본문으로 건너뛰기

Integration Assistant Troubleshooting

Common issues and solutions for the Integration Assistant feature.

MCP Connection Issues​

"fluxloop-mcp package is not installed"​

Cause: The fluxloop-mcp Python package is missing.

Solution:

pip install fluxloop-mcp

Then click Refresh Integration View or restart VSCode.

"MCP Index not found"​

Cause: Knowledge index hasn't been built yet.

Solution:

# Option 1: Use the MCP command
fluxloop-mcp rebuild-index

# Option 2: Run the script (if you have the repo)
cd packages/mcp
./scripts/rebuild_index.sh

The index will be created at ~/.fluxloop/mcp/index/dev.

"Python Environment not available"​

Cause: Python 3.11+ is not installed or not in PATH.

Solution:

  1. Install Python 3.11 or newer
  2. Verify: python3 --version
  3. Restart VSCode
  4. Click Refresh Integration View

MCP Process Fails to Start​

Cause: Python environment issues or corrupted installation.

Solution:

# Reinstall fluxloop-mcp
pip uninstall fluxloop-mcp
pip install fluxloop-mcp

# Verify installation
fluxloop-mcp --version

Check the FluxLoop output channel for detailed error messages.

OpenAI API Issues​

"OpenAI API key is required"​

Cause: No API key configured.

Solution:

  1. Get your API key from OpenAI Platform
  2. Store in VSCode settings: 
    {
      "fluxloop.openaiApiKey": "sk-..."
    }
  3. Or enter when prompted (can be saved to Secret Storage)

"Invalid API key" or "Unauthorized"​

Cause: API key is incorrect or expired.

Solution:

  1. Verify your key at OpenAI Platform
  2. Update in VSCode settings or Secret Storage
  3. Try again

"Rate limit exceeded"​

Cause: Too many API requests.

Solution:

  • Wait a few minutes before retrying
  • Use a different model (e.g., gpt-4o-mini instead of gpt-4o)
  • Check your OpenAI usage limits

API Call Timeout​

Cause: Network issues or slow model response.

Solution:

  • Check your internet connection
  • Switch to a faster model (gpt-4o-mini)
  • Try again in a few moments

Flux Agent Issues​

"Open a file and place the cursor where you need guidance"​

Cause: No file is open in the editor.

Solution:

  1. Open a source file (e.g., src/server.ts, app/main.py)
  2. Optionally select a code block
  3. Run Flux Agent again

"Open a FluxLoop project workspace"​

Cause: No workspace folder is open.

Solution:

  1. Open a folder containing your FluxLoop project
  2. Ensure configs/ directory exists
  3. Run Flux Agent again

Suggestions Are Too Generic​

Cause: Insufficient context or selection.

Solution:

  1. Select specific code: Highlight the exact function/class you want to enhance
  2. Open main files: Open entry points like server.ts or main.py
  3. Add project context: Ensure package.json/requirements.txt exist
  4. Try Knowledge Search first: Get framework-specific info before running agent

Suggestions Don't Match My Framework​

Cause: MCP analysis incorrectly detected framework.

Solution:

  1. Check System Status → MCP Index is ready
  2. Verify your project has clear framework indicators:
    • Express: package.json with "express": "..."
    • FastAPI: requirements.txt with fastapi
    • Next.js: package.json with "next": "..."
  3. Rebuild index if recipes were recently updated: 
    fluxloop-mcp rebuild-index

Performance Issues​

Slow Response Times​

Cause: Large files or complex analysis.

Solutions:

  • Select smaller code blocks instead of entire files
  • Use faster models: Switch to gpt-4o-mini
  • Check MCP index: Ensure it's not rebuilding in the background

Extension Becomes Unresponsive​

Cause: Background process stuck.

Solution:

  1. Check FluxLoop output channel for errors
  2. Restart VSCode
  3. Verify Python processes: 
    ps aux | grep fluxloop-mcp
  4. Kill any stuck processes if needed

Data and Privacy​

Where is Data Stored?​

  • MCP Index: ~/.fluxloop/mcp/index/dev (local only)
  • Suggestions: VSCode workspace state (local only)
  • API Keys: VSCode Secret Storage (encrypted) or workspace settings

What Data is Sent to OpenAI?​

When you run Flux Agent, the following is sent to OpenAI API:

  • Selected code snippet or file content (truncated to ~4000 chars)
  • MCP analysis results (framework detection, file paths)
  • Your structured prompt

Not sent:

  • Full codebase
  • Environment variables or secrets
  • Private repository metadata

Getting Help​

If you encounter issues not covered here:

  1. Check the FluxLoop output channel for detailed logs
  2. Review VSCode Extension Troubleshooting
  3. Visit GitHub Issues
  4. Check MCP Server Docs for protocol-level debugging

Next Steps​