Update README.md

README.md

@@ -1,14 +1,18 @@
 ---
-title: Qwen-Blindfold
+title: Blindfold Model Tool
 emoji: 🙈
 colorFrom: blue
 colorTo: purple
 sdk: gradio
-sdk_version: 6.8.0
+sdk_version: 6.9.0
 app_file: app.py
 pinned: false
-license: apache-2.0
+license: mit
+
 hf_oauth: true
+hf_oauth_scopes:
+ - write-repos
+ - manage-repos
 ---
 
 # 🙈 Blindfold Model Tool
@@ -25,17 +29,18 @@ This Gradio-based tool allows you to:
 
 ## How to use
 
-1. **Enter the model URL**: Provide the Hugging Face model ID or full URL (e.g., `Qwen/Qwen2.5-VL-32B-Instruct`)
-2. **Provide your HF token**: Paste your Hugging Face token with write permissions
+1. **Login with your Hugging Face account**: Click the "Login" button to authenticate
+2. **Enter the model URL**: Provide the Hugging Face model ID or full URL (e.g., `Qwen/Qwen2.5-VL-32B-Instruct`)
 3. **Set output repository name**: Choose a name for your new repository
 4. **Click "Process & Upload"**: The tool will download, process, and upload the model
 
 ## Requirements
 
 - A Hugging Face account
-- A Hugging Face token with **write** permissions (get it at https://huggingface.co/settings/tokens)
 - The source model must be in safetensors format
 
+> **Note**: This tool uses OAuth to authenticate with your Hugging Face account. Your token is handled securely and is never stored.
+
 ## What gets removed
 
 The tool removes:
@@ -43,6 +48,17 @@ The tool removes:
 - Vision configuration sections from `config.json`
 - Image/video token IDs
 
+The architecture is changed to `Qwen3_5MoeForCausalLM` for text-only inference.
+
+## Local Development
+
+To run this locally:
+
+```bash
+pip install -r requirements.txt
+python blindfold_gradio.py
+```
+
 ## License
 
-Apache 2.0
+MIT License