godtier-debugger 1.0.0

AI-powered self-healing code debugger with evolutionary optimization

GodTier Debugger

AI-powered self-healing code debugger with evolutionary optimization.

---

GodTier Debugger finds bugs, fixes them, and makes your code faster — automatically. No cloud APIs, no API keys, no subscriptions. Everything runs locally using AST analysis, symbolic reasoning, and optional LLM inference.

Installation

pip install godtier-debugger

Quick Start

CLI

# Scan a project for bugs
godtier scan myproject/

# Auto-fix a file
godtier fix broken_script.py --write

# AI-powered analysis
godtier analyze complex_module.py

# Security audit
godtier safety untrusted_code.py

# Check system status
godtier info

Python API

from src.debugger.auto_debugger import AutoDebugger
from src.darwin.safety import DarwinSafetyInspector
from src.lazarus.engine import lazarus_protect

# Auto-fix broken code
debugger = AutoDebugger()
result = debugger.debug_code("""
def hello(name)
    print("Hello " + name
""")
print(result.corrected_code)   # Fixed: colon + closing paren
print(result.success)          # True

# Security scan
inspector = DarwinSafetyInspector()
safe, issues = inspector.check("import os; os.system('rm -rf /')")
print(safe)    # False
print(issues)  # ['Forbidden call: os.system', ...]

# Self-healing decorator — function repairs itself at runtime
@lazarus_protect
def risky_division(a, b):
    return a / b

result = risky_division(10, 0)  # No crash — healed automatically

Features

Feature	What it does
Auto-Debugger	AST-based detection & correction of 8 error types (syntax, imports, undefined vars, type mismatches, security risks, etc.)
Darwin Protocol	Evolutionary code optimization — mutate, benchmark (time + memory), keep winners in gene memory (SQLite)
Lazarus Protocol	@lazarus_protect decorator — catches exceptions, generates hot-fixes, patches functions in memory without restart
Safety Scanner	AST security inspector with whitelist/blacklist — blocks eval, exec, os.system, subprocess
Symbolic AI Engine	Knowledge graphs + rule-based reasoning for deep code analysis
Aurora LLM Bridge	Optional local LLM inference (Qwen2.5-14B) for enhanced analysis — zero cloud, zero cost
Multi-Language	Python + JavaScript support
Dashboard	Real-time HTTP + WebSocket monitoring

How It Works

Source Code
    |
    v
[Auto-Debugger] ──> AST Parse ──> 8 Check/Fix Passes ──> Corrected Code
    |
    v
[Darwin Protocol] ──> Mutate ──> Safety Check ──> Benchmark ──> Best Survives
    |
    v
[Lazarus Protocol] ──> Runtime Exception ──> AI Analysis ──> Hot-Patch ──> Retry

No GPU required. The core engine runs on pure Python (AST + symbolic reasoning). If you have an NVIDIA GPU with a local LLM, the Aurora Bridge activates automatically for deeper analysis.

Test Results

Validated with 21 unit tests (100% pass) and 25 real-world stress tests (92% pass rate):

tests/test_aurora_bridge.py        PASSED
tests/test_code_upgrade_system.py  PASSED (9 tests)
tests/test_darwin.py               PASSED
tests/test_god_tier_wrapper.py     PASSED
tests/test_integration.py          PASSED
tests/test_lazarus_unit.py         PASSED (2 tests)
tests/test_router_simulation.py    PASSED
tests/test_specialist.py           PASSED
tests/test_symbolic_parser_unit.py PASSED (4 tests)

Stress tests cover: real-world FastAPI bugs, multi-bug files, security exploits, evolutionary optimization, Lazarus self-healing under production conditions.

Performance

Metric	Value
Analysis speed	< 2ms per file (symbolic)
Auto-fix accuracy	92% on real-world code
Security scan	Catches eval/exec/os.system/subprocess
Memory overhead	< 50MB (no GPU mode)
LLM analysis (optional)	~15s per query (RTX 3060)

Dependencies

Core (installed automatically):

• networkx — Knowledge graphs
• astor — AST manipulation

Optional (pip install godtier-debugger[full]):

• numpy — Numerical computations
• aiohttp — WebSocket dashboard

No PyTorch. No transformers. No cloud APIs. Just fast, local debugging.

Project Structure

src/
├── cli.py              # CLI entry point (godtier command)
├── ai/                 # Symbolic AI + Aurora LLM bridge
├── debugger/           # Auto-debugging engine (8 error types)
├── darwin/             # Evolutionary optimization + safety
├── lazarus/            # Self-healing runtime protocol
├── dashboard/          # Real-time monitoring
├── languages/          # Multi-language support
└── utils/              # Git automation, patching, upgrades

Security

Multiple protection layers:

• Pre-execution AST scanning with whitelist/blacklist
• Sandboxed execution of mutated code
• Automatic validation before applying patches
• Rollback on failed patches

Blocked operations: os.system(), subprocess.Popen(), eval(), exec(), __import__(), globals(), breakpoint()

Contributing

git clone https://github.com/crnlchez1234-afk/godtier-debugger.git
cd godtier-debugger
pip install -e ".[dev]"
pytest tests/ -v

License

MIT License — Copyright 2025 Cruz Sanchez

---

GodTier Debugger — Fix bugs before they fix you.