Troubleshooting

Symptoms: Unable to connect to database, connection timeout errors

Solutions:

• Verify connection string is correct (check for typos)
• Check firewall rules allow connections from Craveva servers
• For cloud databases (MongoDB Atlas, etc.), ensure IP whitelist includes Craveva IPs
• Verify database credentials (username/password)
• Check if database server is running and accessible
• For SSL connections, ensure SSL/TLS is enabled in connection settings
• Test connection from your local machine to verify database is reachable

Symptoms: API calls timing out, slow responses

Solutions:

• Check network connectivity
• Verify API endpoint URLs are correct
• Check if external service is experiencing downtime
• Review rate limits - you may be hitting API rate limits
• Increase timeout settings if appropriate
• Check firewall/proxy settings

Symptoms: Agent returns no response, timeout errors

Solutions:

• Check if agent is active and deployed
• Verify data source connection is working
• Review agent logs for error messages
• Check if LLM model is available and not rate-limited
• Verify agent configuration is correct
• Test agent in preview mode first
• Check billing/credits - insufficient credits can cause failures

Symptoms: Agent gives wrong answers, hallucinates data

Solutions:

• Review and improve agent prompt/instructions
• Verify data mapping is correct - agent may be querying wrong tables
• Check data source has the expected data
• Adjust temperature settings (lower for more factual responses)
• Add more context to agent prompt
• Test with sample queries to identify patterns
• Consider using a more capable LLM model

Symptoms: Slow responses, high latency

Solutions:

• Check database query performance - optimize slow queries
• Review data mapping - too many tables can slow down queries
• Consider using faster LLM models for simple tasks
• Check network latency to data source
• Review agent complexity - simplify if possible
• Monitor usage logs to identify bottlenecks

Symptoms: 401 Unauthorized, 403 Forbidden errors

Solutions:

• Verify API key is correct and not expired
• Check API key has required permissions/scopes
• Ensure API key is included in request headers
• Verify key format is correct (no extra spaces or characters)
• Check if key has been revoked or rotated
• Regenerate API key if needed

Symptoms: 429 Too Many Requests errors

Solutions:

• Implement exponential backoff in your code
• Reduce request frequency
• Check your usage limits in billing dashboard
• Upgrade your plan if you need higher limits
• Batch requests when possible
• Cache responses to reduce API calls

Symptoms: 400 Bad Request, validation errors

Solutions:

• Review API request format and required fields
• Check request body matches API schema
• Verify data types are correct (string vs number, etc.)
• Check for required fields that are missing
• Review API documentation for correct format
• Validate input data before sending requests

Symptoms: Connected but no data visible, empty results

Solutions:

• Wait a few minutes for initial data sync to complete
• Verify data source actually contains data
• Check data source permissions - user may not have read access
• Review data mapping - may be filtering out all data
• Check if tables/collections are correctly identified
• Verify connection is still active and not expired

Symptoms: Missing tables, incorrect field types

Solutions:

• Refresh schema analysis
• Check database user has permissions to view schema
• Verify table/collection names are correct
• For MongoDB, ensure collections have documents (empty collections may not show)
• Reconnect data source if schema seems outdated

Symptoms: Agent deployed but not responding on platform

Solutions:

• Verify deployment credentials are correct
• Check webhook URL is properly configured
• Ensure platform (Slack, WhatsApp, etc.) has verified webhook
• Review deployment logs for error messages
• Test agent in preview mode first
• Check if deployment platform is experiencing issues
• Verify agent is active and not paused

Symptoms: Webhook verification errors, messages not received

Solutions:

• Ensure webhook URL is publicly accessible (not behind firewall)
• Verify webhook URL uses HTTPS
• Check webhook endpoint is responding correctly
• Review webhook signature verification
• Ensure correct HTTP method (POST) is used
• Check platform-specific webhook requirements

Symptoms: Cannot log in, password reset not working

Solutions:

• Verify email and password are correct (check for typos)
• Check if account is locked or suspended
• Try password reset if forgotten
• Clear browser cache and cookies
• Try incognito/private browsing mode
• Check if 2FA is enabled and code is correct
• Contact support if account access is needed

Solutions:

• Optimize database queries - add indexes if needed
• Reduce data mapping complexity
• Use faster LLM models for simple queries
• Cache frequently accessed data
• Review and optimize agent prompts
• Check network latency to data sources

Symptoms: Operations failing, "insufficient credits" errors

Solutions:

• Check credit balance in Billing dashboard
• Add credits if balance is low
• Review usage to identify high-cost operations
• Optimize agent usage to reduce costs
• Set up budget alerts to prevent running out

Solutions:

• Review usage logs to identify high-cost operations
• Check which agents are consuming most credits
• Review LLM model usage - some models are more expensive
• Check for any loops or repeated operations
• Contact support for detailed usage breakdown