1.2. Quickstart: Local Player

This is the fastest path to a working chat session, including with Hugging Face-origin GGUF weights.

Build from source:

git clone https://github.com/ml-cab/juno.git && cd juno
mvn clean package -DskipTests

Download a GGUF (replace the URL with your chosen model):

cd juno/models
wget https://huggingface.co/.../tinyllama-1.1b-chat-v1.0.Q4_K_M.gguf

Run the local interactive console:

Linux / macOS:

./juno local --model-path models/tinyllama-1.1b-chat-v1.0.Q4_K_M.gguf

Windows:

juno.bat local --model-path models\tinyllama-1.1b-chat-v1.0.Q4_K_M.gguf

--model-path is relative to the Juno project directory. To run the REST API alongside the REPL, add --api-port 8080.

Training: ./juno lora --model-path ... on Linux/macOS, juno.bat lora --model-path ... on Windows. See LoRA fine-tuning.

Merging: ./juno merge (or juno.bat merge on Windows) bakes a trained .lora adapter into a new GGUF, so inference needs no sidecar adapter. See Merge mode.

See also¶