Skip to main content

Troubleshooting

Common issues and how to fix them.

Agent Issues​

Agent times out before completing task​

Symptoms: Task stops with timeout error, agent was making progress.

Causes:

  • Task too complex for timeout setting
  • Agent stuck in reflection loop
  • Slow LLM responses

Solutions:

  1. Increase timeout in settings
  2. Break task into smaller steps
  3. Use faster model (gpt-5-mini)

Agent goes in loops​

Symptoms: Same actions repeated, token usage spikes.

Causes:

  • Unclear task description
  • Page content confusing the agent
  • Reflection not accepting valid results

Solutions:

  1. Be more specific in your request
  2. Check page content structure
  3. Try simpler wording

"Cannot read properties of undefined"​

Symptoms: Console shows property access errors.

Causes:

  • Page not fully loaded
  • Element indices changed after action

Solutions:

  1. Add wait between actions
  2. Re-fetch page content before interactions

Extension Issues​

Extension not loading​

Symptoms: No Vibe icon in toolbar, side panel empty.

Solutions:

  1. Check chrome://extensions for errors
  2. Verify Developer mode is enabled
  3. Reload the extension

API key not working​

Symptoms: "Invalid API key" or authentication errors.

Solutions:

  1. Verify key is correct (no extra spaces)
  2. Check key has sufficient quota
  3. For OpenAI: ensure key is not restricted

Side panel won't open​

Symptoms: Click icon but panel stays closed.

Solutions:

  1. Right-click icon → "Open side panel"
  2. Use keyboard shortcut Ctrl+Shift+V
  3. Check for conflicting extensions

CLI Issues​

"OPENAI_API_KEY not set"​

Solution:

export $(< .env)
# or
export OPENAI_API_KEY=sk-your-key

Browser window not appearing​

Symptoms: CLI runs but no browser opens.

Solutions:

  1. Check HEADLESS is not set
  2. Install Playwright browsers: npx playwright install chromium

Timeout on complex tasks​

Solution: Increase timeout:

node cli.js --timeout 300 run "complex task"

Performance Issues​

Slow tool execution​

Causes:

  • Page content too large
  • Too many page elements
  • Network latency

Solutions:

  1. Scroll to relevant section first
  2. Use specific element indices
  3. Enable content truncation

High token usage​

Causes:

  • Large page content
  • Long conversation history
  • Reflection loops

Solutions:

  1. Start fresh conversation for new tasks
  2. Use gpt-5-mini instead of gpt-4
  3. Simplify task descriptions

Debug Mode​

Enable verbose logging:

# CLI
node cli.js --verbose run "task"

# Extension - check console
# Open DevTools → Console → filter "vibe"

Reading Agent Logs​

# Find LLM calls
grep "LLM invoke" console.log

# Find tool executions
grep "Tool execution" console.log

# Find errors
grep -E "Error:|exception" console.log

Getting Help​

If issues persist:

  1. Check GitHub Issues
  2. Include:
    • Error message
    • Console logs
    • Steps to reproduce
  3. For extension: screenshot of side panel
  4. For CLI: full command used