A fine-tuned Qwen 2.5 1.5B model, quantized to Q4_K_M (GGUF), purpose-built for MiniPet — an animated desktop pet that lives on your screen and helps you interact with the SUI blockchain.

Key Features

100% Offline — Runs locally via llama-server sidecar. No API keys, no internet required.
SUI Blockchain Tool Calling — Trained to invoke structured tools: transfer SUI, swap tokens, check balances, bonk pets, send gifts, manage whitelists.
Multilingual — Responds in Vietnamese, English, French, Chinese, Italian, and Korean based on user settings.
Personality — Short, cute, and helpful responses matching a virtual pet companion character.
Pomodoro & Productivity — Can set timers and manage focus sessions via natural language.

Model Details

Property	Value
Base Model	Qwen/Qwen2.5-1.5B-Instruct
Fine-tuning	SUI blockchain tools, pet personality, multilingual responses
Quantization	Q4_K_M (4-bit, GGUF)
File Size	~986 MB
Context Length	2048 tokens
Architecture	Qwen2 (2B params before quantization)

Intended Use

This model is designed exclusively as the embedded AI brain for MiniPet desktop application:

Desktop Pet Chat — Double-click the pet to chat. The model responds with short, personality-driven answers.
On-Chain Actions — When users request blockchain operations, the model outputs structured <tool_call> JSON that the app intercepts and executes via zkLogin.
Context-Aware Assistance — Knows the current time, active app, and user's wallet state.

Supported Tools

Tool	Description
`transfer_sui`	Send SUI to an address or saved contact
`check_wallet_balance`	Query wallet balance on SUI testnet
`swap_sui_to_usdc`	Swap SUI tokens to USDC
`set_pomodoro_timer`	Start/stop focus timer
`bonk_pet`	Interact with on-chain PetNFT (bonk)
`send_pet_gift`	Send a gift to another pet
`rename_pet`	Rename the user's PetNFT
`check_pet_stats`	View pet level and stats
`start_auto_trade` / `stop_auto_trade`	Control simulated auto-trading
`add_fast_transfer_wallet`	Add contact to whitelist

Usage with llama.cpp

bash

# Start local server (used by MiniPet app internally)
llama-server -m qwen-sui-q4_k_m.gguf --port 8080 -c 2048

# Or use the HuggingFace shorthand
llama-server -hf iamquocbao/minipet-qwen-model-SUI:Q4_K_M

Then query via OpenAI-compatible API at http://127.0.0.1:8080/v1/chat/completions.

Usage with MiniPet App

The model is automatically downloaded on first launch. No manual setup needed.

~/Library/Application Support/com.minipet.app/minipet-qwen-model-SUI.gguf

Limitations

Not a general-purpose LLM — Fine-tuned specifically for short pet assistant responses and SUI tool calling. May produce lower quality results for general knowledge tasks.
Simulated Trading — Auto-trade tools are simulated (paper mode) and do not execute real on-chain swaps until mainnet launch.
Context Window — Limited to 2048 tokens. Long conversations may lose earlier context.

License

Apache 2.0 — same as the base Qwen 2.5 model.

iamquocbao/minipet-qwen-model-SUI

MiniPet Qwen SUI — On-Device AI Pet Assistant 🐾🧠