Rapids:
Instant, Auto-Scaling Containers

When your container receives no traffic for 5 minutes, Rapids automatically scales it down to zero running instances. You are not billed during this idle period. When a new request arrives, a new instance spins up in milliseconds and begins processing. This means you never pay for idle resources.

Rapids supports three deployment methods: (1) Docker Image — pull directly from Docker Hub, GitHub Container Registry, or any supported registry. (2) Git Repository — connect your GitHub, GitLab, or Bitbucket repo and deploy on every push. (3) ZIP Upload — upload your source code as a ZIP file from the dashboard or CLI.

Rapids supports any language that can run in a container. For source deployments (Git and ZIP), Cloud Native Buildpacks automatically detect and build Node.js, Python, Go, Java, Ruby, PHP, .NET, and Rust applications. You can also provide your own Dockerfile for full control.

Rapids uses pay-per-use billing based on three components: vCPU-seconds (the CPU time your container uses while active), memory GiB-seconds (the RAM allocated while active), and request count. A generous free tier includes 2 million requests, approximately 69 vCPU-hours, and 139 GiB-hours per month at no cost.

Yes. Every Rapids container gets a subdomain under serverless.danubedata.ro by default, and you can add up to 10 custom domains per container. TLS certificates are provisioned automatically via Let's Encrypt — no configuration required.

Connect your GitHub, GitLab, or Bitbucket repository and Rapids installs a webhook. Every time you push to your configured branch, Rapids automatically triggers a new build using Cloud Native Buildpacks or your Dockerfile, then deploys the new revision with zero downtime.

Rapids offers four resource profiles: Micro (0.25-0.5 vCPU, 128-256 MB RAM), Small (0.5-1 vCPU, 256-512 MB RAM), Medium (1-2 vCPU, 512 MB-1 GB RAM), and Large (2-4 vCPU, 1-2 GB RAM). Each container can scale up to 100 instances and handle up to 1,000 concurrent requests per instance.

All Rapids containers run in our EU datacenter in Falkenstein, Germany. Your source code, container images, and traffic never leave the European Union. We are fully GDPR compliant.

Yes — that is a design goal, not a side effect. Every CLI command accepts --json and returns one envelope: {success, data, error, meta}. Failures carry a machine-readable code and a retryable flag, so an agent branches on the cause instead of parsing English. The typical agentic workflow is preflight the image, apply with an idempotency key, wait for the operation to reach a terminal state, then probe the URL. No screen-scraping and no guesswork.

Every response carries status_details with an explicit operation.terminal boolean. That is the stop condition — never the status string, because inferring terminality from text is how automated loops hang forever. danube operations wait blocks until terminal and honours the server's poll interval, so pacing can change without updating your agent. A wait that times out reports operation.wait_timeout, which means the work has not finished, not that it failed.

Pass --idempotency-key to danube rapids apply. The same key replays the original result instead of creating a second container, so a network timeout followed by a retry cannot leave you with duplicate resources. Deployments are also addressable as operations, so an agent that lost its connection can reattach to the work already in flight rather than starting again.

Run danube rapids diagnose. It correlates status, revisions, events and logs into ranked findings with remediation. It will tell you outright when zero replicas alongside a failed Ready condition means no pod was ever scheduled — usually an image that could not be pulled — so the empty logs are expected and not worth hunting. danube rapids preflight catches that class of failure before you deploy at all, by resolving the digest and architecture and confirming the image can be pulled.

Rapids is approximately 25-50% cheaper than AWS Lambda and Google Cloud Run on vCPU-second pricing, and 40-70% cheaper on request pricing. Unlike Lambda, Rapids runs standard containers with no proprietary runtime constraints. Unlike Cloud Run, your data stays in the EU with native GDPR compliance.