Instructions to use GenbaLabs/super-decider-mlx with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use GenbaLabs/super-decider-mlx with MLX:

# Make sure mlx-lm is installed
# pip install --upgrade mlx-lm

# Generate text with mlx-lm
from mlx_lm import load, generate

model, tokenizer = load("GenbaLabs/super-decider-mlx")

prompt = "Write a story about Einstein"
messages = [{"role": "user", "content": prompt}]
prompt = tokenizer.apply_chat_template(
    messages, add_generation_prompt=True
)

text = generate(model, tokenizer, prompt=prompt, verbose=True)

Notebooks
Google Colab
Kaggle
Local Apps Settings
LM Studio

How to use GenbaLabs/super-decider-mlx with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "GenbaLabs/super-decider-mlx"

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "mlx-lm": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "GenbaLabs/super-decider-mlx"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

MLX LM

How to use GenbaLabs/super-decider-mlx with MLX LM:

Generate or start a chat session

# Install MLX LM
uv tool install mlx-lm
# Interactive chat REPL
mlx_lm.chat --model "GenbaLabs/super-decider-mlx"

Run an OpenAI-compatible server

# Install MLX LM
uv tool install mlx-lm
# Start the server
mlx_lm.server --model "GenbaLabs/super-decider-mlx"
# Calling the OpenAI-compatible server with curl
curl -X POST "http://localhost:8000/v1/chat/completions" \
   -H "Content-Type: application/json" \
   --data '{
     "model": "GenbaLabs/super-decider-mlx",
     "messages": [
       {"role": "user", "content": "Hello"}
     ]
   }'

Hermes Agent

How to use GenbaLabs/super-decider-mlx with Hermes Agent:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "GenbaLabs/super-decider-mlx"

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 GenbaLabs/super-decider-mlx

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use GenbaLabs/super-decider-mlx with OpenClaw:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "GenbaLabs/super-decider-mlx"

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 "GenbaLabs/super-decider-mlx" \
  --custom-provider-id mlx-lm \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

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

super-decider-mlx / runtime /run.sh

GenbaLabs

card, MLX package and runtime

88a6753 verified 1 day ago

Raw History Blame Contribute Delete

1.62 kB

#!/bin/bash
# Serve this checkpoint on a Mac: the MLX engine on :8000, then the structured decision
# server on :8011. The DJEV_* defaults are the published deployment's, so answers match the
# hosted API's.
#
#   MODEL=$(pwd) ./runtime/run.sh        # from a downloaded snapshot
set -euo pipefail
cd "$(dirname "$0")/.."
PY=${PY:-python3}
MODEL=${MODEL:-$(pwd)}
CANVAS=${CANVAS:-128}
export DJEV_DOC_FIRST=${DJEV_DOC_FIRST:-1} DJEV_BIDIR=${DJEV_BIDIR:-qsep} DJEV_BIDIR_MAXPROMPT=${DJEV_BIDIR_MAXPROMPT:-8192}
export DJEV_NOUL_COMMIT=${DJEV_NOUL_COMMIT:-0.8} DJEV_TEMPS=${DJEV_TEMPS:-score=0.375}
export DJEV_ADAPTIVE_THINK=${DJEV_ADAPTIVE_THINK:-choice=0.7,score=0.7,noul=0.85,budget=512,min_tokens=0,max_inflight=2,max_questions=2}
export DJEV_THINK_STEPS=${DJEV_THINK_STEPS:-8} DJEV_MANY_OPTIONS=${DJEV_MANY_OPTIONS:-chunk=20,top=3,think1=0,think2=0}
export DEFAULT_SAMPLES=${DEFAULT_SAMPLES:-1} SERVER_EXTRA_ARGS=${SERVER_EXTRA_ARGS:---constrained --engine-samples}

# Memory floor: the weights are ~19 GB active and an 8k-token prefill peaks around 21 GB.
$PY -m superdecider.server --model "$MODEL" --canvas "$CANVAS" --port 8000 --max-prompt "${MAX_PROMPT:-12288}" &
ENGINE_PID=$!
trap 'kill $ENGINE_PID 2>/dev/null || true' EXIT
until curl -sf 127.0.0.1:8000/v1/models >/dev/null; do
  kill -0 $ENGINE_PID 2>/dev/null || { echo "engine exited"; exit 1; }
  sleep 1
done
echo "engine ready on :8000"

exec $PY runtime/structured_server.py --upstream http://127.0.0.1:8000 --model super-decider \
  --tokenizer "$MODEL" --canvas "$CANVAS" --host "${HOST:-127.0.0.1}" --port "${PORT:-8011}" $SERVER_EXTRA_ARGS

	#!/bin/bash
	# Serve this checkpoint on a Mac: the MLX engine on :8000, then the structured decision
	# server on :8011. The DJEV_* defaults are the published deployment's, so answers match the
	# hosted API's.
	#
	# MODEL=$(pwd) ./runtime/run.sh # from a downloaded snapshot
	set -euo pipefail
	cd "$(dirname "$0")/.."
	PY=${PY:-python3}
	MODEL=${MODEL:-$(pwd)}
	CANVAS=${CANVAS:-128}
	export DJEV_DOC_FIRST=${DJEV_DOC_FIRST:-1} DJEV_BIDIR=${DJEV_BIDIR:-qsep} DJEV_BIDIR_MAXPROMPT=${DJEV_BIDIR_MAXPROMPT:-8192}
	export DJEV_NOUL_COMMIT=${DJEV_NOUL_COMMIT:-0.8} DJEV_TEMPS=${DJEV_TEMPS:-score=0.375}
	export DJEV_ADAPTIVE_THINK=${DJEV_ADAPTIVE_THINK:-choice=0.7,score=0.7,noul=0.85,budget=512,min_tokens=0,max_inflight=2,max_questions=2}
	export DJEV_THINK_STEPS=${DJEV_THINK_STEPS:-8} DJEV_MANY_OPTIONS=${DJEV_MANY_OPTIONS:-chunk=20,top=3,think1=0,think2=0}
	export DEFAULT_SAMPLES=${DEFAULT_SAMPLES:-1} SERVER_EXTRA_ARGS=${SERVER_EXTRA_ARGS:---constrained --engine-samples}

	# Memory floor: the weights are ~19 GB active and an 8k-token prefill peaks around 21 GB.
	$PY -m superdecider.server --model "$MODEL" --canvas "$CANVAS" --port 8000 --max-prompt "${MAX_PROMPT:-12288}" &
	ENGINE_PID=$!
	trap 'kill $ENGINE_PID 2>/dev/null \|\| true' EXIT
	until curl -sf 127.0.0.1:8000/v1/models >/dev/null; do
	kill -0 $ENGINE_PID 2>/dev/null \|\| { echo "engine exited"; exit 1; }
	sleep 1
	done
	echo "engine ready on :8000"

	exec $PY runtime/structured_server.py --upstream http://127.0.0.1:8000 --model super-decider \
	--tokenizer "$MODEL" --canvas "$CANVAS" --host "${HOST:-127.0.0.1}" --port "${PORT:-8011}" $SERVER_EXTRA_ARGS