train.py

import pandas as pd
import pickle
from sklearn.feature_extraction.text import TfidfVectorizer

# 1. Load your 5000 samples
print("👻 Loading Rosetta Stone Dataset...")
try:
    df = pd.read_csv("rosetta_code_dataset.csv")
    print(f"   -> Loaded {len(df)} examples.")
except:
    print("Error: Could not find rosetta_code_dataset.csv")
    exit()

# 2. Train the Brain (TF-IDF Vectorizer)
# This converts English text ("fibonacci in java") into Math Numbers
print("🧠 Training the Ghost Engine...")
vectorizer = TfidfVectorizer()
tfidf_matrix = vectorizer.fit_transform(df['prompt'].values.astype('U'))

# 3. Save the Brain file
# We save the Vectorizer (translator), Matrix (memory), and Code (answers)
output_file = "ghost_brain.pkl"
with open(output_file, "wb") as f:
    pickle.dump((vectorizer, tfidf_matrix, df['code'].values), f)

print(f"✅ SUCCESS! Brain saved as '{output_file}'")
print(f"   Size: {os.path.getsize(output_file) / 1024:.2f} KB (Tiny!)")
print("   Copy this file + ghost_coder.py to your USB stick.")