Instructions to use BrainboxAI/cyber-analyst-4B-verifier-pilot with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use BrainboxAI/cyber-analyst-4B-verifier-pilot with Transformers:

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

pipe = pipeline("text-generation", model="BrainboxAI/cyber-analyst-4B-verifier-pilot")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("BrainboxAI/cyber-analyst-4B-verifier-pilot", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use BrainboxAI/cyber-analyst-4B-verifier-pilot 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 BrainboxAI/cyber-analyst-4B-verifier-pilot:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf BrainboxAI/cyber-analyst-4B-verifier-pilot:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf BrainboxAI/cyber-analyst-4B-verifier-pilot:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf BrainboxAI/cyber-analyst-4B-verifier-pilot: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 BrainboxAI/cyber-analyst-4B-verifier-pilot:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf BrainboxAI/cyber-analyst-4B-verifier-pilot: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 BrainboxAI/cyber-analyst-4B-verifier-pilot:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf BrainboxAI/cyber-analyst-4B-verifier-pilot:Q4_K_M

Use Docker

docker model run hf.co/BrainboxAI/cyber-analyst-4B-verifier-pilot:Q4_K_M

LM Studio
Jan

vLLM

How to use BrainboxAI/cyber-analyst-4B-verifier-pilot with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "BrainboxAI/cyber-analyst-4B-verifier-pilot"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "BrainboxAI/cyber-analyst-4B-verifier-pilot",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/BrainboxAI/cyber-analyst-4B-verifier-pilot:Q4_K_M

SGLang

How to use BrainboxAI/cyber-analyst-4B-verifier-pilot 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 "BrainboxAI/cyber-analyst-4B-verifier-pilot" \
    --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": "BrainboxAI/cyber-analyst-4B-verifier-pilot",
		"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 "BrainboxAI/cyber-analyst-4B-verifier-pilot" \
        --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": "BrainboxAI/cyber-analyst-4B-verifier-pilot",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Ollama
How to use BrainboxAI/cyber-analyst-4B-verifier-pilot with Ollama:
```
ollama run hf.co/BrainboxAI/cyber-analyst-4B-verifier-pilot:Q4_K_M
```

Unsloth Studio

How to use BrainboxAI/cyber-analyst-4B-verifier-pilot 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 BrainboxAI/cyber-analyst-4B-verifier-pilot 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 BrainboxAI/cyber-analyst-4B-verifier-pilot to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for BrainboxAI/cyber-analyst-4B-verifier-pilot to start chatting

How to use BrainboxAI/cyber-analyst-4B-verifier-pilot with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf BrainboxAI/cyber-analyst-4B-verifier-pilot:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "BrainboxAI/cyber-analyst-4B-verifier-pilot:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use BrainboxAI/cyber-analyst-4B-verifier-pilot with Docker Model Runner:
```
docker model run hf.co/BrainboxAI/cyber-analyst-4B-verifier-pilot:Q4_K_M
```

Lemonade

How to use BrainboxAI/cyber-analyst-4B-verifier-pilot with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull BrainboxAI/cyber-analyst-4B-verifier-pilot:Q4_K_M

Run and chat with the model

lemonade run user.cyber-analyst-4B-verifier-pilot-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use BrainboxAI/cyber-analyst-4B-verifier-pilot with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf BrainboxAI/cyber-analyst-4B-verifier-pilot: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 BrainboxAI/cyber-analyst-4B-verifier-pilot:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use BrainboxAI/cyber-analyst-4B-verifier-pilot with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf BrainboxAI/cyber-analyst-4B-verifier-pilot: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 "BrainboxAI/cyber-analyst-4B-verifier-pilot: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"

BrainboxAI commited on 14 days ago

Commit

d984dcb

verified ·

1 Parent(s): c10f896

Upload README.md with huggingface_hub

Browse files

Files changed (1) hide show

README.md +81 -0

README.md ADDED Viewed

	@@ -0,0 +1,81 @@

+---
+language:
+  - he
+  - en
+license: apache-2.0
+base_model: BrainboxAI/cyber-analyst-4B-safetensors
+tags:
+  - cybersecurity
+  - security
+  - vulnerability-verification
+  - finding-verifier
+  - gemma4
+  - unsloth
+  - lora
+  - pilot
+pipeline_tag: text-generation
+pretty_name: Cyber-Analyst 4B — Verifier Pilot
+---
+# Cyber-Analyst 4B — Verifier Pilot
+> ⚠️ **PILOT / experimental.** A small continued-fine-tune on top of
+> [`cyber-analyst-4B`](https://huggingface.co/BrainboxAI/cyber-analyst-4B) to
+> validate a training recipe before scaling. **Not** a replacement for the
+> production model.
+## What this is
+A LoRA delta applied to **`BrainboxAI/cyber-analyst-4B`** that teaches two behaviors:
+1. **Strict finding-verifier** — given the evidence for a suspected web vulnerability
+   (tool, URL, payload, response snippet, PoC), it outputs a JSON verdict:
+   `{"verdict":"true_positive|false_positive|inconclusive","reason":"..."}`.
+   The point is to REJECT unproven findings (500-only ≠ SQLi, HTML-encoded XSS ≠ XSS,
+   SSRF that never left the network, IDOR that returned only the caller's own data,
+   properly-signed JWT, …), not rubber-stamp them.
+2. **Bilingual identity** — knows it is *Cyber-Analyst 4B by BrainboxAI*.
+## Lineage
+```
+Google Gemma-4 E4B  →  cyber-analyst-4B (1.27M security examples, BrainboxAI)
+                     →  THIS pilot (+ verifier + identity delta)
+```
+## Training
+- **Base:** `BrainboxAI/cyber-analyst-4B-safetensors` (weights preserved — this only adds a delta).
+- **Method:** bf16 LoRA, r=16 / α=32, 3 epochs, LR 1e-4, cosine, seq 2048 (Unsloth).
+- **Data:** ~100 balanced verifier examples (true/false/inconclusive, adversarial hard
+  negatives) + a small % bilingual identity. Held-out eval.
+- **Result:** eval_loss ≈ 1.13. Pilot scale — meant to validate the recipe, then scale to ~600.
+## Files in this repo
+| Path | What | Use |
+|---|---|---|
+| `gguf/…Q4_K_M.gguf` | quantized model (~5.3GB) | **LM Studio / Ollama / llama.cpp** |
+| `gguf/…BF16-mmproj.gguf` | vision projector | only for image input (Gemma-4 is multimodal) |
+| `lora/` | LoRA adapter (~147MB) | apply on the base to reproduce / continue |
+| `merged/` | merged 16-bit safetensors | **trainable base for the next round** |
+## Use in LM Studio
+Download `gguf/cyber-analyst-4B-safetensors.Q4_K_M.gguf` and load it. For the verifier
+task, use a system prompt like:
+> You are a STRICT security-finding VERIFIER. Given the evidence, decide true_positive /
+> false_positive / inconclusive. Judge from the EVIDENCE, not the title. Respond with ONLY
+> JSON: {"verdict":"…","reason":"…"}.
+## Limitations
+- **Pilot scale (~100 examples)** — validates format + basic discrimination, not a
+  production verifier. Expect overfit gaps; scale the dataset (~600+) before relying on it.
+- Inherits the base model's limitations (public-data bias, temporal cutoff).
+- Not for autonomous security decisions without human review.
+## Author
+Built by [**Netanel Elyasi**](https://huggingface.co/BrainboxAI), founder of BrainboxAI.