Playgrounds

Playgrounds is a column-based testing workspace where you run multiple prompt configurations side-by-side with identical inputs and compare their outputs.

Each column holds a prompt, a model, and a set of settings; the editor labels them Version 1, Version 2, and so on. You run all columns simultaneously, then examine the results together. Columns are independent - changing one does not affect the others.

What you can compare

• Prompt versions - load the same prompt at different revisions in separate columns to see what changed
• Models - load the same prompt with different models (ChatGPT, Claude, Gemini) to evaluate outputs
• Settings - test different temperature or max token values against the same input
• Different prompts - compare entirely different prompt approaches on the same task

Isolation

Changes made inside a Playground do not affect your saved prompts. You can edit instructions, switch models, and adjust settings freely without modifying anything in your library.

Finding Playgrounds

Playgrounds lives under Tools in the left sidebar. Your saved playgrounds appear in a list — click any to open it, or use New Playground to create one.

The interface

Each version card shows:

• The prompt — a preview of the user message, with the model name (and prompt name, if loaded from your library) above it
• A Run button and a ⋯ menu (Clone the version, or Remove it)
• The Response area, where the output appears after a run

Click the prompt card to open that version's edit panel, where you set:

• Model — and its inference settings, including temperature and max tokens
• System Message — optional instructions for the model
• User Message — the prompt itself
• Variable overrides — per-version values for any {{variable}} in the prompt (see Variables)
• Attachments — files and memories for this version

Workspace controls run across the top of the editor:

• Add Version — add a new column to the right (up to seven)
• Run All — run every version at once with the current variables
• Variables — open the global variables panel
• Files — attach files to every version
• Save / Create — save the current state

Adding versions

Click Add Version and choose how to start the column:

• New Version — start from a blank prompt
• Load Saved — load a prompt from your library, then pick which revision to load

Once a version is loaded, you can edit it freely. Changes inside a playground never modify the saved prompt.

Setting up a comparison

Different prompt versions

1. Add two versions.
2. In each, Load Saved and pick the same prompt.
3. Load a different revision in each version.
4. Enter the same variable values (they are shared by default).
5. Click Run All.

Different models

1. Load your prompt into a version.
2. Clone it into additional versions from the version's ⋯ menu.
3. Change the model in each version's edit panel.
4. Enter variables and click Run All.

Different settings

1. Load the same prompt and model into multiple versions.
2. Adjust temperature or max tokens in each version's model settings.
3. Run all with identical variables.

Variables

Any {{variable_name}} in a prompt becomes a variable. Aisle detects them automatically and lists them in the Variables panel (the Variables button in the toolbar shows the count).

Values set in the Variables panel are the defaults for every version, so all columns run with the same input. To give one version a different value, open its edit panel and set a Variable override — this affects that version only, and you can reset it back to the global value at any time.

Outputs

After a run, each version shows its output in the Response area, with the time the run completed. Use the expand button to read a long output in a full-width panel, or the copy button to copy it. Long outputs scroll within the card.

Model settings

Open a version's edit panel and use the model selector to adjust:

Setting	Description
Temperature	Controls output variability. Lower values (0.0–0.3) produce more consistent outputs. Higher values (0.7–1.0) produce more varied outputs.
Max tokens	Maximum response length. Lower values truncate long responses.

Settings apply to that version only.

Files

Attach files in two places:

• Files in the toolbar attaches a file to every version.
• Attachments in a version's edit panel attaches files (and memories) to that version only.

Where to go next

See Saving and Sharing Playgrounds for how to preserve, name, and share a test session — including a worked example of comparing two prompts side by side.