Weight Prefetch

Weight prefetch streams model weights to Parel storage before a deployment exists. When you create the deployment later, cached weights can shorten cold start time.

Start prefetch

curl https://api.parel.cloud/v1/deployments/prefetch \
  -H "Authorization: Bearer pk-dev-YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"huggingface_id": "Qwen/Qwen2.5-7B-Instruct"}'

Poll status

curl https://api.parel.cloud/v1/deployments/prefetch/Qwen%2FQwen2.5-7B-Instruct \
  -H "Authorization: Bearer pk-dev-YOUR_KEY"

Cancel

curl -X POST https://api.parel.cloud/v1/deployments/prefetch/Qwen%2FQwen2.5-7B-Instruct/cancel \
  -H "Authorization: Bearer pk-dev-YOUR_KEY"

Constraints

Constraint	Value
Per-model size limit	50 GB
Unused prefetch TTL	10 minutes
Storage lifecycle	30 days