language_detector.py

# language_detector.py - FINAL WORKING VERSION (200 lines)
from langdetect import detect, DetectorFactory
import re

DetectorFactory.seed = 0

def detect_input_language(text):
    """Detect if text is Hindi, English or Mixed"""
    try:
        clean_text = re.sub(r'[^\w\s]', '', text)
        if not clean_text.strip():
            return 'mixed'
        
        lang = detect(clean_text)
        
        # Hindi detection
        hindi_chars = re.findall(r'[\u0900-\u097F]', text)
        if lang == 'hi' or hindi_chars:
            # Check if mixed with English
            english_chars = re.findall(r'[a-zA-Z]', text)
            if hindi_chars and english_chars:
                return 'mixed'
            return 'hindi'
        
        # English detection
        if lang == 'en':
            return 'english'
        
        return 'mixed'
    except:
        return 'mixed'

def get_system_prompt(lang_mode, username):
    """Generate system prompt based on language and intent"""
    
    # CORE RULES - COMMON FOR ALL
    core_rules = f"""
    ROLE: AumCore AI - Senior Coding Assistant
    USER: {username}
    
    CRITICAL RULES:
    1. CODE vs CHAT DECISION:
       - CODE WHEN: User says 'code', 'program', 'script', 'function', 'create', 'build'
       - CHAT WHEN: General conversation, greetings, knowledge questions
       - EXAMPLES: 
           * "google drive code" → RAW CODE
           * "hello how are you" → TEXT RESPONSE
           * "koi bhajan aata hai" → TEXT RESPONSE
    
    2. CODE FORMAT:
       - RAW PYTHON CODE ONLY
       - NO markdown blocks (```python```)
       - NO 'python' keyword in response
       - Example: "from google.colab import drive\\ndrive.mount('/content/gdrive')"
    
    3. ERROR HANDLING:
       - If user shows error, analyze and provide corrected code
       - Include brief explanation of fix
    
    4. CODE QUALITY:
       - Production-ready code
       - Error handling included
       - Proper structure
    """
    
    # LANGUAGE SPECIFIC STYLES
    styles = {
        'hindi': """
        STYLE: 100% Hindi (except code)
        EXAMPLES:
        - User: "नमस्ते, कोड बताओ" → RAW CODE
        - User: "क्या हाल है" → "सब ठीक है {username} भाई!"
        - User: "त्रुटि: x परिभाषित नहीं" → "x = 10\\ny = x\\nprint(y)"
        """,
        
        'english': """
        STYLE: 100% English (except code)
        EXAMPLES:
        - User: "hello, give code" → RAW CODE
        - User: "how are you" → "I'm good {username}!"
        - User: "error: x not defined" → "x = 10\\ny = x\\nprint(y)"
        """,
        
        'mixed': """
        STYLE: 60% English + 40% Hindi (natural blend)
        EXAMPLES:
        - User: "hi bhai, code de" → RAW CODE
        - User: "are yaar, kya haal hai" → "Sab badhiya hai {username} bhai!"
        - User: "error aaya: x not defined" → "x = 10\\ny = x\\nprint(y)"
        """
    }
    
    # COMBINE
    full_prompt = f"""{core_rules}
    
    {styles.get(lang_mode, styles['mixed'])}
    
    FINAL REMINDER: Be {username}'s helpful AI assistant.
    Provide accurate code for technical requests.
    Engage naturally in conversation.
    """
    
    return full_prompt.strip()

# SIMPLE CODE GENERATOR (Optional - can be expanded)
def generate_basic_code(task):
    """Generate basic code templates"""
    templates = {
        'web': """
from fastapi import FastAPI
import uvicorn

app = FastAPI()

@app.get("/")
def home():
    return {"message": "Hello from AumCore AI"}

if __name__ == "__main__":
    uvicorn.run(app, host="0.0.0.0", port=8000)
""",
        'data': """
import pandas as pd
import numpy as np

# Load data
df = pd.read_csv("data.csv")

# Basic analysis
print(f"Shape: {df.shape}")
print(f"Columns: {list(df.columns)}")
print(f"Summary:\\n{df.describe()}")
""",
        'drive': """
from google.colab import drive
drive.mount('/content/gdrive')
"""
    }
    
    task_lower = task.lower()
    if 'drive' in task_lower or 'mount' in task_lower:
        return templates['drive']
    elif 'web' in task_lower or 'app' in task_lower:
        return templates['web']
    elif 'data' in task_lower or 'analy' in task_lower:
        return templates['data']
    
    return templates['drive']  # Default

# Test function
if __name__ == "__main__":
    # Test detection
    tests = [
        "नमस्ते",
        "hello world",
        "hi bhai kya haal hai",
        "google drive mount code do"
    ]
    
    for test in tests:
        lang = detect_input_language(test)
        print(f"{test[:20]:20} -> {lang}")
    
    print("\\n✅ language_detector.py ready for AumCore AI")