---
title: ForgeEnv Repair Agent Demo
emoji: 🔧
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: 5.7.1
app_file: app.py
pinned: true
license: apache-2.0
hardware: zero-a10g
tags:
  - openenv
  - self-improvement
  - code-repair
  - schema-drift
short_description: Trained Repair Agent fixes HF scripts under drift
---

# ForgeEnv Repair Agent — Live Demo

Paste a broken HuggingFace training script and the error trace it produced.
The trained Repair Agent (Qwen2.5-3B + LoRA) emits a unified diff that should
restore the script. Inference runs on ZeroGPU (free A10G).

- **Environment server (OpenEnv):**
  <https://huggingface.co/spaces/akhiilll/forgeenv>
- **Trained model (LoRA + repair_library.json):**
  <https://huggingface.co/akhiilll/forgeenv-repair-agent>
- **Project README & plots:**
  <https://github.com/akhiilll/forgeenv>