Debug View
Inspect workflow execution tree and call-level details.
-
Execution Tree
Navigate workflow structure with expandable nodes
-
Call Details
View inputs, outputs, commands, and logs
-
Timeline Analysis
Visualize task durations and identify bottlenecks
-
Resource Efficiency
Analyze CPU, memory, and disk utilization
Requirement
Requires a working connection to your Cromwell server (set via --host or CROMWELL_HOST).
Quick Start
Navigation
| Key | Action |
|---|---|
| Up / Down | Navigate tree |
| Right / Left | Expand / Collapse |
| Enter | Toggle expand |
| Tab | Switch panel focus |
| Shift+E | Expand all nodes |
| Shift+C | Collapse all nodes |
| Y | Copy to clipboard |
| A | Chat with AI |
| Q | Back / Quit |
Quick Actions
Quick actions are context-sensitive and depend on the selected node type. Press number keys 1 to 6 to trigger actions.
Workflow / SubWorkflow Nodes
| Key | Action | Description |
|---|---|---|
| 1 | Inputs | Open modal with workflow inputs |
| 2 | Outputs | Open modal with workflow outputs |
| 3 | Options | View submitted workflow options |
| 4 | Timeline | Open timeline modal (tasks sorted by duration) |
| 5 | Workflow Log | Load and display workflow log |
Task / Shard Nodes
| Key | Action | Description |
|---|---|---|
| 1 | Inputs | Open modal with task inputs |
| 2 | Outputs | Open modal with task outputs |
| 3 | Command | View executed command |
| 4 | Logs | Switch to logs view (stdout/stderr/monitoring) |
| 5 | Efficiency | Analyze resource usage (requires monitoring script) |
| 6 | AI Chat | Open chat with selected context data |
Copy to Clipboard
In modals, press Y to copy content to clipboard.
AI Chat Integration
Press 6 or A on a Task or Shard node to open the AI chat with task context.
Context Selection
Before starting the chat, you'll be prompted to select which data to include in the AI context:
| Option | Description |
|---|---|
| Metadata | Task info, status, timing, command |
| Stderr | Standard error output (last 200 lines) |
| Stdout | Standard output (last 200 lines) |
| Monitoring | Resource efficiency analysis |
| Batch Logs | Google Batch logs (requires Google Batch backend) |
Debugging Failed Tasks
When debugging a failed task, select Metadata + Stderr for the most relevant context. The AI can help interpret error messages and suggest fixes.
Selection Controls
| Key | Action |
|---|---|
| Up / Down | Navigate options |
| Space | Toggle selection |
| Enter | Confirm and start chat |
| Esc | Cancel |
Timeline Analysis
Press 4 on a Workflow or SubWorkflow node to open the timeline modal.
The timeline shows:
- Tasks sorted by duration (longest first)
- Visual timeline bars showing when each task ran relative to workflow start
- Time range for each task (start → end)
- Status icons (✓ Done, ● Running, ✗ Failed, ↺ Preempted)
Subworkflow Timelines
Navigate to a subworkflow node and press 4 to see its specific timeline, separate from the parent workflow.
Resource Efficiency
Press 5 on a Task or Shard node to analyze resource utilization.
Requirement
This feature requires the monitoring script to be configured in Cromwell. See Resource Monitoring.
Displayed metrics:
| Metric | Description |
|---|---|
| CPU | Peak and average usage vs. allocated cores |
| Memory | Peak usage vs. allocated memory |
| Disk | Peak usage vs. allocated disk space |
| Efficiency % | Visual gauge showing utilization percentage |
Cost Optimization
Low efficiency indicates over-provisioned resources. Consider reducing CPU/memory/disk allocations for tasks with < 50% efficiency.
Preemption Summary
For Workflow and SubWorkflow nodes running with preemptible instances, the details panel shows:
- Cost Efficiency — Overall efficiency considering preemption overhead
- Preemptible/Total Tasks — How many tasks used preemptible instances
- Total Attempts — Including retries after preemptions
- Total Preemptions — Number of times tasks were preempted
Problematic Tasks
Tasks with high preemption impact are highlighted:
- Tasks with < 70% cost efficiency
- Tasks where preemption caused > 10% cost overhead
Subworkflows
Preemption stats are calculated per-level. Navigate into subworkflows to see their individual preemption analysis.
Scatter/Shard Summary
For Call nodes with multiple shards (scatter operations), the panel shows:
- Total Shards count
- Status Breakdown — Done, Running, Failed, Preempted counts
- Timing Statistics — Wall clock, min/max/avg per-shard duration
Expand the node to navigate individual shards.