# runpod-serverless

> Create serverless endpoint templates and endpoints on RunPod.io. Supports Python/Node.js runtimes, GPU selection (3090, A100, etc.), and idempotent configuration. Use this skill when a user wants to set up a new serverless endpoint or template on RunPod.

- Author: jarmen423
- Repository: jarmen423/skills
- Version: 20260125165532
- Stars: 2
- Forks: 0
- Last Updated: 2026-02-06
- Source: https://github.com/jarmen423/skills
- Web: https://mule.run/skillshub/@@jarmen423/skills~runpod-serverless:20260125165532

---

---
name: runpod-serverless
description: Create serverless endpoint templates and endpoints on RunPod.io. Supports Python/Node.js runtimes, GPU selection (3090, A100, etc.), and idempotent configuration. Use this skill when a user wants to set up a new serverless endpoint or template on RunPod.
---

# RunPod Serverless Creator

This skill helps you create and configure serverless endpoints on RunPod.io. It handles both the Template creation (software config) and the Endpoint creation (hardware config).

## How to Use

The primary tool is the `scripts/create_serverless.py` script. It is idempotent: if a template or endpoint with the same name exists, it will reuse the template and update the endpoint.

### Prerequisites

- **API Key**: You need a RunPod API Key.
  - Ask the user to provide it or check if `RUNPOD_API_KEY` is in the environment.
  - *Security*: Do not hardcode the key in the script. Pass it via environment variable `RUNPOD_API_KEY`.

### Command

```bash
python runpod-serverless/scripts/create_serverless.py \
  --name <NAME> \
  --runtime <python|node> \
  --gpu <GPU_TYPE> \
  [--disk <GB>] \
  [--min-workers <INT>] \
  [--max-workers <INT>]
```

### Arguments

- `--name` (Required): Unique name for the template and endpoint.
- `--runtime` (Required): `python` or `node`. Maps to standard RunPod serverless base images.
- `--gpu` (Required): GPU type alias (e.g., `3090`, `4090`, `a100`, `a6000`) or specific ID (e.g., `AMPERE_24`).
- `--disk`: Container disk size in GB (default: 10).
- `--min-workers`: Minimum active workers (default: 0 for cold-start serverless).
- `--max-workers`: Maximum active workers (default: 1).

### GPU Options (Aliases)

- `3090` -> `AMPERE_24` (24GB VRAM)
- `4090` -> `ADA_24` (24GB VRAM)
- `a4000` -> `AMPERE_16` (16GB VRAM)
- `a6000` -> `AMPERE_48` (48GB VRAM)
- `a100` -> `AMPERE_80` (80GB VRAM)

See `references/runpod_api.md` for full API details and ID mappings.

## Example Workflow

1. **Ask User** for:
   - Application Name
   - Runtime (Python/Node)
   - Preferred GPU
   - API Key (if not known)

2. **Run Script**:
   ```bash
   $env:RUNPOD_API_KEY="<USER_KEY>"; python runpod-serverless/scripts/create_serverless.py --name my-ai-api --runtime python --gpu 3090
   ```

3. **Verify**: The script outputs the Template ID and Endpoint ID upon success.