Tater is a local-first AI platform that can run local models through llama.cpp, or connect to remote OpenAI-compatible and llama.cpp APIs. It supports voice satellites like VoicePE, Sat1, S3Box, and ReSpeaker XVF3800, plus portals for Discord, Home Assistant, HomeKit, IRC, macOS, Matrix, Meshtastic, Telegram, and XBMC4Xbox.

---

Little Spud Companion App

Little Spud connects to your Tater Spud Hub for chat, TTS, STT, and notifications from your iOS or Android device.

---

Tater Bench

Tater Bench compares local models using repeatable Tater-style accuracy and performance tests across engines, speculative decoding modes, and hardware.

---

🧩 Tater Architecture

Tater is built around a modular system:

• Cores → core systems that extend Tater's capabilities
• Portals → integrations with platforms like Discord, Home Assistant, and more
• Verbas → AI-driven tools and actions Tater can perform
• Integrations → modular provider packages for devices, services, search providers, and external APIs

Tater Shop

Tater Shop manages the catalogs, versions, metadata, and updates for Tater Cores, Portals, and Verbas.

Core-owned Redis keys must use the Core's canonical namespace: <core_id>: or <module_key>:. Standard <module_key>_settings and running-state keys are handled separately. The Delete data uninstall option removes those namespaces automatically. Shared platform namespaces such as tater: are not inferred from a downloaded Core; historical official exceptions are explicitly audited in tateros/core_store.py.

Tater Integrations

Tater Integrations provides modular packages for devices, services, search providers, and external APIs.

---

Installation

Choose the installation that matches your system:

• macOS app, Unraid, and Home Assistant provide guided installation options.
• Docker is the easiest choice for most other Linux servers.
• Local installation is best when you want direct access to the hardware and Python environment.

Model note: Tater is designed around local models and local OpenAI-compatible servers such as Ollama, LM Studio, LocalAI, and Lemonade.

macOS App Installation

1. Download the latest macOS installer

   Download the latest Tater for macOS

2. Install Tater

   Open the DMG, then drag Tater.app into Applications.

3. Launch Tater

   Open Tater from Applications. On first launch, the app prepares its private runtime under:

   ~/.taterassistant/
   

   The app stores its managed Python runtime, virtual environment, runtime settings, logs, updates, and agent_lab data there. It does not use this source checkout's .venv, .runtime, or agent_lab folders. The app also includes the pinned AirPlay sender and receiver, and its startup environment check repairs any missing Python-side AirPlay support automatically.

4. Finish setup in TaterOS

   The app listens on 0.0.0.0:8501 and opens 127.0.0.1:8501 in the native window. If Python 3.11 is not already available, the launcher downloads a standalone CPython 3.11 runtime into ~/.taterassistant/python/ and uses it to build the private venv.

Closing the window keeps Tater running in the menu bar. Use the menu bar item to reopen Tater, open it in a browser, stop, restart, show logs, check for updates, install available updates, or quit.

Once the WebUI is up, continue to Post-Install Setup below.

Unraid Installation

Tater is available in the Unraid Community Apps store.

You can install Tater directly from the Unraid App Store with a one-click template.

Unraid note:

• Add container path mappings for /app/agent_lab and /app/.runtime to persistent, preferably cache-backed storage.
• Also set TZ and map /etc/localtime plus /etc/timezone if you want local time inside the container.

Once the Unraid containers are installed and running, continue to Post-Install Setup below.

Home Assistant Installation

A dedicated Home Assistant add-on repository is available here:

https://github.com/TaterTotterson/hassio-addons-tater

Click the button below to add the repository to Home Assistant:

Once added, the Tater AI Assistant add-on will appear in the Home Assistant Add-on Store.

Install order:

1. Install Tater AI Assistant.
2. Configure your LLM settings in the Tater add-on.
3. Start Tater.

Once the add-ons are running, continue to Post-Install Setup below.

Reachy Mini Installation

The Reachy Tater Satellite app turns Reachy Mini Wireless into a native voice satellite for an existing Tater server. Microphone, speaker, local wake-word detection, user tracking, expressive motion, and optional vision snapshots remain on Reachy, while the robot connects to Tater through its authenticated native satellite connection.

Install path:

1. Install Reachy Tater Satellite from the Reachy Mini app store and start it.
2. Open the app settings from Reachy's web interface and enter the address of your running Tater server.
3. Create a satellite pairing code in Tater, enter it once in the Reachy settings, and save the connection.
4. Optionally enable Allow vision snapshots for the Reachy Vision Verba, then say Hey Reachy.

Docker Installation

Use this method on a Linux server with Docker installed.

1. Create persistent storage

Run these commands from the directory where you want to keep Tater's data:

mkdir -p tater-data/agent_lab tater-data/runtime

• agent_lab stores models, documents, logs, and Tater's internal Redis data.
• runtime stores settings and native satellite pairing credentials.

Keep both directories when you update or recreate the container. Fast local storage is recommended; avoid network-mounted or unusually slow storage for agent_lab.

2. Start Tater

docker pull ghcr.io/tatertotterson/tater:latest

docker run -d --name tater_webui \
  --restart unless-stopped \
  --network host \
  --cap-add NET_BIND_SERVICE \
  -e TZ=America/Chicago \
  -e HTMLUI_PORT=8501 \
  -v "$(pwd)/tater-data/agent_lab:/app/agent_lab" \
  -v "$(pwd)/tater-data/runtime:/app/.runtime" \
  ghcr.io/tatertotterson/tater:latest

Change TZ if needed. Tater uses host networking for WebUI, voice, discovery, and media services.

3. Open TaterOS

Open one of these addresses:

• http://localhost:8501 on the server
• http://<server-ip>:8501 from another device on the same network

If port 8501 is already in use, change HTMLUI_PORT in the command. Continue to Post-Install Setup after TaterOS opens.

NVIDIA Docker

The NVIDIA image is available for amd64 systems with an NVIDIA driver and NVIDIA Container Toolkit installed:

docker pull ghcr.io/tatertotterson/tater:nvidia

Use the Docker command above, add --gpus all, and replace the final image name with ghcr.io/tatertotterson/tater:nvidia.

Local Installation

Use this method on Linux or macOS when you want Tater to run directly on the host.

1. Download Tater

git clone https://github.com/TaterTotterson/Tater.git
cd Tater

2. Run setup

sh setup_tater.sh

The setup menu asks which runtime fits your system, creates .venv, installs dependencies, and saves the selected profile. Tater supports Python 3.11 through 3.13; on supported Linux systems, setup can install a private Python runtime when the system version is unsuitable.

Profile	Choose it for
CPU	Most Linux PCs and generic ARM systems
macOS	Apple Silicon Macs
NVIDIA	Linux PCs and servers with NVIDIA GPUs
AMD ROCm	Supported Linux systems with AMD GPUs or Ryzen AI
Jetson	NVIDIA Jetson systems
Thor	Jetson Thor systems
Edge	Pi-class or remote-only systems that connect to a Spud Hub

To skip the menu, pass the profile name directly—for example:

sh setup_tater.sh cpu

The Edge profile requires the operating system's redis-server package. If macOS setup reports missing build tools, install ffmpeg and cmake with Homebrew and rerun setup.

3. Start Tater

sh run_ui.sh

Tater listens on 0.0.0.0:8501 by default. Open http://localhost:8501, or use http://<computer-ip>:8501 from another device. To use a different port:

HTMLUI_PORT=8601 sh run_ui.sh

Continue to Post-Install Setup after TaterOS opens. Model downloads and voice acceleration are configured inside TaterOS under Settings -> Models and Settings -> Voice Pipeline.

---

Post-Install Setup

After Tater is running, open TaterOS and finish the first-run setup:

1. Configure your base model in Settings -> Models -> LLM / Vision:
   • choose OpenAI-Compatible API for a local server such as Ollama, LM Studio, LocalAI, Lemonade, or vLLM
   • choose llama.cpp GGUF to load a GGUF model through Tater's native llama.cpp engine
   • for built-in local models, download llama.cpp GGUF models from the Models tab first, then select a downloaded model in the LLM tab
   • for OpenAI-compatible providers, set the endpoint host/port and model name
2. Optional:
   • add more Base servers for round-robin regular AI calls
   • enable Beast Mode and set per-head model settings for Astraeus/Hermes

Hydra model settings are saved by TaterOS and used at runtime. Base, Terminal, Beast Mode routing, and Vision can each use llama.cpp or a configured remote provider.

Local Models

• Use the Models tab to download llama.cpp GGUF files from Hugging Face. The LLM tab lists downloaded models only.
• Model caches live under agent_lab/models/llm/ by default:
  • llama-cpp for GGUF models and matching mmproj*.gguf vision projectors
• The Hugging Face browser uses the token saved in Integration Manager -> Hugging Face for private/gated models and better Hub rate limits.
• llama.cpp uses the native llama-server engine built by setup. It uses GPU offload by default when the installed build supports it. Set TATER_LLAMA_CPP_N_GPU_LAYERS=0 for CPU-only or TATER_LLAMA_CPP_SERVER_BIN to point at a custom llama-server binary.

Vision

• Vision can use an OpenAI-compatible API, the loaded Base model, or a dedicated local vision model.
• If Base is already loaded and vision-capable, Tater reuses it instead of loading the same model twice.
• Dedicated vision models are managed separately from Base.

Advanced Notes

• Local context length is configured in Settings -> Models -> LLM / Vision.
• Thinking suppression is enabled by default for local providers when supported.
• run_ui.sh starts Uvicorn with --no-access-log to suppress per-request log spam. Shutdown waits at most eight seconds for long-lived WebSocket and event-stream connections before cancelling them, so a stale satellite or browser connection cannot block a restart. Set HTMLUI_GRACEFUL_SHUTDOWN_TIMEOUT_SECONDS to override that connection-drain deadline.

License and trademarks

This project is licensed under the GNU Affero General Public License v3.0.

The Tater name and logo are governed by the Tater Trademark Policy.