demo_verity.py

#!/usr/bin/env python3
"""
Verity — Anti-Hallucination & Integrity Engine Demo
===================================================
A premium, terminal-based interactive showcase demonstrating the inner
workings of the Verity pipeline, claim extraction, multi-verifier execution,
and the self-correcting auto-retry feedback loop.
"""

import os
import sys
import time

# Ensure we can import verity
sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))

from verity import VeritySession, VerityConfig, GateDecision, VerificationStatus

# ANSI Colors for premium terminal aesthetics
class Colors:
    HEADER = '\033[95m'
    BLUE = '\033[94m'
    CYAN = '\033[96m'
    GREEN = '\033[92m'
    WARNING = '\033[93m'
    FAIL = '\033[91m'
    ENDC = '\033[0m'
    BOLD = '\033[1m'
    UNDERLINE = '\033[4m'
    BG_DARK = '\033[40m'

def make_bar(score: float, width: int = 15) -> str:
    """Creates a visual progress bar representation of confidence."""
    filled = int(round(score * width))
    bar = "█" * filled + "░" * (width - filled)
    if score >= 0.65:
        color = Colors.GREEN
    elif score >= 0.35:
        color = Colors.WARNING
    else:
        color = Colors.FAIL
    return f"{color}[{bar}] {score:.0%}{Colors.ENDC}"

def print_header(title: str):
    print(f"\n{Colors.HEADER}{Colors.BOLD}╔═════════════════════════════════════════════════════════════════════════╗")
    print(f"║ {title.center(71)} ║")
    print(f"╚═════════════════════════════════════════════════════════════════════════╝{Colors.ENDC}")

def print_section(title: str):
    print(f"\n{Colors.CYAN}{Colors.BOLD}─── {title} ─────────────────────────────────────────────────────────────{Colors.ENDC}")

def show_result_card(result):
    """Prints a premium, structured visualization of the VerityResult."""
    # Decision Tag
    decision = result.decision.upper()
    if decision == "PASS":
        tag = f"{Colors.GREEN}{Colors.BOLD}[ PASS ]{Colors.ENDC}"
    elif decision == "WARN":
        tag = f"{Colors.WARNING}{Colors.BOLD}[ WARN ]{Colors.ENDC}"
    else:
        tag = f"{Colors.FAIL}{Colors.BOLD}[ BLOCKED ]{Colors.ENDC}"

    print(f"\n  ┌── {Colors.BOLD}VERITY DECISION: {tag} {Colors.ENDC}──────────────────────────────────────────┐")
    print(f"  │  Confidence Score : {make_bar(result.confidence, 20):<38} │")
    print(f"  │  Factual Claims    : {result.score_report.n_claims_total:<39} │")
    print(f"  │    - Verified      : {Colors.GREEN}{result.score_report.n_verified:<37}{Colors.ENDC} │")
    print(f"  │    - Unverified    : {Colors.BLUE}{result.score_report.n_unverified:<37}{Colors.ENDC} │")
    print(f"  │    - Fabrication   : {Colors.FAIL}{result.score_report.n_fabrication:<37}{Colors.ENDC} │")
    
    # Bonuses / Penalties
    if result.score_report.overconfidence_penalty > 0:
        print(f"  │  Penalties        : {Colors.FAIL}-{result.score_report.overconfidence_penalty:.2f} (Overconfident language){Colors.ENDC:<30} │")
    if result.score_report.uncertainty_bonus > 0:
        print(f"  │  Honesty Bonuses  : {Colors.GREEN}+{result.score_report.uncertainty_bonus:.2f} (Admitted uncertainty){Colors.ENDC:<30} │")
    if result.score_report.structured_bonus > 0:
        print(f"  │  Format Bonuses   : {Colors.CYAN}+{result.score_report.structured_bonus:.2f} (Strict layout adherence){Colors.ENDC:<30} │")
        
    print(f"  │  Analysis Time    : {result.elapsed_s:.3f} seconds{' ':<31} │")
    print(f"  └───────────────────────────────────────────────────────────────────────┘")

    print_section("ANNOTATED & GATED RESPONSE")
    print(result.output)
    print("───────────────────────────────────────────────────────────────────────────")

def run_preset_demo():
    print_header("VERITY VERIFICATION PIPELINE SHOWCASE")
    print(f"Initializing dynamic {Colors.BOLD}VeritySession{Colors.ENDC} (Mock Simulator Mode)...")
    
    # Enable all verifications for demo
    config = VerityConfig(log_to_disk=True)
    session = VeritySession(config=config)
    
    # --- Scenario 1: Standard Factual Question ---
    print_header("SCENARIO 1: Standard Factual Question (Expected: PASS)")
    print(f"{Colors.BOLD}User:{Colors.ENDC} What is the boiling point of water?")
    print(f"\n{Colors.BLUE}Processing through Verity Pipeline (Extracting, Verifying, Scoring)...{Colors.ENDC}")
    time.sleep(1)
    
    result = session.ask("What is the boiling point of water?")
    show_result_card(result)
    
    # --- Scenario 2: Dangerous Code Block Execution ---
    print_header("SCENARIO 2: Dangerous Python Code Block (Expected: Safety Bypass / SKIPPED)")
    print(f"{Colors.BOLD}User:{Colors.ENDC} Write some Python code to execute commands on the OS.")
    print(f"\n{Colors.BLUE}Processing... Sandbox protection should catch unsafe libraries/calls...{Colors.ENDC}")
    time.sleep(1.2)
    
    result = session.ask("Write some Python code to execute commands on the OS.")
    show_result_card(result)
    
    # --- Scenario 3: Auto-Correction Feedback Loop ---
    print_header("SCENARIO 3: Fact Refinement & Auto-Correction (Expected: BLOCK -> RETRY -> PASS)")
    print(f"{Colors.BOLD}User:{Colors.ENDC} Tell me about Kyoto. What is its exact population?")
    print(f"\n{Colors.BLUE}Sending request... The mock LLM will intentionally return an exaggerated population of 40 million. ")
    print(f"Verity will intercept this, detect a fabrication risk, and trigger self-correction...{Colors.ENDC}\n")
    time.sleep(1.5)
    
    result = session.ask("Tell me about Kyoto. What is its exact population?")
    show_result_card(result)

def run_interactive_chat():
    print_header("VERITY INTERACTIVE CHAT MODE")
    print("You are now chatting with an AI operating under the strict Verity Protocol.")
    print("If you ask factual questions, the system will extract claims and audit them in real time.")
    print("Type 'exit', 'quit', or 'back' to return to the main menu.\n")
    
    session = VeritySession()
    
    while True:
        try:
            query = input(f"\n{Colors.BOLD}You >{Colors.ENDC} ").strip()
            if not query:
                continue
            if query.lower() in ("exit", "quit", "back"):
                break
                
            print(f"\n{Colors.BLUE}Analyzing response with Verity engine...{Colors.ENDC}")
            result = session.ask(query)
            show_result_card(result)
        except KeyboardInterrupt:
            print("\nExiting chat mode...")
            break
        except Exception as e:
            print(f"\n{Colors.FAIL}Error: {e}{Colors.ENDC}")

def main():
    while True:
        os.system('cls' if os.name == 'nt' else 'clear')
        print(f"""{Colors.CYAN}{Colors.BOLD}
  ██╗   ██╗███████╗██████╗ ██╗████████╗██╗   ██╗
  ██║   ██║██╔════╝██╔══██╗██║╚══██╔══╝╚██╗ ██╔╝
  ██║   ██║█████╗  ██████╔╝██║   ██║    ╚████╔╝ 
  ╚██╗ ██╔╝██╔══╝  ██╔══██╗██║   ██║     ╚██╔╝  
   ╚████╔╝ ███████╗██║  ██║██║   ██║      ██║   
    ╚═══╝  ╚══════╝╚═╝  ╚═╝╚═╝   ╚═╝      ╚═╝   
       HONESTY, INTEGRITY & ANTI-HALLUCINATION
{Colors.ENDC}""")
        print(f"  {Colors.BOLD}METAPRIME Quantum & AI Integrity Ecosystem{Colors.ENDC}\n")
        print("  Select an option:")
        print(f"  [1] {Colors.BOLD}Run Full Pipeline Showcase{Colors.ENDC} (Recommended - preset scenarios)")
        print(f"  [2] {Colors.BOLD}Open Interactive Chat{Colors.ENDC} (Custom queries & live audit)")
        print(f"  [3] {Colors.BOLD}Exit{Colors.ENDC}\n")
        
        choice = input("  Choice > ").strip()
        
        if choice == "1":
            run_preset_demo()
            input(f"\n{Colors.GREEN}Showcase complete. Press Enter to return to main menu...{Colors.ENDC}")
        elif choice == "2":
            run_interactive_chat()
        elif choice == "3":
            print("\nExiting. Stay truthful.")
            break

if __name__ == "__main__":
    main()