AIKit ✨

AIKit is a comprehensive platform to quickly get started to host, deploy, build and fine-tune large language models (LLMs).

AIKit offers two main capabilities:

Inference: AIKit uses LocalAI, which supports a wide range of inference capabilities and formats. LocalAI provides a drop-in replacement REST API that is OpenAI API compatible, so you can use any OpenAI API compatible client, such as Kubectl AI, Chatbot-UI and many more, to send requests to open LLMs!
Fine-Tuning: AIKit offers an extensible fine-tuning interface. It supports Unsloth for fast, memory efficient, and easy fine-tuning experience.

👉 For full documentation, please see AIKit website!

Features

🐳 No GPU, Internet access or additional tools needed except for Docker!
🤏 Minimal image size, resulting in less vulnerabilities and smaller attack surface with a custom distroless-based image
🎵 Fine-tune support
🚀 Easy to use declarative configuration for inference and fine-tuning
✨ OpenAI API compatible to use with any OpenAI API compatible client
📸 Multi-modal model support
🖼️ Image generation support
🦙 Support for GGUF (llama), GPTQ or EXL2 (exllama2), and GGML (llama-ggml) and Mamba models
🚢 Kubernetes deployment ready
📦 Supports multiple models with a single image
🖥️ Supports AMD64 and ARM64 CPUs and GPU-accelerated inferencing with NVIDIA GPUs
🔐 Ensure supply chain security with SBOMs, Provenance attestations, and signed images
🌈 Supports air-gapped environments with self-hosted, local, or any remote container registries to store model images for inference on the edge.

Quick Start

You can get started with AIKit quickly on your local machine without a GPU!

docker run -d --rm -p 8080:8080 ghcr.io/sozercan/llama3.1:8b

After running this, navigate to http://localhost:8080/chat to access the WebUI!

API

AIKit provides an OpenAI API compatible endpoint, so you can use any OpenAI API compatible client to send requests to open LLMs!

curl http://localhost:8080/v1/chat/completions -H "Content-Type: application/json" -d '{
    "model": "llama-3.1-8b-instruct",
    "messages": [{"role": "user", "content": "explain kubernetes in a sentence"}]
  }'

Output should be similar to:

{
  // ...
    "model": "llama-3.1-8b-instruct",
    "choices": [
        {
            "index": 0,
            "finish_reason": "stop",
            "message": {
                "role": "assistant",
                "content": "Kubernetes is an open-source container orchestration system that automates the deployment, scaling, and management of applications and services, allowing developers to focus on writing code rather than managing infrastructure."
            }
        }
    ],
  // ...
}

That's it! 🎉 API is OpenAI compatible so this is a drop-in replacement for any OpenAI API compatible client.

Pre-made Models

AIKit comes with pre-made models that you can use out-of-the-box!

If it doesn't include a specific model, you can always create your own images, and host in a container registry of your choice!

CPU

[!NOTE] AIKit supports both AMD64 and ARM64 CPUs. You can run the same command on either architecture, and Docker will automatically pull the correct image for your CPU.

Depending on your CPU capabilities, AIKit will automatically select the most optimized instruction set.

Model	Optimization	Parameters	Command	Model Name	License
🦙 Llama 3.2	Instruct	1B	`docker run -d --rm -p 8080:8080 ghcr.io/sozercan/llama3.2:1b`	`llama-3.2-1b-instruct`	Llama
🦙 Llama 3.2	Instruct	3B	`docker run -d --rm -p 8080:8080 ghcr.io/sozercan/llama3.2:3b`	`llama-3.2-3b-instruct`	Llama
🦙 Llama 3.1	Instruct	8B	`docker run -d --rm -p 8080:8080 ghcr.io/sozercan/llama3.1:8b`	`llama-3.1-8b-instruct`	Llama
🦙 Llama 3.3	Instruct	70B	`docker run -d --rm -p 8080:8080 ghcr.io/sozercan/llama3.3:70b`	`llama-3.3-70b-instruct`	Llama
Ⓜ️ Mixtral	Instruct	8x7B	`docker run -d --rm -p 8080:8080 ghcr.io/sozercan/mixtral:8x7b`	`mixtral-8x7b-instruct`	Apache
🅿️ Phi 3.5	Instruct	3.8B	`docker run -d --rm -p 8080:8080 ghcr.io/sozercan/phi3.5:3.8b`	`phi-3.5-3.8b-instruct`	MIT
🔡 Gemma 2	Instruct	2B	`docker run -d --rm -p 8080:8080 ghcr.io/sozercan/gemma2:2b`	`gemma-2-2b-instruct`	Gemma
⌨️ Codestral 0.1	Code	22B	`docker run -d --rm -p 8080:8080 ghcr.io/sozercan/codestral:22b`	`codestral-22b`	MNLP
QwQ		32B	`docker run -d --rm -p 8080:8080 ghcr.io/sozercan/qwq:32b`	`qwq-32b-preview`	Apache 2.0

NVIDIA CUDA

[!NOTE] To enable GPU acceleration, please see GPU Acceleration.

Please note that only difference between CPU and GPU section is the --gpus all flag in the command to enable GPU acceleration.

Model	Optimization	Parameters	Command	Model Name	License
🦙 Llama 3.2	Instruct	1B	`docker run -d --rm --gpus all -p 8080:8080 ghcr.io/sozercan/llama3.2:1b`	`llama-3.2-1b-instruct`	Llama
🦙 Llama 3.2	Instruct	3B	`docker run -d --rm --gpus all -p 8080:8080 ghcr.io/sozercan/llama3.2:3b`	`llama-3.2-3b-instruct`	Llama
🦙 Llama 3.1	Instruct	8B	`docker run -d --rm --gpus all -p 8080:8080 ghcr.io/sozercan/llama3.1:8b`	`llama-3.1-8b-instruct`	Llama
🦙 Llama 3.3	Instruct	70B	`docker run -d --rm --gpus all -p 8080:8080 ghcr.io/sozercan/llama3.3:70b`	`llama-3.3-70b-instruct`	Llama
Ⓜ️ Mixtral	Instruct	8x7B	`docker run -d --rm --gpus all -p 8080:8080 ghcr.io/sozercan/mixtral:8x7b`	`mixtral-8x7b-instruct`	Apache
🅿️ Phi 3.5	Instruct	3.8B	`docker run -d --rm --gpus all -p 8080:8080 ghcr.io/sozercan/phi3.5:3.8b`	`phi-3.5-3.8b-instruct`	MIT
🔡 Gemma 2	Instruct	2B	`docker run -d --rm --gpus all -p 8080:8080 ghcr.io/sozercan/gemma2:2b`	`gemma-2-2b-instruct`	Gemma
⌨️ Codestral 0.1	Code	22B	`docker run -d --rm --gpus all -p 8080:8080 ghcr.io/sozercan/codestral:22b`	`codestral-22b`	MNLP
QwQ		32B	`docker run -d --rm --gpus all -p 8080:8080 ghcr.io/sozercan/qwq:32b`	`qwq-32b-preview`	Apache 2.0
📸 Flux 1 Dev	Text to image	12B	`docker run -d --rm --gpus all -p 8080:8080 ghcr.io/sozercan/flux1:dev`	`flux-1-dev`	FLUX.1 [dev] Non-Commercial License

Apple Silicon (experimental)

[!NOTE] To enable GPU acceleration on Apple Silicon, please see Podman Desktop documentation. For more information, please see GPU Acceleration.

Apple Silicon is an experimental runtime and it may change in the future. This runtime is specific to Apple Silicon only, and it will not work as expected on other architectures, including Intel Macs.

Only gguf models are supported on Apple Silicon.

Model	Optimization	Parameters	Command	Model Name	License
🦙 Llama 3.2	Instruct	1B	`podman run -d --rm --device /dev/dri -p 8080:8080 ghcr.io/sozercan/applesilicon/llama3.2:1b`	`llama-3.2-1b-instruct`	Llama
🦙 Llama 3.2	Instruct	3B	`podman run -d --rm --device /dev/dri -p 8080:8080 ghcr.io/sozercan/applesilicon/llama3.2:3b`	`llama-3.2-3b-instruct`	Llama
🦙 Llama 3.1	Instruct	8B	`podman run -d --rm --device /dev/dri -p 8080:8080 ghcr.io/sozercan/applesilicon/llama3.1:8b`	`llama-3.1-8b-instruct`	Llama
🅿️ Phi 3.5	Instruct	3.8B	`podman run -d --rm --device /dev/dri -p 8080:8080 ghcr.io/sozercan/applesilicon/phi3.5:3.8b`	`phi-3.5-3.8b-instruct`	MIT
🔡 Gemma 2	Instruct	2B	`podman run -d --rm --device /dev/dri -p 8080:8080 ghcr.io/sozercan/applesilicon/gemma2:2b`	`gemma-2-2b-instruct`	Gemma

What's next?

👉 For more information and how to fine tune models or create your own images, please see AIKit website!

# README