Instructions to use Illaitar/aidnd-arbiter with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use Illaitar/aidnd-arbiter with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf Illaitar/aidnd-arbiter:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf Illaitar/aidnd-arbiter:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf Illaitar/aidnd-arbiter:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf Illaitar/aidnd-arbiter:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf Illaitar/aidnd-arbiter:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf Illaitar/aidnd-arbiter:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf Illaitar/aidnd-arbiter:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf Illaitar/aidnd-arbiter:Q4_K_M

Use Docker

docker model run hf.co/Illaitar/aidnd-arbiter:Q4_K_M

Llama new
LM Studio
Jan
Ollama
How to use Illaitar/aidnd-arbiter with Ollama:
```
ollama run hf.co/Illaitar/aidnd-arbiter:Q4_K_M
```
Unsloth Desktop

How to use Illaitar/aidnd-arbiter with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Illaitar/aidnd-arbiter:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "Illaitar/aidnd-arbiter:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use Illaitar/aidnd-arbiter with Docker Model Runner:
```
docker model run hf.co/Illaitar/aidnd-arbiter:Q4_K_M
```

Lemonade

How to use Illaitar/aidnd-arbiter with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull Illaitar/aidnd-arbiter:Q4_K_M

Run and chat with the model

lemonade run user.aidnd-arbiter-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use Illaitar/aidnd-arbiter with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Illaitar/aidnd-arbiter:Q4_K_M

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default Illaitar/aidnd-arbiter:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use Illaitar/aidnd-arbiter with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Illaitar/aidnd-arbiter:Q4_K_M

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "Illaitar/aidnd-arbiter:Q4_K_M" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

aidnd-arbiter — арбитр свободных действий (Qwen3.5-9B + LoRA)

Решает, как разрешать freeform-действие игрока в движке AI-DnD. На вход — действие, краткая сцена и оценка правдоподобия 0..1. На выход — один JSON:

{"resolution": "auto_success|auto_fail|roll", "ability": …, "skill": …, "dc": …,
 "target": …, "lasting_effect": …, "reason": …}

auto_success — тривиальное, без риска;
auto_fail — невозможное;
roll — рискованное: навык 5e + DC (ниже DC — выше правдоподобие).

Как получился

База: unsloth/Qwen3.5-9B (Apache-2.0).
Метод: QLoRA SFT (4-bit), 170 примеров, 3 эпохи; учили только ответ ассистента. Затем мердж в базу и квантизация в Q4_K_M GGUF.
Оценка (30 отложенных, по 10 на тип resolution; метрика — совпадение с эталоном, DC с допуском ±2):

resolution skill (roll) dc±2 (roll) full

база Qwen3.5-9B 10% 0% 0% 10%

aidnd-arbiter 83% 80% 100% 77%

База почти не держит схему арбитра; адаптер — обязателен. DC калибруется отлично (±2 в 100% бросков). Часть «промахов» — спорные (невозможное действие как очень сложный бросок DC≈22 вместо auto_fail), а не грубые ошибки.

	resolution	skill (roll)	dc±2 (roll)	full
база Qwen3.5-9B	10%	0%	0%	10%
aidnd-arbiter	83%	80%	100%	77%

Запуск локально через Ollama

hf download Illaitar/aidnd-arbiter aidnd-arbiter-q4_k_m.gguf Modelfile --local-dir aidnd-arbiter
cd aidnd-arbiter && ollama create aidnd-arbiter -f Modelfile
ollama run aidnd-arbiter

Требуется Ollama 0.17.1+ (Modelfile использует RENDERER/PARSER qwen3.5).

В движке AI-DnD роль arbiter (decide_resolution) указывает на aidnd-arbiter (AIDND_ARBITER_MODEL, с откатом на базовую модель, если адаптера нет на сервере).