troubleshooting

1# ComfyUI Troubleshooting Guide
2 
3> Render completes but looks WRONG (artifacts, wrong subject/pose/color, a
4> ControlNet/mask/LoRA not taking, a refiner degrading it)? That's not an error.
5> Use the debug-render skill (`list_packs` with `action: "skill_read"`,
6> `name: "debug-render"`) to localize the bad stage with run-to-node
7> (`panel_run` `to_node_id`) by previewing intermediate steps. This guide is for
8> runs that fail with an error, OOM, or missing node.
9 
10## Error Diagnosis Strategy
11 
12When a workflow fails, follow this approach:
13 
141. Get the error. Use `get_history(action="diagnose")` to retrieve the execution result with the full traceback, plus any missing models/nodes
152. Check logs. Use `get_system_stats (action:"logs")` with keyword filters like `"error"`, `"warning"`, `"traceback"`
163. Identify the failing node. The history response includes the `node_id` and `node_type` that failed
174. Cross-reference inputs. Use `create_workflow (action:"node_info")` to verify the failing node's expected input schema
185. Check models. Use `list_local_models` to verify all referenced model files exist
19 
20## Out of Memory (OOM)
21 
22### Error Pattern
23 
24```
25torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate X MiB.
26GPU 0 has a total capacity of 24.00 GiB of which X MiB is free.
27```
28 
29Or:
30 
31```
32RuntimeError: CUDA error: out of memory
33```
34 
35### Root Cause
36…