#!/usr/bin/env python3
"""
Farnsworth Model Downloader
Downloads and configures local LLMs for Farnsworth:
- Ollama models
- GGUF models for llama.cpp
- Embedding models
"""
import argparse
import os
import sys
import subprocess
import urllib.request
import hashlib
from pathlib import Path
from typing import Optional
# Model configurations
OLLAMA_MODELS = {
"deepseek-r1:1.5b": {
"description": "DeepSeek-R1 Distilled 1.5B - Best reasoning at size",
"size": "~1.2GB",
"recommended": True,
},
"qwen3:0.6b": {
"description": "Qwen3 0.6B - Ultra-lightweight",
"size": "~400MB",
"recommended": False,
},
"phi3:mini": {
"description": "Phi-3 Mini - GPT-3.5 class reasoning",
"size": "~2.4GB",
"recommended": False,
},
"llama3.2:1b": {
"description": "Llama 3.2 1B - Balanced performance",
"size": "~700MB",
"recommended": False,
},
}
EMBEDDING_MODELS = {
"all-MiniLM-L6-v2": {
"description": "Fast, lightweight embeddings",
"dimensions": 384,
"recommended": True,
},
"all-mpnet-base-v2": {
"description": "Higher quality embeddings",
"dimensions": 768,
"recommended": False,
},
}
def print_banner():
"""Print download banner."""
print("""
╔═══════════════════════════════════════════╗
║ Farnsworth Model Downloader ║
╚═══════════════════════════════════════════╝
""")
def check_ollama() -> bool:
"""Check if Ollama is installed."""
try:
result = subprocess.run(
["ollama", "--version"],
capture_output=True,
text=True,
)
return result.returncode == 0
except FileNotFoundError:
return False
def download_ollama_model(model_name: str) -> bool:
"""Download a model using Ollama."""
print(f"\n📦 Downloading {model_name}...")
try:
result = subprocess.run(
["ollama", "pull", model_name],
check=True,
)
print(f"✅ {model_name} downloaded successfully")
return True
except subprocess.CalledProcessError as e:
print(f"❌ Failed to download {model_name}: {e}")
return False
except FileNotFoundError:
print("❌ Ollama not found. Install from: https://ollama.ai")
return False
def download_embedding_model(model_name: str, data_dir: Path) -> bool:
"""Download embedding model using sentence-transformers."""
print(f"\n📦 Downloading embedding model: {model_name}...")
try:
from sentence_transformers import SentenceTransformer
# Download to cache
model = SentenceTransformer(model_name)
# Test it works
test_embedding = model.encode("test")
print(f"✅ {model_name} loaded (dimensions: {len(test_embedding)})")
return True
except ImportError:
print("❌ sentence-transformers not installed")
print(" Run: pip install sentence-transformers")
return False
except Exception as e:
print(f"❌ Failed to download {model_name}: {e}")
return False
def list_available_models():
"""List available models."""
print("\n📋 Available LLM Models (via Ollama):\n")
for name, info in OLLAMA_MODELS.items():
rec = " ⭐ RECOMMENDED" if info.get("recommended") else ""
print(f" {name}{rec}")
print(f" {info['description']}")
print(f" Size: {info['size']}\n")
print("\n📋 Available Embedding Models:\n")
for name, info in EMBEDDING_MODELS.items():
rec = " ⭐ RECOMMENDED" if info.get("recommended") else ""
print(f" {name}{rec}")
print(f" {info['description']}")
print(f" Dimensions: {info['dimensions']}\n")
def download_recommended(data_dir: Path) -> bool:
"""Download recommended models."""
print("\n🚀 Downloading recommended models...\n")
success = True
# Check Ollama
if check_ollama():
print("✅ Ollama detected")
# Download recommended Ollama model
for name, info in OLLAMA_MODELS.items():
if info.get("recommended"):
if not download_ollama_model(name):
success = False
break
else:
print("⚠️ Ollama not installed")
print(" For best performance, install Ollama: https://ollama.ai")
print(" Then run: ollama pull deepseek-r1:1.5b")
# Download embedding model
for name, info in EMBEDDING_MODELS.items():
if info.get("recommended"):
if not download_embedding_model(name, data_dir):
success = False
break
return success
def interactive_setup(data_dir: Path):
"""Interactive model selection."""
print("\n🎯 Interactive Model Setup\n")
# LLM selection
print("Select an LLM backend:\n")
print(" 1. Ollama (recommended - easy setup)")
print(" 2. llama.cpp (advanced - manual GGUF files)")
print(" 3. Skip LLM setup")
choice = input("\nChoice [1]: ").strip() or "1"
if choice == "1":
if not check_ollama():
print("\n⚠️ Ollama not installed.")
print("Install from: https://ollama.ai")
print("Then run this script again.")
else:
print("\nSelect a model:\n")
models = list(OLLAMA_MODELS.keys())
for i, name in enumerate(models, 1):
info = OLLAMA_MODELS[name]
rec = " ⭐" if info.get("recommended") else ""
print(f" {i}. {name}{rec} ({info['size']})")
model_choice = input(f"\nChoice [1]: ").strip() or "1"
try:
model_idx = int(model_choice) - 1
if 0 <= model_idx < len(models):
download_ollama_model(models[model_idx])
except ValueError:
print("Invalid choice")
elif choice == "2":
print("\n📝 llama.cpp Setup")
print("1. Download a GGUF model from Hugging Face")
print("2. Place it in:", data_dir / "models")
print("3. Update configs/models.yaml with the path")
# Embedding model
print("\n\nSelect an embedding model:\n")
models = list(EMBEDDING_MODELS.keys())
for i, name in enumerate(models, 1):
info = EMBEDDING_MODELS[name]
rec = " ⭐" if info.get("recommended") else ""
print(f" {i}. {name}{rec}")
emb_choice = input(f"\nChoice [1]: ").strip() or "1"
try:
emb_idx = int(emb_choice) - 1
if 0 <= emb_idx < len(models):
download_embedding_model(models[emb_idx], data_dir)
except ValueError:
print("Invalid choice")
def main():
"""Main entry point."""
parser = argparse.ArgumentParser(
description="Download models for Farnsworth"
)
parser.add_argument(
"--list", "-l",
action="store_true",
help="List available models"
)
parser.add_argument(
"--recommended", "-r",
action="store_true",
help="Download recommended models"
)
parser.add_argument(
"--interactive", "-i",
action="store_true",
help="Interactive model selection"
)
parser.add_argument(
"--ollama",
type=str,
help="Download specific Ollama model"
)
parser.add_argument(
"--embedding",
type=str,
help="Download specific embedding model"
)
parser.add_argument(
"--data-dir",
type=str,
default="./data",
help="Data directory"
)
args = parser.parse_args()
print_banner()
data_dir = Path(args.data_dir)
data_dir.mkdir(parents=True, exist_ok=True)
if args.list:
list_available_models()
elif args.recommended:
download_recommended(data_dir)
elif args.interactive:
interactive_setup(data_dir)
elif args.ollama:
download_ollama_model(args.ollama)
elif args.embedding:
download_embedding_model(args.embedding, data_dir)
else:
# Default: interactive setup
interactive_setup(data_dir)
print("\n✨ Model setup complete!")
print("Run 'python main.py' to start Farnsworth")
if __name__ == "__main__":
main()
