Skip to content
TAIP

Products / For AI builders

DevSpace

Available

Jupyter or VS Code on a GPU in seconds. Idle environments shut themselves down.

DevSpace turns a Kubernetes cluster into a self-service development platform for AI and data science teams. Users pick a template, pick hardware — whole GPUs or MIG slices — and get an isolated environment with persistent storage and the right libraries, mounted at whichever subdirectory of their workspace they pick from a file browser. Each environment carries its own oauth2-proxy sidecar, so only the owner and the colleagues they name can reach it, even with the URL. An activity probe watches the IDE and scales idle environments to zero, returning GPUs to the pool with state preserved. Power users reach the same environment over the platform SSH gateway: scp, port forwarding, and editor remoting included, with keys and tokens managed once in ConsoleX rather than per app.

All products

Specification

Version
v1.7.3 — generally available
IDEs
Jupyter · VS Code · Marimo · Streamlit · Gradio · Deepnote — extensible via CRD
Isolation
Per-environment oauth2-proxy sidecar · NetworkPolicy · named-user sharing
Idle
Activity-probe sidecar → scale-to-zero or delete
Accelerators
Whole GPUs and MIG slices, shown and quota-counted as distinct resources
Access
Platform SSH gateway · taip_ tokens from ConsoleX with devspace scopes
Languages
English · 简体中文 (UI and docs)

Proof, not promises

See it in one block.

No proprietary SDKs, no rewrites — DevSpace meets your tools where they already are.

real SSH, no kubeconfig 
# your laptop → a GPU environment, through the platform SSH gateway
$ ssh alice+env=jupyter@gateway.intra.example
(jupyter) $ nvidia-smi --query-gpu=name --format=csv,noheader
NVIDIA A100-SXM4-80GB
# idle for 2h → scaled to zero, PVC intact, GPU back in the pool▌

Native SSH UX — shell, scp, sftp, and port forwarding — authenticated by the key you registered once in ConsoleX. One gateway for every workspace, no kubectl anywhere.

Capabilities

What DevSpace gives you

01

Notebook, IDE, app — same plane

Jupyter, Marimo, Streamlit, Gradio, VS Code Server out of the box. Add any new IDE through an EnvironmentTemplate CRD — no code changes, no controller redeploy.

02

Isolated by default

Each environment runs as its own StatefulSet with an oauth2-proxy sidecar. Only the owner — and the colleagues they explicitly share with — can reach the IDE, even if someone else discovers the URL. NetworkPolicy locks the rest down, and an append-only audit log records every action. Sharing is per environment, not per namespace, and the UI says plainly what it grants: an IDE is code execution as you.

03

Idle shutdown that frees GPUs

An activity sidecar polls the IDE — reading Jupyter's real kernel activity, not just liveness. When it goes quiet, DevSpace scales to zero with state preserved, or deletes — operator's choice per template. Restart takes seconds, not a re-provision.

04

Web terminal and SSH, through one gateway

Open a shell in the browser, or reach the same pod with ssh user+env=name@gateway from your own terminal — plain scp and sftp, ssh -L port forwarding, and VS Code Remote-SSH all work. DevSpace no longer runs a bastion of its own: keys live once on the ConsoleX Profile page, and API tokens are the platform's taip_ tokens scoped to devspace.

05

Mount exactly the directory you meant

Pick the subdirectory of your workspace PVC to mount by browsing it, not by typing a path and finding out later. Extra PVCs mount alongside it, and a restart action cycles an environment deliberately — a share edit never kills the session you are working in.

How it works

From template to running IDE in seconds.

  1. Step 01

    Pick a template

    Jupyter, Marimo, Streamlit, Gradio, or VS Code — and any custom EnvironmentTemplate the platform team has authored.

  2. Step 02

    Get an isolated environment

    DevSpace boots a StatefulSet with your storage, libraries, and an oauth2-proxy sidecar. Only you can reach it.

  3. Step 03

    Idle = scale to zero

    When the IDE goes quiet, DevSpace stops it — state preserved — or deletes it. GPUs go back to the pool automatically.

Who it's for

Built for these teams

  • Data scientists prototyping on GPUs
  • Research teams sharing finite hardware
  • Platform teams done supporting one-off notebook servers

Pairs well with

Other builder products

ConsoleX

 Available

Log in, get a governed Kubernetes workspace. No kubectl, no tickets.

On first SSO login every user gets an isolated namespace with quotas, default-deny networking, storage, and a web terminal — provisioned automatically, reconciled continuously. It is also the platform's front door: the token authority every TAIP app trusts, the SSH gateway into every workspace, and the taip CLI the cluster serves itself.

Learn more

TrainX

 Available

A template is a shape. Your project fills it. Kubernetes runs the job.

Self-describing templates render straight into UI forms — or your project declares itself in a trainx.yaml and runs on a platform shape. Live quota checks, streaming logs, parsed progress, MLflow-tracked parameters and metrics, resumable runs, and declared outputs you can hand to the next step.

Learn more

ModelSphere

 Available

Your own Hugging Face Hub. Change one env var — every client just works.

A self-hosted, HF-compatible model and dataset registry: transformers, datasets, huggingface-cli, git-lfs, and the Xet protocol all work unchanged — plus a browsable Hub UI with model cards, a paging dataset viewer, file history, OIDC, audit, quotas, and pull-through caching of the public Hub.

Learn more