Collected data
See, preview, download and delete the files, records and inference results field devices gathered, across every device
Required permission: View
This screen shows the raw material field devices (Edge) gathered and uploaded, across the whole tenant. Edge Fleet looks at one device at a time; this screen brings together what every device sent so you can find it and tidy it up.
Collected data is kept apart from datasets. When a device uploads a file without naming a dataset, the file does not go into a training dataset; it lands here. Moving collected files into a dataset is not available yet.
Every member can view; deleting files needs dataset_write (DEVELOPER, OPERATOR, ADMIN). If the Edge Fleet feature is off for the tenant, this screen is hidden too.
What goes where
| What the device sends | Where you see it | How long it is kept |
|---|---|---|
A file with no dataset (the watchdir collector, the local API /api/v1/blobs) | the Files (파일) tab here | until deleted (or the retention an administrator set) |
A file with a dataset (dataset_id) | that dataset's files | follows the dataset |
Records from collectors (http, push, modbus) | the Records (레코드) tab here and on the device detail | 90 days after arrival (default) |
| Results of inference run on the device | the Inference results (추론 결과) tab here and on the device detail | 90 days after arrival (default) |
Deleting a device keeps the collected files it uploaded, and they still show which device sent them.
Files tab
use-scr-collected-files| No. | Area | Description |
|---|---|---|
| ① | Summary | Number of files, size, devices that uploaded files, file retention (kept indefinitely or deleted after N days). |
| ② | Tabs | Files (파일) · Records (레코드) · Inference results (추론 결과). |
| ③ | Filters | File name search, device, collector, received date (from–to, the end date included). Filters stay in the address, so they survive a visit to the detail page. |
| ④ | View toggle | Table view and grid view (images as thumbnails). |
| ⑤ | Table | File name, device (opens the device detail), collector, size, received time. Click a header to sort. |
An empty collector (—) means the file came in through the device's local API. Received is when the server received it; a file the device held while offline and uploaded later may have been captured much earlier.
Deleting files
- In the table view, tick the files to delete. The box in the header ticks the whole page.
- Press Delete selected (선택 삭제) that appears above the table.
- The button only enables once you type
삭제exactly. Confirming removes the original from storage.
File detail
use-scr-collected-fileClick a file to open its detail: the preview on top, then the origin (device, collector, original path on the device, received time, size, SHA-256, upload id) and the metadata the device sent. Download (다운로드) at the top right fetches the original; Delete (삭제) enables only after you type the file name.
Previews are drawn from the original only, because collected files are not validated like dataset files and have no downsampled copy.
| Format | Preview |
|---|---|
| Images (PNG, JPEG, …) | the original image |
| PLY point clouds | 3D viewer up to 50 MB; download only above that |
| JSON, CSV, TSV | the content up to 2 MB; download only above that |
| Anything else | download only |
Records and inference results tabs
use-scr-collected-recordsAll devices' records and inference results, newest first: the same tables as on the device detail with a device column added. The device and date filters are shared with the files tab, so you can look at the same device and period while switching tabs. Records can be narrowed further by kind and inference results by model name (exact match).
Payloads and results are shortened to one line; Details (자세히) expands the full JSON. Load more (더 보기) at the bottom loads older entries.
Records and inference results are deleted after a set period (90 days by default), which the tab states at the top. Upload anything you need to keep as a file.