Troubleshooting

Common issues and their solutions.

Data Sources

Data source isn't processing

Symptoms:

  • Status stuck at "Processing"

  • Process logs show errors

  • No new documents indexed

Solutions:

  1. Check Process Logs

    • Navigate to Data → [Your Data Source] → Logs

    • Look for specific error messages

    • Common issues: authentication, connectivity, format

  2. Verify Authentication

    • Re-authenticate if using OAuth (Google Drive, OneDrive)

    • Check API tokens haven't expired

    • Verify credentials are correct

  3. Test Connectivity

    • For websites: Ensure URL is publicly accessible

    • Check for robots.txt restrictions

    • Verify SSL certificates are valid

  4. Check File Formats

    • Ensure files are in supported formats

    • Try re-uploading files

    • Check file isn't corrupted

  5. Manual Retry

    • Click "Retry Processing"

    • If issue persists, delete and recreate

Confluence pages not syncing

Common Issues:

Authentication Expired:

  • Re-authenticate: Data → Confluence → Settings → Reconnect

Insufficient Permissions:

  • Ensure Confluence account has read access to spaces

  • Check space permissions in Confluence admin

Space Selection:

  • Verify correct spaces are selected in settings

  • Some pages may be restricted

Solution:

Google Drive files not appearing

Possible Causes:

  1. Shared Drive Access:

    • Enable "Shared Drives" during OAuth

    • Grant access to specific drives

  2. File Type Filtering:

    • Check file type settings

    • Ensure PDFs/Docs aren't excluded

  3. Folder Selection:

    • Verify correct folders are selected

    • Check folder permissions

Solution:

AI Responses

Agent can't answer questions

Symptoms:

  • "I don't have information about that"

  • No citations in response

  • Generic or off-topic answers

Diagnosis:

  1. Check Data Coverage

    • Search knowledge base for relevant content

    • Verify data sources contain the information

    • Check if documents processed successfully

  2. Verify Data Source Connection

    • Go to Agent settings

    • Confirm data sources are selected

    • Ensure sources are active (not disabled)

  3. Test Query Variations

  4. Review Citations

    • Check what sources the AI attempted to use

    • Verify relevance of cited documents

Solutions:

  • Add Missing Information: Create KB article or add data source

  • Improve Data Quality: Better document structure, clear headings

  • Try Different Strategy: Switch from Redwood to Cedar for ambiguous queries

  • Adjust Instructions: Provide more context in system prompt

Responses are inaccurate

Common Causes:

  1. Outdated Information

    • Data source hasn't synced recently

    • Old documents still in index

  2. Conflicting Information

    • Multiple sources with different information

    • Priority not set correctly

  3. Hallucination

    • AI generating unsupported information

    • Enable "Private Data Only" mode

Solutions:

  1. Refresh Data Sources

  2. Enable Private Data Only

  3. Review and Edit in Inbox

  4. Add Source Prioritization

Responses are too slow

Symptoms:

  • Response time > 5 seconds

  • Timeout errors

  • Poor user experience

Diagnosis:

Check response time breakdown:

  • Vector search: Should be < 200ms

  • LLM completion: Should be < 2s

  • Total: Should be < 3s

Solutions:

  1. Switch to Faster Strategy

  2. Reduce topK

  3. Enable Caching

  4. Use Streaming

  5. Check System Status

API Issues

Authentication failures

Error: 401 Unauthorized

Causes:

  • Invalid API key

  • Expired key

  • Incorrect header format

Solutions:

  1. Verify API Key

  2. Check Key Status

    • Settings → API Keys

    • Ensure key is active (not revoked)

  3. Generate New Key

    • Settings → API Keys → Generate New Key

    • Update your application

Rate limit exceeded

Error: 429 Too Many Requests

Headers:

Solutions:

  1. Implement Exponential Backoff

  2. Upgrade Plan

    • Free: 20 req/min → Pro: 100 req/min

    • Contact sales for higher limits

  3. Optimize Request Patterns

    • Batch similar requests

    • Cache frequent queries

    • Reduce polling frequency

Invalid request format

Error: 400 Bad Request

Common Issues:

  1. Missing Required Fields

  2. Incorrect Data Types

  3. Invalid JSON

    • Validate JSON before sending

    • Check for trailing commas

    • Ensure proper escaping

Integration Issues

Browser extension not working

Symptoms:

  • Extension icon is grayed out

  • No response when clicking

  • "Not logged in" error

Solutions:

  1. Check Login Status

    • Click extension icon

    • Click "Login" if not authenticated

    • Allow pop-ups if blocked

  2. Update Extension

    • Chrome Web Store → Twig AI → Update

    • Restart browser

  3. Clear Extension Data

    • Right-click extension icon → Options

    • Click "Clear Data"

    • Log in again

  4. Check Permissions

    • Extension may need additional permissions

    • Review and accept prompts

Slack integration not responding

Symptoms:

  • Bot doesn't reply in channels

  • Commands not recognized

  • "Bot not installed" message

Solutions:

  1. Verify Installation

    • Slack workspace → Apps

    • Check Twig AI is installed

    • Reinstall if missing

  2. Check Channel Permissions

    • Invite bot to channel: /invite @TwigAI

    • Verify bot has read/write permissions

  3. Test with Direct Message

    • DM @TwigAI to test bot

    • If working, issue is channel-specific

  4. Reconnect Integration

    • Twig AI → Integrations → Slack

    • Click "Reconnect"

    • Authorize again

Zendesk app not loading

Symptoms:

  • Sidebar empty

  • "Failed to load" error

  • App not appearing

Solutions:

  1. Refresh Ticket

    • Reload the ticket page

    • Clear browser cache

  2. Check App Installation

    • Zendesk Admin → Apps → Manage

    • Verify Twig AI is enabled

    • Check configuration settings

  3. Verify API Connection

    • Twig AI → Settings → Integrations → Zendesk

    • Test connection

    • Update credentials if needed

  4. Browser Console Errors

    • Open DevTools (F12)

    • Check Console tab for errors

    • Report to support if needed

Performance Issues

High memory usage

Symptoms:

  • Browser tab using lots of RAM

  • Slow UI responsiveness

  • Tab crashes

Solutions:

  1. Limit Open Conversations

    • Close old Playground sessions

    • Clear conversation history

  2. Disable Streaming

    • Use non-streaming mode for lower memory

    • Streaming keeps connection open

  3. Use Dedicated App

    • Desktop app (if available) uses less memory

    • Or use API instead of web UI

Slow dashboard loading

Symptoms:

  • Analytics page takes > 10s to load

  • Charts not rendering

  • Timeouts

Solutions:

  1. Reduce Date Range

    • Select shorter time period (e.g., 7 days vs 90 days)

    • Large datasets take longer to query

  2. Clear Cache

    • Browser cache may be corrupted

    • Hard refresh: Ctrl+Shift+R (Cmd+Shift+R on Mac)

  3. Check Data Volume

    • Large organizations may need longer load times

    • Contact support for optimization

Still Having Issues?

Collect Diagnostic Information

Before contacting support, gather:

  1. Error Details

    • Full error message

    • Error code

    • Screenshot if visual issue

  2. Request Details

    • Timestamp

    • Request ID (from API response headers)

    • User ID / Organization ID

  3. Environment

    • Browser type and version

    • Operating system

    • Network connection type

  4. Steps to Reproduce

    • What you were doing

    • What you expected to happen

    • What actually happened

Contact Support

Emergency Support (Enterprise Only)

For production-critical issues:

  • Phone: Provided in enterprise agreement

  • Slack Connect: Direct channel to support team

  • Response Time: 4 hours guaranteed

Last updated