Skip to content
Back to skills

Deployment And Interfaces

ASecurity

"Routes min(DALL·E) command-line, notebook, Tkinter GUI, and

  • 247 stars
  • 0 votes
  • 0 copies
  • 4 views
  • Added September 8, 2026
toolspythonbashapibackend

Works with

  • cli
  • api

Security analysis

A100/100

Pro scans all 6 files and shows the line behind each finding

Scanned September 8, 2026

npx -y skills add VectorSpaceLab/AREX-Skill --skill deployment-and-interfaces --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Deployment And Interfaces?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Deployment And Interfaces
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/vectorspacelab-deployment-and-interfaces/badge)](https://www.skillsdirectory.com/skills/vectorspacelab-deployment-and-interfaces)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: deployment-and-interfaces
description: "Routes min(DALL·E) command-line, notebook, Tkinter GUI, and
  Replicate/Cog interface workflows beyond direct MinDalle API calls."
disable-model-invocation: true
metadata:
  disco-role: operating
license: MIT
---

# deployment-and-interfaces

Use this sub-skill when the task is about min(DALL·E) public interfaces outside a direct `MinDalle` API call: command-line usage, Colab/notebook patterns, the Tkinter UI, Replicate/Cog deployment, predictor filenames, output paths, or interface-specific troubleshooting.

## Start here

- Read [references/cli-reference.md](references/cli-reference.md) for the original `image_from_text.py` flags, defaults, save behavior, ASCII preview, and the verified absence of an installed console command.
- Read [references/interface-workflows.md](references/interface-workflows.md) for the Colab/notebook flow, Tkinter controls and limits, Replicate predictor inputs/defaults/ranges, filename sanitation, output extensions, Cog pins, and deployment caveats.
- Read [references/troubleshooting.md](references/troubleshooting.md) when the user reports a missing command, slow downloads/inference, CPU/FP16 problems, display errors, Cog/CUDA issues, unexpected Replicate filenames, or progressive-output counts.

## Bundled helpers

- [scripts/min_dalle_cli_template.py](scripts/min_dalle_cli_template.py) is the safe replacement for the original command-line script. It defaults to `--dry-run`; add `--run` only when the user explicitly wants model construction, possible downloads, and inference.
- [scripts/replicate_filename_sanitize.py](scripts/replicate_filename_sanitize.py) previews the Replicate-style output basename without requiring Cog or model loading.

Useful no-network checks:

```bash
python scripts/min_dalle_cli_template.py --help
python scripts/min_dalle_cli_template.py --text "artificial intelligence" --no-mega --top-k 256
python scripts/replicate_filename_sanitize.py --self-test
python scripts/replicate_filename_sanitize.py --text "Dali painting of WALL·E"
```

## Boundary routes

- For direct `MinDalle` construction, `generate_image`, `generate_image_stream`, sampling arguments, tensor/PIL image semantics, and reusable model lifetime, route to [../text-to-image-generation/SKILL.md](../text-to-image-generation/SKILL.md).
- For model weight downloads, `models_root`, cache layout, device selection, dtype/backend behavior, CUDA/CPU trade-offs, and environment setup, route to [../model-assets-and-runtime/SKILL.md](../model-assets-and-runtime/SKILL.md).
- Do not tell future agents to run the original repository scripts or notebook. Use this sub-skill's bundled helpers and distilled references instead.

Files in this skill

  • SKILL.md2.7 KB
  • references/cli-reference.md3.8 KB
  • references/interface-workflows.md4.9 KB
  • references/troubleshooting.md5.3 KB
  • scripts/min_dalle_cli_template.py6.7 KB
  • scripts/replicate_filename_sanitize.py2.1 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.