12 node git curl ollama
You can also install Ollama from the official macOS app and keep it running in the menu bar instead of using the Homebrew formula.
4

Clone the repository

git clone https://github.com/TrinaxCode/TrinaxAI.git ~/trinaxai
cd ~/trinaxai
5

Create the Python virtual environment

python3 -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
python -m pip install -r requirements.txt
pip install -e .
6

Install and build the PWA

cd chat-pwa
npm install
npm run build
cd ..
7

Run the installer

bash install.sh
The installer creates .env, generates a local HTTPS certificate (and attempts to add it to your login keychain automatically), pulls any requested Ollama models, and optionally enables LaunchAgent autostart.

​
Installer flags

Pass any of these flags to install.sh to skip prompts or override defaults:
FlagDescription
--interactiveGuided install; prompts for optional choices (default)
--non-interactiveFully automatic install — no prompts. Useful for CI/scripts
--no-modelsSkip downloading all Ollama models (including vision)
--no-visionSkip vision model download only
--no-autostartDo not enable boot auto-start
--no-startDo not start TrinaxAI at the end of the install
--profile 8gb|16gb|max|ultraOverride the auto-detected hardware profile
--lan-systemEnable LAN system-control endpoints and generate an admin token
Profile reference
ProfileRAM targetNotes
8gb≤ 8 GBllama3.2:1b, qwen2.5-coder:1.5b, nomic-embed-text
16gb9–19 GBllama3.2:3b, qwen2.5-coder:3b, bge-m3
max20–31 GBqwen2.5-coder:7b, larger context window
ultra≥ 32 GBqwen2.5-coder:14b, qwen2.5vl:7b — ideal for M2/M3 Max/Ultra
Example: CI install with no prompts: 
bash install.sh --non-interactive --no-models --profile 16gb

​
Gatekeeper and quarantine

macOS quarantines files downloaded from the internet. If you see “cannot be opened because it is from an unidentified developer” or a quarantine error, clear the attribute: 
xattr -d com.apple.quarantine install.sh
Cloning with git clone does not trigger quarantine — only files downloaded via a browser or curl -O are affected.

​
Trusting the local HTTPS certificate

TrinaxAI generates a self-signed certificate for localhost and your LAN IP. The installer tries to add it to your login keychain automatically via security add-trusted-cert. If it doesn’t auto-trust:
  1. Open Keychain Access → select the login keychain.
  2. Import ~/trinaxai/chat-pwa/certs/trinaxai-local.crt.
  3. Double-click the certificate → expand Trust → set “When using this certificate” to Always Trust.
  4. Close and enter your password to save.
After trusting, Safari and Chrome will open https://localhost:3334 without a warning.
On iPhone and iPad, navigate to https://[YOUR-LAN-IP]:3334 in Safari, tap the warning, and choose Visit Website. To permanently trust the cert, AirDrop the .crt file to your device and install it via Settings → General → VPN & Device Management, then enable full trust under Settings → General → About → Certificate Trust Settings.

​
Service management

​
Start and stop

# Start all services
./startup_ai.sh

# Shut down only AI (RAG + Ollama); PWA stays available
./shutdown_ai.sh

# Shut down everything
.venv/bin/python service_manager.py stop-all --base-dir "$PWD"

# Check status
.venv/bin/python service_manager.py status --base-dir "$PWD"
You can also use the CLI shorthand: 
trinaxai start
trinaxai stop

​
LaunchAgent autostart

TrinaxAI registers a LaunchAgent in ~/Library/LaunchAgents/com.trinaxcode.trinaxai.plist. Manage it with: 
# Enable autostart
.venv/bin/python service_manager.py enable-autostart --base-dir "$PWD"

# Disable autostart
.venv/bin/python service_manager.py disable-autostart --base-dir "$PWD"

# Verify the agent is loaded
launchctl list | grep trinax
Tail service logs: 
tail -f logs/supervisor.log
tail -f logs/rag_api.log
tail -f logs/frontend.log
The supervisor keeps the PWA available at all times. If you shut down the AI with ./shutdown_ai.sh or from the PWA settings, the next boot will not restart Ollama/RAG — it respects the state you left it in.

​
Update

Run the guided updater from the repository root. It asks whether to back up first, pull the latest code, update models, change autostart, and restart services. Python and npm dependencies always update automatically: 
cd ~/trinaxai
./update.sh
Common non-interactive flags: 
./update.sh --non-interactive    # Full update, no prompts, safe defaults
./update.sh --no-backup          # Skip the pre-update backup
./update.sh --models             # Also pull/update configured Ollama models
./update.sh --restart            # Restart services after update

​
Uninstall

The guided uninstaller stops services, removes the LaunchAgent, and asks which generated/runtime files to remove. Your source code is always kept: 
cd ~/trinaxai
./uninstall.sh
To also remove Ollama models during uninstall: 
./uninstall.sh --remove-models
RAG index data (storage/) and uploaded files (local_sources/) are kept by default. Pass --remove-data to delete them.

​
Port reference

PortServiceNotes
3333RAG API (FastAPI)Backend — binds to 0.0.0.0 for LAN access
3334PWA (chat interface)Open https://localhost:3334 in your browser
11434OllamaBinds to 127.0.0.1 by default
Accessing from a phone or tablet: open https://[YOUR-LAN-IP]:3334 from any device on the same Wi-Fi. Find your Mac’s LAN IP: 
ipconfig getifaddr en0    # Wi-Fi on most Macs
ipconfig getifaddr en1    # Ethernet or secondary interface
Do not expose ports 3333, 3334, or 11434 to the internet. For remote access, use a VPN such as Tailscale or WireGuard. See SECURITY.md for the full threat model.
LAN system control is disabled by default. Sensitive endpoints require either a localhost caller or a valid TRINAXAI_ADMIN_TOKEN. Enable it during install with --lan-system, or set TRINAXAI_ALLOW_LAN_SYSTEM=1 and TRINAXAI_ADMIN_TOKEN in .env manually.

​
macOS folder access permissions

When you first index a folder such as Documents, Desktop, or Downloads, macOS will show a permission prompt. Click Allow to let TrinaxAI read those locations. You can review and change these permissions at any time in System Settings → Privacy & Security → Files and Folders.

Build docs developers (and LLMs) love

Get started for freeTalk to us