Skip to main content

Troubleshooting

PR Generation Issues​

PR Has Merge Conflicts​

The generated PR has conflicts with the target branch.

Solutions:

  1. Click Regenerate on the stale flag row to create a fresh PR
  2. Pull the branch and resolve conflicts manually: 
    git checkout janitor/remove-{flag-key}
    git merge main
    # Resolve conflicts
    git push
  3. If the flag has changed significantly, dismiss it and re-scan

LLM Analysis Returned Low Confidence​

The AI is unsure about safely removing this flag.

Why this happens:

  • Complex nested conditionals (3+ levels deep)
  • Dynamic flag references (e.g., isEnabled(variableName))
  • Flag used in test assertions
  • Unusual language patterns

Solutions:

  1. Review the analysis notes for specific concerns
  2. Manually inspect the flagged references
  3. Consider refactoring the flag usage to be more analyzable
  4. Dismiss if you've confirmed it's safe manually

Connection Issues​

Repository Not Found​

Causes:

  • Token doesn't have access to the repository
  • Repository was renamed or deleted
  • Organization access not granted

Solutions:

  1. Verify the token has access to the repository
  2. Reconnect the provider in Settings → Integrations
  3. Check that the repository exists and is accessible

Token Expired​

Solution: Reconnect the provider in Settings → Integrations

LLM Provider Unavailable​

Causes:

  • API key expired or invalid
  • Rate limit exceeded
  • Provider outage

Solutions:

  1. Check your API key in Settings → Integrations → LLM Provider
  2. Wait for rate limit to reset (typically 1 minute)
  3. The system will automatically fall back to regex analysis

Analysis Issues​

Flag is Referenced in Unexpected Ways​

Examples:

  • Dynamic key construction: isEnabled("prefix-" + flagName)
  • String interpolation in templates: `${isEnabled("flag") ? A : B}`
  • Flag stored in a variable: const isEnabled = boolVariation("flag", user)

Solutions:

  1. The LLM can handle many of these cases — review the analysis
  2. If the LLM also misses it, manually inspect the code
  3. Consider standardizing flag usage patterns across your codebase

Generated PR Breaks Tests​

Causes:

  • Test files reference the removed flag
  • The removed conditional had side effects

Solutions:

  1. Check if test files were correctly updated
  2. Review the diff for any missed references
  3. Manually fix the test assertions
  4. Update the test to not depend on the removed flag

Compliance Issues​

LLM Processing Disabled by Policy​

Your organization's compliance policy has disabled LLM processing.

What happens:

  • The Janitor falls back to regex-based analysis
  • Analysis confidence is lower (35%)
  • PRs include "Manual Review Required" warnings

To enable LLM:

  1. Go to Settings → Integrations → Compliance
  2. Set mode to Approved or Strict
  3. Configure approved providers

Data Masking is Blocking Analysis​

Redaction rules are too aggressive and are redacting valid code patterns.

Solutions:

  1. Review your redaction rules in Compliance settings
  2. Disable specific rules that are too broad
  3. Use more specific regex patterns

Getting Help​

If you're still having issues: