Instructions to use TitleOS/ADBait-1B-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use TitleOS/ADBait-1B-GGUF with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="TitleOS/ADBait-1B-GGUF",
	filename="adbait-1b-fp16.gguf",
)

llm.create_chat_completion(
	messages = "No input example has been defined for this model task."
)

Notebooks
Google Colab
Kaggle
Local Apps

llama.cpp

How to use TitleOS/ADBait-1B-GGUF with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf TitleOS/ADBait-1B-GGUF:Q8_0
# Run inference directly in the terminal:
llama-cli -hf TitleOS/ADBait-1B-GGUF:Q8_0

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf TitleOS/ADBait-1B-GGUF:Q8_0
# Run inference directly in the terminal:
llama-cli -hf TitleOS/ADBait-1B-GGUF:Q8_0

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 TitleOS/ADBait-1B-GGUF:Q8_0
# Run inference directly in the terminal:
./llama-cli -hf TitleOS/ADBait-1B-GGUF:Q8_0

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 TitleOS/ADBait-1B-GGUF:Q8_0
# Run inference directly in the terminal:
./build/bin/llama-cli -hf TitleOS/ADBait-1B-GGUF:Q8_0

Use Docker

docker model run hf.co/TitleOS/ADBait-1B-GGUF:Q8_0

LM Studio
Jan
Ollama
How to use TitleOS/ADBait-1B-GGUF with Ollama:
```
ollama run hf.co/TitleOS/ADBait-1B-GGUF:Q8_0
```

Unsloth Studio

How to use TitleOS/ADBait-1B-GGUF 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 TitleOS/ADBait-1B-GGUF 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 TitleOS/ADBait-1B-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for TitleOS/ADBait-1B-GGUF to start chatting

Docker Model Runner
How to use TitleOS/ADBait-1B-GGUF with Docker Model Runner:
```
docker model run hf.co/TitleOS/ADBait-1B-GGUF:Q8_0
```

Lemonade

How to use TitleOS/ADBait-1B-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull TitleOS/ADBait-1B-GGUF:Q8_0

Run and chat with the model

lemonade run user.ADBait-1B-GGUF-Q8_0

List all available models

lemonade list

ADBait: Dynamic Android ADB Honeypot

ADBait is a fine-tuned language model designed to act as the backend brain for a dynamic Android Debug Bridge (ADB) honeypot. Built on top of ibm-granite/granite-4.0-h-1b, this model is trained to generate highly convincing, context-aware Android 14 shell environments to trap, delay, and analyze automated exploitation scripts and human attackers.

Rather than relying on static, hardcoded directory trees or predictable regex responses, ADBait generates realistic synthetic terminal outputs on the fly. Attackers will find outputs to be filled with accurate, dynamic data, such as popular installed user apps.

Model Details

Architecture: FP16 merged model converted to FP16 & Q8_0 GGUFs.
Base Model: ibm-granite/granite-4.0-h-1b
Target Environment: Android 14 (API Level 34) shell simulation.
Format: GGUFs

Training Data

This model was fine-tuned exclusively against the TitleOS/ADB-CursedHoneycomb dataset. The dataset consists of 250 curated, synthetic ChatML conversational structures mapping standard, aggressive, and exploratory ADB commands to their corresponding Android 14 outputs generated by Gemini 3.1 Flash-Lite.

Hardware & Fine-tuning

Training was executed on a single NVIDIA Tesla P40 (24GB VRAM). Due to the Pascal architecture's hardware constraints, the training pipeline utilized QLoRA in 4-bit precision with a strict float16 compute type to bypass the lack of bfloat16 and Flash Attention support. The FP16 & Q8_0 merged GGUFs are available here.

Intended Use

ADBait is meant for security researchers and network administrators. Deploy this model behind a socket listener handling the ADB protocol handshake. Once the handshake completes, pipe the incoming shell commands directly into the model as user prompts, and return the model's generation to the socket as the shell output.

License

See license.md for the modified Mozilla Public License 2.0 license this model is provided under.

Disclaimer: This is a honeypot tool. Do not use this model as a source of factual Android documentation, as it is explicitly trained to hallucinate directory structures, process lists, and system variables to deceive attackers. Do not give ADBait access to real data.