This page walks the upload flow for a custom model. For the concept overview see Models. The Models entry must appear in your sidebar — if not, this deployment doesn’t have custom models on your deployment.Documentation Index
Fetch the complete documentation index at: https://docs.cyberun.cloud/llms.txt
Use this file to discover all available pages before exploring further.
Before you start
- The file on disk you want to upload.
- A team you belong to. Any member can upload.
Upload
- Open Models in the sidebar.
- Click Upload model.
- Pick the file.
-
Fill in the metadata:
Field Notes Display name Human-readable label shown in parameter pickers. Up to 255 characters. Type ComfyUI models subdirectory name — checkpoints,loras,vae,controlnet, … Free-form string up to 50 characters; pick the directory your workflows expect.Path Required. Relative path inside the agent’s ComfyUI models directory (e.g. models/checkpoints/sdxl_base_1.0.safetensors). Must not be absolute and must not contain... - Click Upload. The dashboard uploads the file and registers the record; it then appears in the list.
What happens next
The file sits in platform storage. Each agent whose label set matches and that team owns the model pulls it into its local cache. The agent reportssyncing while the download runs. Once
cached, subsequent tasks on that agent reuse the local copy.
You don’t have to push the file to any specific machine.
Edit or delete
From a model’s row:- Edit — change the display name, type, path, or status
(
active/disabled). The file bytes themselves aren’t replaced through edit; upload a new record to ship different bytes. - Delete — remove the record. Agents keep the file cached locally until they evict it; new agents won’t see it.
Bulk delete
Multi-select rows in the list and use the bulk-actions bar.Troubleshooting
Upload succeeds but workflow can’t find the model — verify the Path matches what the workflow’s parameter expects. The path is exactly what the agent sees relative to its ComfyUI models directory. Agent stays onsyncing — the agent is downloading the model
on first use. Subsequent runs are fast.
Related
- Models — concept.
- Custom nodes — same pattern for node packages.
- Workflows — where models are referenced.
