Skip to main content

Documentation Index

Fetch the complete documentation index at: https://avala.ai/docs/llms.txt

Use this file to discover all available pages before exploring further.

There are two ways to upload local files into Avala: the Add Dataset wizard in the web app (drag-and-drop) and the avala datasets upload CLI command. Both write to the same managed storage, both enforce the same per-user 10 GB cap, and both create a normal Avala dataset that you can label, share, and export. Use the wizard for ad-hoc uploads of a few hundred files. Use the CLI when you have many files, large files, or want to script the upload.

​
Storage cap

Local uploads count against a 10 GB per-user quota. Each individual file is capped at 2 GB. Files larger than 2 GB are not currently supported via either the wizard or the CLI — split them, or use one of the cloud-storage data sources instead. You can see your current usage on the local-upload step of the Add Dataset wizard, or fetch it from the API: 
curl https://api.avala.ai/api/v1/datasets/manual-upload/quota/ \
  -H "X-Avala-Api-Key: $AVALA_API_KEY"
# {"used": 1342177280, "limit": 10737418240}
If you need a higher cap, contact support@avala.ai.

​
Web wizard (drag-and-drop)

  1. Open Mission Control and click Add Dataset.
  2. Step 1 — Common data: name the dataset, pick the data type (image, video, lidar, splat, MCAP, etc.), and set labels.
  3. Step 2 — Source: pick Local files.
  4. Step 3 — Upload: drag files into the drop zone or click to browse. The picker accepts only file types that match the data type from step 1 — for example, a Lidar dataset accepts .pcd, .bin, .las, .laz, .ply and rejects everything else with an inline reason. Per-file progress shows on each row.
  5. Click Submit once every selected file has finished uploading. (The wizard keeps the button disabled while any file is still pending, uploading, or failed — partial batches would create a dataset with only some of the files you picked.) The dataset is created in your account immediately and is ready to label.
If a file fails (e.g. network blip), use the Retry button on the row. The wizard never re-uploads files that already succeeded.

​
CLI (avala datasets upload)

Install the SDK: 
pip install avala
Authenticate by setting an API key (create one in Settings → API Keys): 
export AVALA_API_KEY=avk_...
Then run the upload — the CLI creates the dataset and uploads all files in one shot: 
avala datasets upload \
  --source ./recordings/2026-04-26/ \
  --name "PCD recordings 2026-04-26" \
  --slug pcd-recordings-2026-04-26 \
  --data-type lidar
The CLI walks the source directory, opens parallel connections to S3, and shows a progress bar. If a single upload fails the command stops with a non-zero exit code; rerun the command to retry.

​
Common flags

FlagWhat it does
--source <path>Local file or directory containing files to upload. Required.
--name <name>Display name for the new dataset. Required.
--slug <slug>URL-friendly identifier for the dataset. Required.
--data-type <type>One of image, video, lidar, splat, mcap. Required.
--visibility <vis>private (default) or public.
--dry-runList files that would upload, with their sizes, and exit.
--workers NParallel uploads (default 8).

​
Quota errors

If your upload would push you over 10 GB, the API returns HTTP 413 and the CLI prints: 
ERROR: Storage quota exceeded.
Used: 9.7 GB / 10 GB.
This upload would add 600 MB.
Free up space at https://avala.ai/@<org>/datasets, or contact support to raise the cap.
The CLI stops before issuing any further presigned URLs, so partial uploads are bounded.

​
After the upload

Either path produces a normal dataset. From there you can:
  • Define a project over it and assign annotation tasks.
  • Trigger an export once labelling is done.
  • Browse items via client.datasets.get(<uid>).items() in the Python SDK.
Your files live in Avala-managed storage. Your dataset’s provider_config handles access, so you don’t need separate AWS credentials to view them.