persist code data

Dockerfile

@@ -19,8 +19,13 @@ RUN apt-get update && apt-get install -y \
     python3-pip \
     nodejs \
     npm \
+    unzip \
+    rsync \
     && rm -rf /var/lib/apt/lists/*
 
+# Install Hugging Face CLI for dataset access
+RUN pip3 install --no-cache-dir huggingface_hub[cli]
+
 # Install code-server
 RUN curl -fsSL https://code-server.dev/install.sh | sh
 
@@ -44,9 +49,6 @@ EXPOSE 7860
 # Create config directory
 RUN mkdir -p /home/coder/.config/code-server
 
-# Install unzip for extracting extensions
-RUN sudo apt-get update && sudo apt-get install -y unzip && sudo rm -rf /var/lib/apt/lists/*
-
 # Install AI coding assistant extensions that support Gemini
 RUN code-server --install-extension Continue.continue || true
 
@@ -59,5 +61,9 @@ RUN echo "bind-addr: 0.0.0.0:7860" > /home/coder/.config/code-server/config.yaml
     echo "password: ${PASSWORD}" >> /home/coder/.config/code-server/config.yaml && \
     echo "cert: false" >> /home/coder/.config/code-server/config.yaml
 
-# Start code-server
-CMD ["code-server", "--bind-addr", "0.0.0.0:7860", "--auth", "password", "/home/coder/workspace"]
+# Copy startup script
+COPY --chown=coder:coder startup.sh /home/coder/startup.sh
+RUN chmod +x /home/coder/startup.sh
+
+# Start code-server with persistence
+CMD ["/home/coder/startup.sh"]

README.md

@@ -10,25 +10,55 @@ app_port: 7860
 
 # VS Code on Web - Hugging Face Spaces
 
-This Space deploys VS Code in your browser using code-server.
+This Space deploys VS Code in your browser using code-server with persistent storage via Hugging Face Datasets.
 
 ## Features
 
 - 🌐 VS Code running in your browser
+- 💾 Persistent storage using Hugging Face Datasets
 - 🐍 Pre-installed Python 3 and pip
 - 📦 Node.js and npm included
 - 🛠️ Build tools and common utilities
 - 🔒 Password-protected access
+- 🤖 AI coding assistant (Continue) with Gemini support
 
-## Access
+## Setup
+
+### 1. Create a Hugging Face Dataset
+
+Create a new private dataset on Hugging Face to store your VS Code data:
+https://huggingface.co/new-dataset
+
+### 2. Configure Environment Variables
+
+In your Space settings, add these secrets:
+
+- `HF_TOKEN`: Your Hugging Face access token (with write permissions)
+- `HF_DATASET`: Your dataset name (e.g., `username/vscode-data`)
+
+### 3. Access Your VS Code
 
 Once deployed, access your VS Code instance through this Space. The default password is `huggingface`.
 
+## How Persistence Works
+
+The startup script automatically:
+- Downloads your saved workspace, settings, and extensions from the HF dataset on startup
+- Auto-saves every 5 minutes while running
+- Saves all data when the Space shuts down
+
+**Persisted data includes:**
+- VS Code settings and configuration
+- Installed extensions
+- Workspace files and folders
+- Extension settings (including API keys)
+
 ## Configuration
 
 - **Port**: 7860
-- **Default Password**: `huggingface`
+- **Default Password**: `huggingface` (change in Dockerfile)
 - **Workspace Directory**: `/home/coder/workspace`
+- **Auto-save interval**: Every 5 minutes
 
 ## Installed Tools
 
@@ -37,10 +67,12 @@ Once deployed, access your VS Code instance through this Space. The default pass
 - Node.js with npm
 - Build essentials (gcc, g++, make)
 - Text editors (vim, nano)
-- curl, wget
+- Hugging Face CLI
+- Continue extension (supports Gemini, GPT-4, Claude, etc.)
 
 ## Notes
 
-- You can install additional VS Code extensions through the browser interface
-- Files created in the workspace persist during the session
-- For persistent storage, consider using Hugging Face Datasets or external storage
+- Without `HF_TOKEN` and `HF_DATASET` set, the Space runs without persistence
+- The first startup takes longer as it sets up the environment
+- Large workspaces may take time to sync
+- Your data is stored in your private HF dataset and not accessible to others

startup.sh

@@ -0,0 +1,95 @@
+#!/bin/bash
+set -e
+
+echo "Starting VS Code with Hugging Face persistence..."
+
+# Check if HF_TOKEN and HF_DATASET are set
+if [ -z "$HF_TOKEN" ] || [ -z "$HF_DATASET" ]; then
+    echo "Warning: HF_TOKEN or HF_DATASET not set. Running without persistence."
+    exec code-server --bind-addr 0.0.0.0:7860 --auth password /home/coder/workspace
+    exit 0
+fi
+
+# Login to Hugging Face
+echo "Logging in to Hugging Face..."
+huggingface-cli login --token "$HF_TOKEN" --add-to-git-credential
+
+# Directories to persist
+CONFIG_DIR="/home/coder/.config/code-server"
+LOCAL_DIR="/home/coder/.local/share/code-server"
+WORKSPACE_DIR="/home/coder/workspace"
+BACKUP_DIR="/tmp/hf_backup"
+
+# Create backup directory
+mkdir -p "$BACKUP_DIR"
+
+# Function to download from HF dataset
+download_from_hf() {
+    echo "Downloading persisted data from Hugging Face dataset: $HF_DATASET"
+    
+    # Try to download the backup archive
+    if huggingface-cli download "$HF_DATASET" backup.tar.gz --repo-type dataset --local-dir "$BACKUP_DIR" 2>/dev/null; then
+        echo "Found existing backup, restoring..."
+        
+        # Extract the backup
+        cd /home/coder
+        tar -xzf "$BACKUP_DIR/backup.tar.gz" 2>/dev/null || echo "No previous backup found or extraction failed"
+        
+        echo "Restoration complete!"
+    else
+        echo "No existing backup found. Starting fresh."
+    fi
+}
+
+# Function to upload to HF dataset
+upload_to_hf() {
+    echo "Saving data to Hugging Face dataset..."
+    
+    # Create backup archive
+    cd /home/coder
+    tar -czf "$BACKUP_DIR/backup.tar.gz" \
+        --exclude='*.log' \
+        --exclude='*.sock' \
+        --exclude='CachedData' \
+        --exclude='Cache' \
+        --exclude='GPUCache' \
+        .config/code-server \
+        .local/share/code-server \
+        workspace 2>/dev/null || true
+    
+    # Upload to HF dataset
+    huggingface-cli upload "$HF_DATASET" "$BACKUP_DIR/backup.tar.gz" backup.tar.gz --repo-type dataset --create || \
+        echo "Upload failed, continuing anyway..."
+}
+
+# Setup trap to save on exit
+trap 'upload_to_hf' EXIT SIGTERM SIGINT
+
+# Download existing data
+download_from_hf
+
+# Start background sync every 5 minutes
+(
+    while true; do
+        sleep 300
+        upload_to_hf
+    done
+) &
+
+SYNC_PID=$!
+
+echo "Starting code-server..."
+# Start code-server
+code-server --bind-addr 0.0.0.0:7860 --auth password /home/coder/workspace &
+CODE_SERVER_PID=$!
+
+# Wait for code-server to exit
+wait $CODE_SERVER_PID
+
+# Kill background sync
+kill $SYNC_PID 2>/dev/null || true
+
+# Final upload
+upload_to_hf
+
+echo "Shutdown complete."