What is a Playground View?
A playground view saves a complete snapshot of your current prompt playground session, allowing you to preserve your work, share configurations with teammates, or return to previous experiments.
Why Playground Views ?
Playground Views make prompt experimentation reproducible, shareable, and collaborative. Instead of losing progress or re-creating setups from scratch, you can save the exact state of your Playground and return to it anytime. With Playground Views, you can:- Preserve every detail — Save your full session state, including model configuration, parameters, inputs, and results, so experiments remain consistent and comparable.
- Revisit past experiments — Quickly reload prior sessions to review outputs, test improvements, or continue iterating without losing context.
- Collaborate effortlessly — Share saved views with teammates so everyone can see, reproduce, and refine the same experiment setup.
- Track prompt evolution — Keep snapshots of key milestones as you optimize prompts over time, making it easy to see what changed and why.
- Ensure reproducibility — Use saved views to validate results, reproduce runs, or debug changes in model behavior.
- Work confidently — Auto-save keeps your latest progress safe, while manual view saving lets you version experiments intentionally.
Creating a New View
You can create a new view at any time by saving your current playground progress. Simply click the Save as New View button and give your view a descriptive name.What Gets Saved in a View
When you save a playground view, the following elements are preserved:LLM Configuration
- Provider (OpenAI, Azure OpenAI, Vertex AI, AWS Bedrock, etc.)
- Specific model selection
- All model parameters (temperature, max_tokens, top_p, etc.)
- Custom model endpoint settings (if applicable)
Prompt Setup
- Complete conversation history with all message roles (system, user, assistant, tool)
- Message content and any attached images
- Tool calls and function calling configurations
Generated Results (when working with datasets)
- All responses generated during your session
- Performance metrics (latency, token counts)
- Evaluation scores and feedback (if evaluators are applied)
Data Context
- Connected data source (dataset or span)
Auto-Save Feature
Playground views automatically save your work as you make changes, so you never lose progress. A small visual indicator will show when your changes are being saved. If you prefer, you can turn off auto-save at any time in Settings. When auto-save is disabled, you’ll need to save your prompt manually — either by selecting Save from the dropdown menu or by pressing ⌘+S
Managing Your Views
- Reload a view: Access any saved view from the Views menu to instantly restore that exact configuration
- Duplicate: Allows you to duplicate your current view to a new one.
- Exit a view: Return to a clean, stateless playground at any time by clicking “Exit View”