fixmyslop 0.2.2

AI wrote your code. FixMySlop fixes it.

FixMySlop

FixMySlop — The free, 100% local AI-powered code janitor that finds and fixes bugs, security holes, and AI slop in your projects. No cloud, no API keys, no BS.

• Cleans messy AI-generated code (hallucinations, bad patterns, weak security)
• Gives a clear Slop Score, detailed audit report, and one-click fix generation
• Works offline with Ollama, llama.cpp, or vLLM so your code stays private

---

Visual Tour

Screenshot Guide

Screenshot File	What It Shows
screenshots/Search.png	Scan setup flow (project selection + scan controls)
screenshots/report1.png	Report view with issue list and quality insights
screenshots/fix.png	Fix workflow and diff-style patch review
screenshots/cli.png	CLI scan output for terminal-first users

---

Features

• ⚡ Turbo + Deep scan modes for speed vs depth
• 🔍 Full repository scanning with .gitignore awareness
• 🛡️ Security checks (Bandit + LLM context analysis)
• 🧹 AI smell detection (hallucinated imports, weak validation, broad exception handling)
• 📊 Slop Score (0-100) to summarize code health
• 🔧 LLM-powered fix generation with unified diffs
• 📄 Report export in Markdown, HTML, JSON, CSV, and PDF
• 🎨 Polished PyQt5 desktop GUI with dark/light themes
• 💻 Full CLI for automation and CI-friendly usage
• 🤖 Auto model fallback when configured model is missing
• 🪟 Windows-safe terminal output handling
• 🏠 100% local-first architecture

---

Why FixMySlop?

FixMySlop combines static analysis + LLM reasoning + local privacy in one workflow.

Compared To	Typical Limitation	FixMySlop Advantage
Ruff/Pylint-only workflows	Strong syntax/style checks, limited semantic AI context	Adds LLM-based deep reasoning on top of static tools
Cloud AI code reviewers	Code leaves your machine, subscription cost	100% local and private
IDE-only assistants	Great inline help, weaker full-repo audit/report flow	Full-repo scans, report panel, Slop Score, fix workflow

---

Installation

Install from PyPI

# Base install (CLI + core engine)
pip install fixmyslop

# Install with GUI support
pip install fixmyslop[gui]

# Install with static analysis tools
pip install fixmyslop[analysis]

# Install everything
pip install fixmyslop[gui,analysis]

After install, run:

fixmyslop scan ./your-project --mode turbo

Prerequisites

• Python 3.10+
• A local OpenAI-compatible backend (Ollama recommended): https://ollama.ai

Quick Start

# Clone
git clone https://github.com/MrSpideyNihal/FixMySlop.git
cd FixMySlop

# Install dependencies
pip install -r requirements.txt

# Pull a local model (example)
ollama pull qwen2.5-coder:3b

# Launch GUI
python main.py

# Or run CLI
python main.py scan ./your-project --mode turbo

FixMySlop can auto-detect an available model, so manual model config is optional.

Optional Static Tools

pip install ruff bandit
pip install semgrep

Troubleshooting

• Backend not reachable: run ollama serve first
• Model not found: run ollama pull qwen2.5-coder:7b (or another installed coder model)
• Windows terminal shows garbled Unicode: switch to UTF-8 code page (chcp 65001) or use Windows Terminal/CMD

---

Usage

GUI

python main.py

Main sections:

• Home
• Scan
• Report
• Fix
• Settings

CLI Recipes

# Quick scan
python main.py scan ./myproject

# Turbo mode (fast top issues)
python main.py scan ./myproject --mode turbo

# Deep mode (full pass)
python main.py scan ./myproject --mode deep

# Deep static-only
python main.py scan ./backend --mode deep --no-llm --output json

# Save report
python main.py scan ./myproject --save slop-report.md --output markdown

# Use explicit model
python main.py scan ./myproject --model qwen2.5-coder:3b --mode turbo

# List backend models
python main.py models

Scan Modes

Mode	Focus	Typical per-file estimate
turbo	Prioritize high-impact findings quickly	~20-40s per file
deep	Broader and more exhaustive finding coverage	~15-30s per file
--no-llm	Static analysis only	< 1s per file

---

Configuration

Config file path: ~/.fixmyslop/config.yaml

model: ""
base_url: http://localhost:11434/v1
api_key: ollama
temperature: 0.2
theme: dark
font_size: 14
use_ruff: true
use_bandit: true
use_semgrep: false
auto_backup: true

If model is empty, FixMySlop auto-selects the best available model from your running local backend.

---

Supported Languages

Python, JavaScript, TypeScript, Go, Rust, Java, C++, C, C#, Ruby, PHP, Swift, Kotlin

---

Architecture

FixMySlop/
├── macros.py
├── main.py
├── core/
│   ├── scanner.py
│   ├── analyzer.py
│   ├── llm_client.py
│   ├── fix_engine.py
│   └── ...
├── ui/
│   ├── panels/
│   ├── widgets/
│   └── assets/
├── cli/
├── utils/
└── tests/

Design rules:

• core/ never imports ui/
• Constants are centralized in macros.py
• Heavy work runs in QThread
• Path handling uses pathlib.Path

---

Running Tests

pytest tests/ -v

---

Contributing

Contributions are welcome.

1. Open an issue for bugs or feature ideas
2. Fork and create a branch
3. Add tests for behavior changes
4. Run pytest tests/ -q
5. Open a PR with a clear summary

---

Roadmap

• More language-specific rules and higher-quality fix templates
• Better safe auto-apply for low-risk fixes
• Pre-commit and CI integration helpers
• Optional richer report visualizations

---

Acknowledgments

• Ollama
• llama.cpp
• vLLM
• Ruff
• Bandit
• Semgrep
• PyQt5
• fpdf2

---

License

MIT — free forever and open source.

---

Loved this project? Please star the repo.