setup_model.sh

#!/bin/bash
set -e

# Default URL (Nanbeige4.1-3B-f32-GGUF - Q8_0)
# DEFAULT_URL="https://huggingface.co/prithivMLmods/Nanbeige4.1-3B-f32-GGUF/resolve/main/Nanbeige4.1-3B.Q8_0.gguf"
DEFAULT_URL="https://huggingface.co/microsoft/Phi-3-mini-4k-instruct-gguf/blob/main/Phi-3-mini-4k-instruct-q4.gguf"
MODEL_URL=${1:-$DEFAULT_URL}
MODEL_DIR="model"
ENGINE_DIR="engine"

echo ">>> Setting up Ultra-Fast LLM..."

# 1. Setup Model Directory
if [ ! -d "$MODEL_DIR" ]; then
    mkdir -p "$MODEL_DIR"
fi

# 2. Download Model
if [ ! -f "$MODEL_DIR/model.gguf" ]; then
    echo ">>> Downloading model from $MODEL_URL..."
    wget -O "$MODEL_DIR/model.gguf" "$MODEL_URL" --show-progress
else
    echo ">>> Model already exists using existing..."
fi

# 3. Create Template (Phi-3 default)
if [ ! -f "$MODEL_DIR/template.txt" ]; then
    echo "<|user|>\n{{prompt}}<|end|>\n<|assistant|>" > "$MODEL_DIR/template.txt"
    echo ">>> Created default Phi-3 template."
fi

# 4. Check for Virtual Environment
if [ ! -d ".venv" ]; then
    echo ">>> Creating Python virtual environment..."
    python3 -m venv .venv
    source .venv/bin/activate
    pip install -r requirements.txt
else
    source .venv/bin/activate
fi

# 5. Compile Engine
echo ">>> Compiling C++ Optimization Engine..."
if [ -d "$ENGINE_DIR" ]; then
    cd "$ENGINE_DIR"
    g++ -O2 -march=native -shared -fPIC -o libbatch.so batch_server.cpp \
      -I"../.venv/lib/python3.12/site-packages/include" \
      -L"../.venv/lib/python3.12/site-packages/llama_cpp/lib" \
      -lllama -Wl,-rpath,"../.venv/lib/python3.12/site-packages/llama_cpp/lib"
    cd ..
else
    echo "Error: engine directory not found!"
    exit 1
fi

echo ">>> Setup Complete!"
echo "Run server with: cd api && ../.venv/bin/uvicorn main:app --host 0.0.0.0 --port 8000"