ComfyUI Node
Atlas Assess Output 🧪
A ComfyUI node in Atlas/09 · QA & Gates with 14 inputs and 7 outputs.
Atlas Assess Output 🧪
- camera_view
- solve
- source_image
- depth
- report
- assessment_json
- json_path
- verdict
- image_provenance
- assessed_image
- evidence_path
◄solve_summary—►
◄enabledfalse►
◄providerlmstudio►
◄model►
◄base_url►
◄extra_instructions►
◄file_pathatlas_debug/output_assessment.json►
◄api_key►
◄offload_modeltrue►
◄fallback_to_sourcetrue►
CategoryAtlas/09 · QA & Gates
Inputs (14)
| Name | Type | Default | Description |
|---|---|---|---|
| camera_view | IMAGE | Final camera-view image. AtlasBlockoutViewport.shaded is valid only after Render Proxy Passes; the node detects its headless all-zero state. | |
| solve | ATLAS_SOLVE | — | |
| source_imageopt | IMAGE | Honest fallback for an unbaked/blank viewport pass. Projection checks remain inconclusive when this image is used. | |
| depthopt | ATLAS_DEPTH_MAP | — | |
| solve_summaryopt | STRING | Optional upstream debug/export summary appended to the internally derived scene-health JSON. | |
| enabledopt | BOOLEAN | false | Run the VLM. OFF still writes deterministic solve health. atlas_run_workflow(..., assess_output=True) enables every terminal assessor for agentic runs. |
| provideropt | COMBO | lmstudio | 4 options: ollama, lmstudio, llamacpp, openai |
| modelopt | STRING | — | |
| base_urlopt | STRING | — | |
| extra_instructionsopt | STRING | — | |
| file_pathopt | STRING | atlas_debug/output_assessment.json | Stable path, relative to ComfyUI's working directory. |
| api_keyopt | STRING | Saved in the workflow; prefer OPENAI_API_KEY. | |
| offload_modelopt | BOOLEAN | true | — |
| fallback_to_sourceopt | BOOLEAN | true | — |
Outputs (7)
| Name | Type | Description |
|---|---|---|
| report | STRING | — |
| assessment_json | STRING | — |
| json_path | STRING | — |
| verdict | STRING | — |
| image_provenance | STRING | — |
| assessed_image | IMAGE | — |
| evidence_path | STRING | — |