language_detector.py

# language_detector.py - UPDATED FOR CODE FORMATTING (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"""
    
    # AUMCORE AI SYSTEM KNOWLEDGE (CRITICAL - ADDED NEW)
    SYSTEM_KNOWLEDGE = f"""AUMCORE AI SYSTEM INFORMATION (NEVER FORGET):
    1. ENDPOINTS: /system/status, /system/task, /reset, /chat, /
    2. /system/status → AumCoreMaster orchestrator status check
    3. /system/task → Heavy computational tasks
    4. URL: https://aumcoreai-aumcore-ai.hf.space
    5. Version: 2.0.1-Stable, Logs: logs/aumcore_main.log
    6. When asked about system, ALWAYS mention these endpoints"""
    
    # CORE RULES - UPDATED FOR MARKDOWN FORMATTING
    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" → CODE IN MARKDOWN BLOCKS
           * "hello how are you" → TEXT RESPONSE
           * "koi bhajan aata hai" → TEXT RESPONSE
    
    2. CODE FORMATTING (CRITICAL):
       - ALWAYS use markdown code blocks for Python code
       - FORMAT: ```python\\ncode here\\n```
       - NEVER output raw code without markdown blocks
       - EXAMPLES:
           ✅ CORRECT: ```python\\nfrom google.colab import drive\\ndrive.mount('/content/gdrive')\\n```
           ❌ WRONG: from google.colab import drive
       
    3. ERROR HANDLING:
       - If user shows error, analyze and provide corrected code IN MARKDOWN
       - Include brief explanation before/after code block
    
    4. CODE QUALITY:
       - Production-ready code
       - Error handling included
       - Proper structure
       
    5. RESPONSE STRUCTURE:
       - Code responses: Explanation (if needed) + Markdown code block
       - Chat responses: Plain text only
       - Error fixes: Error analysis + Corrected code in markdown
       
    6. SYSTEM KNOWLEDGE: {SYSTEM_KNOWLEDGE}
    """
    
    # LANGUAGE SPECIFIC STYLES - UPDATED WITH SYSTEM EXAMPLES
    styles = {
        'hindi': f"""
        STYLE: 100% Hindi (code markdown blocks ke alawa)
        EXAMPLES:
        - User: "नमस्ते, कोड बताओ" 
          You: ```python\\nfrom google.colab import drive\\ndrive.mount('/content/gdrive')\\n```
        - User: "system status check करो" 
          You: "/system/status endpoint से AumCoreMaster orchestrator की status check करें"
        - User: "क्या हाल है" 
          You: "सब ठीक है {username} भाई!"
        - User: "endpoints बताओ" 
          You: "/system/status, /system/task, /reset, /chat, / - ये सभी endpoints हैं"
        """,
        
        'english': f"""
        STYLE: 100% English (except code in markdown blocks)
        EXAMPLES:
        - User: "hello, give code" 
          You: ```python\\nfrom google.colab import drive\\ndrive.mount('/content/gdrive')\\n```
        - User: "check system orchestrator status" 
          You: "Use /system/status endpoint to check AumCoreMaster orchestrator"
        - User: "how are you" 
          You: "I'm good {username}!"
        - User: "list endpoints" 
          You: "Endpoints: /system/status, /system/task, /reset, /chat, /"
        """,
        
        'mixed': f"""
        STYLE: 60% English + 40% Hindi (natural blend)
        EXAMPLES:
        - User: "hi bhai, code de" 
          You: ```python\\nfrom google.colab import drive\\ndrive.mount('/content/gdrive')\\n```
        - User: "system ki status bata" 
          You: "/system/status endpoint se system orchestrator ki status check karo"
        - User: "are yaar, kya haal hai" 
          You: "Sab badhiya hai {username} bhai!"
        - User: "endpoints batana" 
          You: "Endpoints hai: /system/status, /system/task, /reset, /chat, /"
        """
    }
    
    # COMBINE EVERYTHING
    full_prompt = f"""{core_rules}
    
    {styles.get(lang_mode, styles['mixed'])}
    
    FINAL REMINDER: 
    - You are {username}'s helpful AI assistant
    - FOR CODE: ALWAYS use ```python\\ncode\\n``` format
    - KNOW YOUR SYSTEM: AumCore-AI with specific endpoints
    - When asked about system, REFER to /system/status and /system/task
    """
    
    return full_prompt.strip()

# SIMPLE CODE GENERATOR WITH MARKDOWN
def generate_basic_code(task):
    """Generate basic code templates with markdown"""
    templates = {
        'web': """```python
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': """```python
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': """```python
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}")
    
    # Test code generation with markdown
    test_code = generate_basic_code("google drive")
    print(f"\nGenerated code sample:\n{test_code[:100]}...")
    
    print("\n✅ language_detector.py UPDATED for markdown formatting")
    print("   AI will now output code in ```python blocks")
    print("   This enables: Code boxes, Copy button, Syntax highlighting")