Installation

Install

Prebuilt Binaries

Prebuilt binaries are available here

Using Go Install
go install github.com/binuud/avatar/cmd/avatar@latest

mkdir dataVolume
export FILE_STORE_PATH=`pwd`/dataVolume
export SERVABLE_DIRECTORY=`pwd`/dataVolume/static
avatar

(Tested and working on version 0.1.1)

Using Docker

The ui displays system metrics for CPU, RAM and GPU usage. These stats will not be visible, when run via docker.

Pull the docker image and start the container. The docker image has a very small footprint.

docker pull binuud/avatar:latest

mkdir dataVolume
docker run --rm -v ./dataVolume:/dataVolume \
  -e FILE_STORE_PATH=/dataVolume \
  -p 9090:9090 -p 9080:9080 binuud/avatar:latest

(Tested and working on version 0.1.0)

Access UI from Browser

http://localhost:9080
Build from source

Git Repo For Avatar

git clone https://github.com/binuud/avatar.git
cd avatar
make vendor
make build
make run

(Tested and working on version 0.1.1)

Supported Environment Variables
  • LLM_HOST - The LLM Host, the default value connects to port 11234 (mlx serve)
  • MQTT_HOST - Not required in this release
  • LLM_MODEL - The default LLM Model
  • IMAGE_MODEL - The default Image model
  • VISION_MODEL - The default vision model
  • FILE_STORE_PATH - The path where the memories, avatars, generated assets are stored
  • SERVABLE_DIRECTORY - The path from which static files can be served

Sample data for supported Environment variables

export LLM_HOST="http://localhost:11234"
export MQTT_HOST="tcp://localhost:1883"

export LLM_MODEL="lmstudio-community/Qwen3.5-9B-MLX-4bit"
export IMAGE_MODEL="Runpod/FLUX.2-klein-4B-mflux-4bit"
export VISION_MODEL="lmstudio-community/Qwen3.5-9B-MLX-4bit"

export FILE_STORE_PATH=$PWD/dataVolume
export SERVABLE_DIRECTORY=$PWD/dataVolume/static