Table of contents

Templates

Templates turn a full pod configuration — image, ports, environment, resources — into a one-click launch. Podstack offers two kinds:

  • 1-click templates (catalog) — pre-built, GPU-ready apps maintained by Podstack (ComfyUI, Ollama, vLLM, Jupyter, and more). You pick one inside the launch wizard.
  • Your saved templates — pod configurations you save yourself and reuse across your team. These live on the Templates page.

Using a 1-click template

The Podstack catalog is browsed inside the launch wizard, on the Choose a Template step:

  1. Go to Pods and click Launch Pod.
  2. Pick a GPU on the Instance Type step (see Launch a Pod).
  3. On the Choose a Template step, find your app. Filter with the Search templates by name, image, GPU… box or the All Categories dropdown. Templates are grouped into collapsible categories.
  4. Click the template card. Its image, ports, environment variables, and startup command are applied, and you jump to the Configure step.
  5. Adjust anything you like, then click Launch Pod.

Prefer to build from scratch? Click Custom Configuration at the top of the step instead.

The 1-click catalog

Every image is GPU-ready, expects persistent storage at /data, exposes SSH on port 22, and runs its app on its native port. Each has its own reference page under Container Images with ports and environment variables.

Image Generation

  • a1111 — Automatic1111 Stable Diffusion WebUI
  • sd-forge — SD WebUI Forge (faster A1111 fork)
  • comfyui — ComfyUI node-graph image/video generator
  • kohya-ss — Kohya LoRA / fine-tune trainer

LLM Training

LLM Serving & Inference

  • vllm — vLLM high-throughput inference server
  • sglang — SGLang structured-generation server
  • ollama — Ollama local LLM runner
  • text-gen-webui — oobabooga text-generation-webui
  • tei — Hugging Face Text Embeddings Inference
  • triton — NVIDIA Triton Inference Server

ML Frameworks & Notebooks

Speech & Audio

  • cloudtts — MOSS-TTS + Chatterbox TTS
  • whisperx — WhisperX speech-to-text + diarization

Science & HPC

  • alphafold — AlphaFold protein structure prediction
  • gromacs — GROMACS molecular dynamics
  • octave — GNU Octave numerical computing

Video, Graphics & Vision

Base

JupyterHub login. On images with JupyterHub, log in with any username and your Podstack API token (psk_...) as the password. Generate one at Account > API Tokens.

Save your own template

Any configuration you build in the launch wizard can be saved for reuse:

  1. In the launch wizard, get to the Configure step and set everything the way you want.
  2. In the Summary panel, click Save as Template.
  3. In the Save as Template dialog, enter a Template Name, an optional Description, and choose Save to Project (the template becomes available to that project’s members).
  4. Click Save Template.

Your saved templates then appear:

  • On the Templates page (My Templates), and
  • In the launch wizard under the User Templates category, tagged User template.

Managing saved templates

The Templates page (titled Launch Templates) lists your saved pod and VM configurations as cards. Filter by project with the All Projects dropdown. Each card shows the name, description, image, resource chips, and tags.

Per-template actions:

  • Launch Pod — opens the launch wizard pre-filled with this template, jumping straight to the Instance Type step.
  • Edit (pencil) — opens the Edit Template dialog. You can change the Docker Image, CPU Cores, Memory (GB), GPU Type, GPU Count, Exposed Ports, Environment Variables, Startup Command, and Tags. Click Save Changes.
  • Delete (trash) — removes the template after confirmation.

To create a template from scratch, click Create Template on the Templates page — it opens the launch wizard, where you configure and then Save as Template.

Editing or deleting a template does not affect pods already launched from it.

Next steps