---
tags:
- ml-intern
---
# 🧙‍♂️ Wizard-Vibe Core

> **Sandbox-First Architecture** — single-file SSE streaming code generator with Reflect-Select self-healing and A2A native deploy.

## Quick Start

```bash
chmod +x sandbox.sh && ./sandbox.sh
```

Or with hot-reload:

```bash
HOT_RELOAD=1 ./sandbox.sh
```

## Architecture

```
core.py          ← Single-file: SSE server + orchestrator + self-heal + deploy
hot_reload.py    ← File watcher for dev mode (auto-restart on changes)
static/          ← Liquid Glass UI (minimalist HTML/CSS/JS)
sandbox.sh       ← One-command bootstrap
Dockerfile       ← Containerized deployment
```

## API

| Endpoint | Method | Description |
|----------|--------|-------------|
| `/` | GET | Liquid Glass UI |
| `/api/health` | GET | Health check |
| `/api/stream` | POST | SSE streaming code generation |
| `/api/publish` | POST | GitHub + A2A deploy |
| `/api/status?session_id=X` | GET | Session status |
| `/api/preview?session_id=X` | GET | Sandbox iframe content |
| `/.well-known/agent.json` | GET | A2A agent card |

## Models

- Vision/UI → `microsoft/Phi-3-vision-128k-instruct`
- Logic/Backend → `deepseek-ai/DeepSeek-V3-0324`
- Code/Infra → `Qwen/Qwen3-Coder-30B-A3B-Instruct`
- Fallback → `mistralai/Mistral-7B-Instruct-v0.3`

## Links

- **HF Repo**: https://huggingface.co/dryymatt/Wizard-Vibe-Core
- **HF Space**: https://dryymatt-wizard-vibe-core.hf.space
- **GitHub Mirror**: https://github.com/M523zappin/Wizard-Vibe-Core

---

*Built with the Reflect-Select self-healing architecture. Every line validated in sandbox before publish.*

<!-- ml-intern-provenance -->
## Generated by ML Intern

This model repository was generated by [ML Intern](https://github.com/huggingface/ml-intern), an agent for machine learning research and development on the Hugging Face Hub.

- Try ML Intern: https://smolagents-ml-intern.hf.space
- Source code: https://github.com/huggingface/ml-intern

## Usage

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = 'dryymatt/Wizard-Vibe-Core'
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id)
```

For non-causal architectures, replace `AutoModelForCausalLM` with the appropriate `AutoModel` class.