Instructions to use flowxai/sentinel-gate with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use flowxai/sentinel-gate with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="flowxai/sentinel-gate")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("flowxai/sentinel-gate")
model = AutoModelForCausalLM.from_pretrained("flowxai/sentinel-gate", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

MLX

How to use flowxai/sentinel-gate 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("flowxai/sentinel-gate")

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)

llama-cpp-python

How to use flowxai/sentinel-gate with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="flowxai/sentinel-gate",
	filename="gguf/sentinel-gate-4b-Q4_K_M.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use flowxai/sentinel-gate 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 flowxai/sentinel-gate:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf flowxai/sentinel-gate:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf flowxai/sentinel-gate:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf flowxai/sentinel-gate: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 flowxai/sentinel-gate:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf flowxai/sentinel-gate: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 flowxai/sentinel-gate:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf flowxai/sentinel-gate:Q4_K_M

Use Docker

docker model run hf.co/flowxai/sentinel-gate:Q4_K_M

LM Studio
Jan

vLLM

How to use flowxai/sentinel-gate with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "flowxai/sentinel-gate"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "flowxai/sentinel-gate",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/flowxai/sentinel-gate:Q4_K_M

SGLang

How to use flowxai/sentinel-gate with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "flowxai/sentinel-gate" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "flowxai/sentinel-gate",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "flowxai/sentinel-gate" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "flowxai/sentinel-gate",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Ollama
How to use flowxai/sentinel-gate with Ollama:
```
ollama run hf.co/flowxai/sentinel-gate:Q4_K_M
```

Unsloth Studio

How to use flowxai/sentinel-gate with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for flowxai/sentinel-gate to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for flowxai/sentinel-gate to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for flowxai/sentinel-gate to start chatting

How to use flowxai/sentinel-gate with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "flowxai/sentinel-gate"

Configure the model in Pi

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

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use flowxai/sentinel-gate 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 "flowxai/sentinel-gate"

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 flowxai/sentinel-gate

Run Hermes

hermes

Atomic Chat new

OpenClaw new

How to use flowxai/sentinel-gate with OpenClaw:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "flowxai/sentinel-gate"

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 "flowxai/sentinel-gate" \
  --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"

MLX LM

How to use flowxai/sentinel-gate with MLX LM:

Generate or start a chat session

# Install MLX LM
uv tool install mlx-lm
# Interactive chat REPL
mlx_lm.chat --model "flowxai/sentinel-gate"

Run an OpenAI-compatible server

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

Docker Model Runner
How to use flowxai/sentinel-gate with Docker Model Runner:
```
docker model run hf.co/flowxai/sentinel-gate:Q4_K_M
```

Lemonade

How to use flowxai/sentinel-gate with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull flowxai/sentinel-gate:Q4_K_M

Run and chat with the model

lemonade run user.sentinel-gate-Q4_K_M

List all available models

lemonade list

bogdanraduta commited on 20 days ago

Commit

9c8e219

verified ·

1 Parent(s): 6ebd961

Add fp16 merged model (Qwen3-4B LoRA)

Browse files

Files changed (1) hide show

README.md +4 -82

README.md CHANGED Viewed

@@ -1,87 +1,9 @@
 ---
 license: apache-2.0
-language:
-  - en
-base_model: Qwen/Qwen3-4B
-library_name: transformers
 pipeline_tag: text-generation
 tags:
-  - lora
-  - regulatory
-  - compliance
-  - escalation
-  - decision-gate
-  - mlx
-  - gguf
-  - flowx
 ---
-# FlowX Sentinel Gate (4B)
-**FlowX Sentinel Gate** is an escalation decision gate for regulated workflows: given a
-complete case (domain facts + the applicable policy schema), it decides **ESCALATE** (route
-to a human) vs **DECIDE** (safe to automate), and for escalations returns a **category**, a
-structured rationale, a **calibrated confidence**, the required human action, and an audit
-trail. It is a LoRA fine-tune of **Qwen3-4B**, built by FlowX.AI to sit after the
-[Semantic Mapper](https://huggingface.co/flowxai/semantic-mapper) in a compliance pipeline.
-> Decision-support tool, **not legal advice**. It gates automation; a human owns escalated
-> cases. Deploy with the confidence threshold your risk posture requires.
-## What it does
-Input: a case JSON (facts + `policy_schema`) + `Decide: ESCALATE or DECIDE?`. Output: an
-oracle decision. For ESCALATE: `action`, `escalation_category` (one of six), the
-category-specific block, `confidence_score`, `confidence_reasoning`, `human_action_required`,
-`audit_trail`. For DECIDE: the safe-to-automate rationale + confidence + audit trail.
-**Six escalation categories:** MISSING_REQUIRED_DOCUMENTATION, POLICY_VIOLATION,
-BOUNDARY_CONDITION, INSUFFICIENT_CONFIDENCE, CONFLICTING_SIGNALS, EXTERNAL_DEPENDENCY.
-## Evaluation (held-out, n=71: 46 escalate / 17 decide)
-| Metric | Result | Note |
-|---|---|---|
-| **False-negative rate** | **0.000** (0/46) | never misses an escalation &mdash; the safety-critical metric |
-| **Action accuracy** (ESCALATE vs DECIDE) | **1.000** | |
-| **False-positive rate** | **0.000** (0/17) | never over-escalates |
-| JSON validity | 0.89 | structured output; pair with the deterministic repair step |
-| Category accuracy (on true-escalate) | 0.61 | the human-routing label; see limitations |
-The decision itself (automate vs escalate) is the gate's job, and it is **perfect on this
-held-out set** &mdash; no missed escalations, no over-escalation, action decided correctly
-every time. The `escalation_category` is a secondary routing hint and is right ~61% of the
-time (categories legitimately overlap for some cases).
-## Files & formats
-| Path | Format | Runs on |
-|---|---|---|
-| `/` (root) | fp16 safetensors | CUDA / servers, vLLM |
-| `mlx-int4/`, `mlx-int8/` | MLX quantized | Apple Silicon |
-| `gguf/*.gguf` | GGUF Q8_0 / Q4_K_M | CUDA + CPU (llama.cpp / Ollama) |
-System prompt: `You are an escalation gate for regulated decisions.\nDetermine: ESCALATE or DECIDE? Output ONLY JSON.`
-(decode with `enable_thinking=False`).
-## Training
-LoRA (rank 32 / scale 16 / dropout 0.05), Qwen3-4B base, MLX-LM, cosine LR 5e-5&rarr;5e-6,
-max sequence length 2048. Data: **471 realistic-synthetic escalation cases** (400 train / 71
-held-out), balanced ~35% DECIDE / 65% ESCALATE across the six categories and four regulated
-domains (banking, insurance, logistics, labor), each grounded in a real regulatory citation.
-## Limitations & responsible use
-- **Not legal advice** &mdash; an automation gate; escalated cases must be handled by a human.
-- **Category label is ~61% accurate** &mdash; use the ESCALATE/DECIDE decision (perfect on
-  held-out) as the gate; treat the category as a routing suggestion, not ground truth.
-- JSON validity is 0.89 &mdash; deploy with the deterministic repair/retry step.
-- Evaluated on 71 synthetic cases; validate on your own case distribution before production.
-## License & attribution
-Apache-2.0. Copyright 2026 FlowX.AI. See `NOTICE`. Base model: Qwen3-4B (Apache-2.0).
-Escalation scenarios are realistic synthetic, grounded in real regulatory citations.
-_Author: Bogdan Răduță, Head of Research, FlowX.AI._

 ---
+library_name: mlx
 license: apache-2.0
+license_link: https://huggingface.co/Qwen/Qwen3-4B/blob/main/LICENSE
 pipeline_tag: text-generation
+base_model: mlx-community/Qwen3-4B-4bit
 tags:
+- mlx
 ---