Common Issues
Solutions to frequently encountered problems
This page covers solutions to frequently encountered problems in Agents.
Workflow Not Triggering
Symptoms
- Workflow never runs
- No executions appear
- Events seem to be ignored
Causes and Solutions
Workflow Not Released
Check: Workflow status shows "Draft" instead of "Active"
Solution:
- Open the workflow editor
- Click "Release"
- Confirm the release
Wrong Event Type
Check: Event type doesn't match expected events
Solution:
- Verify the trigger event type
- Ensure it matches when you expect the workflow to run
- For meetings, check if you need
MEETING_ENDEDvsMEETING_ENDED_FOR_DEAL_ROOM
Trigger Rules Filtering Events
Check: Trigger has filter rules that exclude events
Solution:
- Review trigger configuration
- Check any filter conditions
- Temporarily remove filters to test
No Matching Events
Check: No events have occurred since release
Solution:
- Use MANUAL trigger to test
- Wait for next real event
- Verify events are being generated
Organization Mismatch
Check: Events are for a different organization
Solution:
- Verify workflow is in correct organization
- Check user permissions
Verification Steps
- Check workflow is released and Active
- View workflow trigger configuration
- Check execution list for any attempts
- Use manual trigger to test
Node Execution Failures
Symptoms
- Node shows error status
- Workflow stops at a specific node
- Error messages in execution log
Common Causes
Timeout
Error: "Execution timed out"
Causes:
- Operation takes too long
- External service slow
- Too much data to process
Solutions:
- Simplify the operation
- Use summary instead of full transcript
- Check external service status
- Consider splitting into multiple nodes
Invalid Configuration
Error: "Invalid configuration" or validation errors
Causes:
- Missing required fields
- Invalid values
- Template syntax errors
Solutions:
- Review all required fields
- Check for typos in configuration
- Validate Liquid templates
- Verify CEL expressions
Missing Input Data
Error: "Property not found" or null errors
Causes:
- Upstream node didn't produce expected output
- Data structure different than expected
- Conditional path skipped a node
Solutions:
- Check upstream node output
- Add null checks to conditions
- Verify data flow through workflow
Retry Exhaustion
Error: "Max retries exceeded"
Causes:
- Persistent failure condition
- External service down
- Rate limiting
Solutions:
- Check external service status
- Review error from first attempt
- Increase retry limit if appropriate
- Add longer backoff
Integration Problems
Symptoms
- Integration nodes fail
- Authentication errors
- "Connection failed" messages
HubSpot Issues
OAuth Token Expired
Error: "Invalid access token" or "401 Unauthorized"
Solution:
- Go to Settings → Integrations
- Disconnect HubSpot
- Reconnect with fresh OAuth
Missing Permissions
Error: "Insufficient permissions"
Solution:
- Review required HubSpot scopes
- Reconnect with correct permissions
- Verify HubSpot user has admin access
Object Not Found
Error: "Object not found" or "404"
Causes:
- Deal/contact ID is invalid
- Object was deleted
- Wrong object type
Solution:
- Verify the object ID exists in HubSpot
- Check object type matches
Salesforce Issues
Session Expired
Error: "Session expired or invalid"
Solution:
- Reconnect Salesforce integration
- Ensure Salesforce user account is active
Field Access Denied
Error: "Field not accessible"
Solution:
- Check Salesforce field-level security
- Ensure connected user has field access
- Use different field if restricted
Slack Issues
Channel Not Found
Error: "Channel not found"
Causes:
- Channel ID is wrong
- Bot not added to channel
- Private channel access
Solution:
- Verify channel ID
- Add Slack bot to the channel
- Use channel picker if available
Rate Limited
Error: "Rate limited" or "429"
Solution:
- Add Wait nodes between messages
- Reduce message frequency
- Batch multiple notifications
AI Node Issues
Symptoms
- AI output is wrong
- Parsing errors
- Unexpected results
Common Problems
Output Parse Failure
Error: "Failed to parse output" or type mismatch
Causes:
- AI response doesn't match return type
- Inconsistent output format
Solutions:
- Add explicit format instructions to prompt
- Use simpler return type
- Include examples in prompt
- Try higher model tier
Hallucination
Symptoms: AI includes information not in input
Solutions:
- Add "Only include information from the provided content"
- Be more specific about sources
- Use "If not mentioned, say 'Not discussed'"
Too Long / Too Short
Symptoms: Output length not as expected
Solutions:
- Add explicit length constraints
- "Maximum 3 bullet points"
- "Exactly 2-3 sentences"
- "Under 100 words"
Wrong Focus
Symptoms: AI focuses on wrong aspects
Solutions:
- Be more specific about what to include/exclude
- Add "Focus on:" section
- List items in priority order
- Add "Do not include:" section
Model Tier Too Low
Symptoms: Poor quality for complex tasks
Solutions:
- Upgrade to medium or high tier
- Simplify the task
- Break into multiple simpler prompts
AI Timeout
Error: "AI request timed out"
Causes:
- Input too large
- Complex task
- Model overloaded
Solutions:
- Use transcript summary instead of full transcript
- Simplify prompt
- Reduce input size
- Try at different time
Data Flow Issues
Symptoms
- Wrong data reaching nodes
- Missing data in output
- Unexpected null values
Common Problems
Referencing Wrong Node Output
Cause: Using json when data is from different step
Solution:
- Use
steps.nodeName.valuefor specific nodes - Check data flow in execution log
- Verify node connections
Array vs Single Item
Cause: Expecting array but getting item, or vice versa
Solutions:
- Check node variant (per-item vs batch)
- Use Select Many to flatten arrays
- Use proper Liquid iteration
Null in Chain
Cause: Null value breaks property chain
Solutions:
- Add null checks in conditions
- Use If nodes to handle null cases
- Provide defaults
Data Lost Between Nodes
Cause: Node doesn't pass through all data
Solutions:
- Check what the node outputs
- Use Zip to combine data
- Use Broadcast to attach context
Performance Issues
Symptoms
- Workflows take too long
- Timeouts
- High costs
Solutions
Reduce Input Size
// Use summary instead of full transcript
{{ json.callRecording.transcriptSummary }}
Use Appropriate Model Tier
| Task | Tier |
|---|---|
| Simple classification | Low |
| Standard analysis | Medium |
| Complex reasoning | High |
Optimize Workflow Structure
- Remove unnecessary nodes
- Combine operations where possible
- Use parallel branches effectively
Add Caching
- Don't re-process unchanged data
- Store results for reuse
- Use conditional execution
Release Issues
Can't Release Workflow
Causes:
- Validation errors in workflow
- Disconnected nodes
- Missing required configuration
Solution:
- Check for validation errors
- Ensure all nodes are connected
- Review required fields in each node
- Check trigger configuration
Release Doesn't Take Effect
Causes:
- Old version still Active
- Cache issues
- Timing
Solution:
- Verify workflow version status is Active
- Wait a few minutes
- Check which version is Active
- Try releasing again
Recovery Steps
General Recovery Process
- Identify the issue - Check execution logs
- Understand the cause - Review error message
- Fix the configuration - Update workflow
- Test the fix - Use manual trigger
- Release - Make the fixed version Active
- Monitor - Watch next few executions
Recovering from DLQ
- Fix the root cause first
- Review DLQ messages
- Retry messages if appropriate
- Discard if obsolete
Rolling Back
- Identify last working version
- Release that version
- Monitor to confirm fix
- Investigate issue in new version