---
title: Axon v26
emoji: ⚡
colorFrom: indigo
colorTo: purple
sdk: docker
pinned: true
license: mit
short_description: "Free AI Coding Assistant - 5 Models - 25 Tools - 100% Local"
---

<div align="center">

# ⚡ Axon v26

### The Ultimate Free AI Coding Assistant

**5 Models • 25 Tools • 100% Local • Zero API Keys**

[![Built with llama.cpp](https://img.shields.io/badge/llama.cpp-Powered-blue?style=for-the-badge&logo=cplusplus)](https://github.com/ggerganov/llama.cpp)
[![Models](https://img.shields.io/badge/Models-5-green?style=for-the-badge)](https://huggingface.co/Qwen)
[![Tools](https://img.shields.io/badge/Tools-25-orange?style=for-the-badge)](https://huggingface.co/spaces/AIencoder/Axon)
[![License](https://img.shields.io/badge/License-MIT-yellow?style=for-the-badge)](LICENSE)
[![AVX2 Optimized](https://img.shields.io/badge/AVX2-Optimized-red?style=for-the-badge)](https://huggingface.co/datasets/AIencoder/llama-cpp-wheels)

---

[**Try Axon Now →**](https://huggingface.co/spaces/AIencoder/Axon)

</div>

---

## 🚀 What is Axon?

Axon is a **free, privacy-first AI coding assistant** that runs entirely locally using llama.cpp. No API keys, no rate limits, no data collection - just powerful AI coding tools at your fingertips.

Built from the ground up after spending 2 days building llama-cpp-python wheels that didn't exist. Now you don't have to.

---

## 🤖 5 Powerful Models

Choose the right model for your task - from instant responses to complex reasoning.

| Model | Size | Speed | Best For |
|-------|------|-------|----------|
| 🧠 **DeepSeek V2 Lite** | ~9GB | ⭐⭐ | Complex logic, MoE architecture |
| ⚖️ **Qwen2.5 Coder 7B** | ~4.5GB | ⭐⭐⭐ | Balanced quality & speed |
| 🚀 **Qwen2.5 Coder 3B** | ~2GB | ⭐⭐⭐⭐ | Fast & highly capable |
| 💨 **Qwen2.5 Coder 1.5B** | ~1GB | ⭐⭐⭐⭐⭐ | Quick tasks |
| 🔬 **Qwen2.5 Coder 0.5B** | ~0.3GB | ⚡ | Instant responses |

> Models download automatically on first use. Storage persists between sessions.

---

## 🛠️ 25 Tools

### Core Tools
| Tool | Description |
|------|-------------|
| 💬 **Chat** | Conversational coding help with streaming responses |
| ⚡ **Generate** | Create code from natural language descriptions |
| 🔍 **Explain** | Understand any code (Brief / Normal / Detailed modes) |
| 🔧 **Debug** | Find and fix bugs with error context |
| 📋 **Review** | Code quality, security & performance review |

### Advanced Tools
| Tool | Description |
|------|-------------|
| 🔐 **Security Scan** | Find vulnerabilities (SQL injection, XSS, etc.) |
| 📊 **Complexity** | Big O analysis for time & space |
| 🔄 **Convert** | Translate between 22+ programming languages |
| 🧪 **Test** | Generate comprehensive unit tests |
| 📝 **Document** | Add docstrings, comments & inline docs |
| 🚀 **Optimize** | Performance improvements & refactoring |
| 🔀 **Diff** | Compare two code snippets |
| 📐 **Pseudocode** | Convert code to pseudocode/flowcharts |
| 🎓 **Interview** | Generate coding challenges & solutions |

### Builders
| Tool | Description |
|------|-------------|
| 🗄️ **SQL Builder** | Natural language → SQL queries |
| 🐚 **Shell Builder** | Natural language → Bash/PowerShell commands |
| ⏰ **Cron Builder** | Create cron schedule expressions |
| 🎯 **Regex Builder** | Pattern creation with explanations |
| 🔗 **API Builder** | Generate REST endpoint boilerplate |

### Data Tools
| Tool | Description |
|------|-------------|
| 📦 **Mock Data** | Generate realistic test data (JSON, CSV, etc.) |
| 🔄 **Format Converter** | Convert between JSON/YAML/XML/CSV/TOML |

### NEW in v26
| Tool | Description |
|------|-------------|
| 🎨 **Refactor** | Restructure code for better design patterns |
| 📊 **Benchmark** | Generate performance benchmark code |
| 🔗 **Dependency Analyzer** | Analyze imports & dependencies |
| 📋 **Changelog** | Generate changelogs from code diffs |
| 💡 **Suggest** | AI-powered improvement suggestions |

---

## 🎤 Voice Input

Speak your code requests using Whisper-powered speech-to-text. Just click the microphone and talk naturally.

---

## 🌙 Dark Mode

Toggle between light and dark themes. Your preference is saved automatically.

---

## 💾 Export

Save your chat history and generated code for later reference.

---

## ⚡ Performance

Axon uses **AVX2-optimized** llama-cpp-python wheels for **2-3x faster** inference compared to basic builds.

| Build Type | Tokens/sec (3B) | Compatibility |
|------------|-----------------|---------------|
| Basic | ~10-15 | All x86_64 |
| **AVX2 (Axon)** | **~30-40** | Intel Haswell+ / AMD Zen+ (2013+) |

---

## 🔒 Privacy First

| Feature | Axon | Cloud Alternatives |
|---------|------|--------------------|
| 100% Local Processing | ✅ | ❌ |
| No API Keys Required | ✅ | ❌ |
| No Rate Limits | ✅ | ❌ |
| No Data Collection | ✅ | ❌ |
| Works Offline | ✅ | ❌ |
| Free Forever | ✅ | ❌ |

Your code **never** leaves your browser. Period.

---

## 💻 Supported Languages

Python • JavaScript • TypeScript • Go • Rust • Java • C++ • C# • C • PHP • Ruby • Swift • Kotlin • Scala • R • Julia • Perl • HTML/CSS • SQL • Bash • PowerShell • Lua

---

## 🛞 Pre-built Wheels

Tired of building llama-cpp-python from source? Use our AVX2-optimized wheels:

```bash
# Python 3.10
pip install https://huggingface.co/datasets/AIencoder/llama-cpp-wheels/resolve/main/llama_cpp_python-0.3.16-cp310-cp310-manylinux_2_31_x86_64.whl

# Python 3.11
pip install https://huggingface.co/datasets/AIencoder/llama-cpp-wheels/resolve/main/llama_cpp_python-0.3.16-cp311-cp311-manylinux_2_31_x86_64.whl
```

**Features:**
- AVX2 + FMA + F16C enabled
- 2-3x faster than basic builds
- Works on Intel Haswell+ (2013+) and AMD Zen+ (2018+)

[**Browse all wheels →**](https://huggingface.co/datasets/AIencoder/llama-cpp-wheels)

---

## 🏗️ Tech Stack

| Component | Technology |
|-----------|------------|
| Inference | [llama.cpp](https://github.com/ggerganov/llama.cpp) via [llama-cpp-python](https://github.com/abetlen/llama-cpp-python) |
| Wheels | [AIencoder/llama-cpp-wheels](https://huggingface.co/datasets/AIencoder/llama-cpp-wheels) (AVX2 optimized) |
| UI | [Gradio](https://gradio.app/) |
| Speech | [faster-whisper](https://github.com/SYSTRAN/faster-whisper) |
| Models | GGUF format from HuggingFace |
| Hosting | HuggingFace Spaces (Docker) |

---

## 🚀 Self-Hosting

Want to run Axon on your own machine?

```bash
# Clone the space
git clone https://huggingface.co/spaces/AIencoder/Axon
cd Axon

# Build and run
docker build -t axon .
docker run -p 7860:7860 -v axon_data:/data axon
```

Then open `http://localhost:7860`

---

## 📊 Changelog

### v26 (Current) - The FINAL Version
- ✨ Added 6 new tools (25 total)
- 🎨 Redesigned UI with better UX
- ⚡ AVX2-optimized wheels for 2-3x speed boost
- 🔧 Gradio 6.0 compatibility fixes
- 📦 Optimized storage usage

### v6
- 🚀 Initial public release
- 🤖 8 models (reduced to 5 due to storage)
- 🛠️ 19 tools
- 🎤 Whisper voice input

---

## 🙏 Credits

- [Qwen](https://huggingface.co/Qwen) - Amazing coding models
- [DeepSeek](https://huggingface.co/deepseek-ai) - Logic-focused models
- [ggerganov](https://github.com/ggerganov) - llama.cpp
- [abetlen](https://github.com/abetlen) - llama-cpp-python
- [Gradio](https://gradio.app/) - UI framework
- [SYSTRAN](https://github.com/SYSTRAN) - faster-whisper

---

## 📄 License

MIT License - Use it, modify it, share it!

---

## ⭐ Support

If Axon helps you code faster, consider:
- ⭐ Starring the Space
- 🐛 Reporting issues
- 💡 Suggesting features
- 📢 Sharing with friends

---

<div align="center">

**Built with ❤️ and mass caffeine by [AIencoder](https://huggingface.co/AIencoder)**

*No sleep was had in the making of those wheels.*

</div>