VS Code + Gemini Code Assist Integration

Google's Gemini Code Assist for VS Code provides powerful AI-powered coding assistance. When combined with Eclair, it becomes a data-aware development environment that can discover, download, and analyze datasets directly within your IDE.

Prerequisites

Before setting up VS Code with Gemini Code Assist and Eclair:

• ✅ Eclair server is installed and running
• ✅ Visual Studio Code is installed
• ✅ Google Cloud account with Gemini API access
• ✅ Gemini API key

Installation & Setup

1. Install Gemini Code Assist Extension

Install the Gemini Code Assist extension from the VS Code marketplace:

# Install via command line
code --install-extension Google.geminicodeassist

Or search for "Gemini Code Assist" in the VS Code Extensions panel.

2. Enable Agentic Mode

To use MCP servers with Gemini Code Assist, you need to enable Agentic mode:

1. Open VS Code
2. Press Ctrl+Shift+P (or Cmd+Shift+P on Mac)
3. Type > Open User Settings JSON
4. Add this configuration:

{
   "geminicodeassist.updateChannel": "Insiders"
}

1. Restart VS Code

3. Configure API Key

Set your Gemini API key in an .env file in your project directory:

echo "GEMINI_API_KEY=your_actual_api_key_here" >> .env

Keep Your API Key Secure

Never commit API keys to version control. Use environment files or secure key management.

4. Configure MCP Server

Create the Gemini configuration file at ~/.gemini/settings.json:

mkdir -p ~/.gemini
cat > ~/.gemini/settings.json << 'EOF'
{
  "mcpServers": {
    "eclair": {
      "httpUrl": "http://localhost:8080/mcp",
      "timeout": 5000
    }
  },
  "selectedAuthType": "gemini-api-key"
}
EOF

5. Copy System Prompt

Copy the system prompt to help Gemini understand Eclair's capabilities:

cp src/eclair/client/gemini/gemini.md ./GEMINI.md

6. Activate Gemini Sidebar

1. Restart VS Code
2. Open the Gemini Sidebar (sparkle ✨ icon)
3. Set Gemini to Agentic mode
4. Verify MCP server connection by typing /mcp

Gemini Code Assist + Eclair is ready! You now have powerful AI-assisted dataset discovery and analysis directly in VS Code.

Usage Examples

Fashion Dataset Analysis Example

Find a fashion-related dataset and visualize some data examples

Step 1: Automatic Tool Usage Gemini immediately uses Eclair's search-datasets tool:

Step 2: Notebook Generation Gemini creates a complete Jupyter notebook with:

• Proper data loading code using Eclair's instructions
• Comprehensive data analysis
• Multiple visualization types

Step 3: Visualization Results The generated notebook produces the requested visualizations:

Further examples

Try these capabilities as well

1. Intelligent Dataset Discovery

Gemini understands context and can find datasets based on: - Domain (computer vision, NLP, time series, etc.) - Size requirements - Data format preferences - Specific use cases

2. Automatic Code Generation

Based on Eclair's metadata, Gemini generates: - Data loading scripts with proper error handling - Preprocessing pipelines - Visualization code - Analysis workflows - Complete Jupyter notebooks

3. Interactive Development

• Ask follow-up questions about datasets
• Request modifications to generated code
• Get explanations for data analysis steps
• Iterate on visualizations and analysis

4. Context Awareness

Gemini maintains context about: - Previously discussed datasets - Your project requirements - Code that's already been generated - Analysis results

Troubleshooting

Common Issues

"MCP servers not loaded"

• Ensure ~/.gemini/settings.json is properly configured
• Verify Eclair server is running: curl http://localhost:8080/mcp
• Restart VS Code and check Agentic mode is enabled

"API key not found"

• Check .env file contains GEMINI_API_KEY=your_key
• Verify the API key is valid and has proper permissions
• Ensure the .env file is in your project root

"Dataset access denied"

• Some datasets require authentication
• Check dataset license and access requirements
• Try alternative datasets with open access

"Code execution errors"

• Install required dependencies: pip install datasets pandas matplotlib
• Check Python environment is properly configured
• Verify dataset URLs are still valid

Debug Steps

1. Verify Eclair Connection:

   curl http://localhost:8080/mcp/health
   eclair-client --tool ping
   

2. Check Configuration Files:

   cat ~/.gemini/settings.json
   cat .env
   ls -la GEMINI.md
   

3. Test MCP Integration: In Gemini Code Assist, type /mcp to verify tools are loaded

Configuration Issues

If MCP servers aren't loading:

1. Check Settings Format:

   {
     "mcpServers": {
       "eclair": {
         "httpUrl": "http://localhost:8080/mcp",
         "timeout": 5000
       }
     },
     "selectedAuthType": "gemini-api-key"
   }
   

2. Verify Server Status: Eclair server must be running before starting VS Code