Unify customer data, automate workflows, and scale experiences with Einstein AI, Copilot, and Data Cloud in one CRM platform.
110M visits/mo+1.1%#1 Business & Marketing
JEV PLAYGROUND ARCADE
Interactive testing environment for Jev decision model with eight playable cabinets
About JEV PLAYGROUND ARCADE
JEV PLAYGROUND ARCADE provides a hands-on interface to evaluate TypeSafe AI’s Jev decision model through eight distinct game-like scenarios. Users supply application state in text, JSON, or string arrays and pose typed questions to receive calibrated decisions with confidence scores in approximately 70–500 milliseconds. The system handles four decision types: choosing one labeled option from multiple choices, scoring state on an ordered rubric, calculating calibrated probabilities for conditions, and confidence-gated routing that escalates uncertain decisions to higher-cost models. Each cabinet simulates a real Jev call using arcade tokens, allowing users to experiment with decision logic without deploying code. The arcade includes preset scenarios like ticket triage, sentiment analysis, refund detection, and resume grading, as well as a free-play sandbox for custom queries and raw JSON inspection. High scores track performance metrics such as latency and confidence across sessions.
Key features
- Typed question input with application state
- Choice selection with labeled options
- Rubric scoring with fractional results
- Calibrated probability calculations
- Confidence-gated routing to fallback models
- High score tracking for performance metrics
- Free-play sandbox for custom testing
- Eight preset game-like scenarios
Use cases
- Testing decision logic before production deployment
- Evaluating model performance across different scenarios
- Training teams on calibrated AI decision-making
Pros
- Supports multiple input formats (text, JSON, string arrays)
- Returns calibrated decisions with confidence scores
- Four distinct decision types (choice, scoring, probability, routing)
- Real-time performance metrics (latency, confidence)
- Preset scenarios for common use cases
Cons
- Requires arcade tokens for each decision call
- Limited to Jev decision model via Vercel AI Gateway
- No offline or local deployment option
- Single-player mode for most cabinets
Frequently asked questions about JEV PLAYGROUND ARCADE
What is JEV PLAYGROUND ARCADE?
JEV PLAYGROUND ARCADE is an interactive interface designed to evaluate TypeSafe AI’s Jev decision model through eight game-like scenarios. Users input application state and pose typed questions to receive calibrated decisions with confidence scores in approximately 70–500 milliseconds.
Who should use JEV PLAYGROUND ARCADE?
The tool is intended for developers, data scientists, and product teams who need to test and refine decision-making logic without deploying code. It is particularly useful for those working with Jev’s decision model for tasks like ticket triage, sentiment analysis, or resume grading.
How does JEV PLAYGROUND ARCADE work?
Users supply application state in text, JSON, or string arrays and ask typed questions. The system returns typed decisions with calibrated confidence scores, simulating real Jev calls. Each scenario in the arcade uses tokens to represent these calls, allowing experimentation without live deployment.
What types of decisions can JEV PLAYGROUND ARCADE handle?
The arcade supports four decision types: choosing one labeled option from multiple choices, scoring state on an ordered rubric, calculating calibrated probabilities for conditions, and confidence-gated routing that escalates uncertain decisions to higher-cost models.
Can I create custom scenarios in JEV PLAYGROUND ARCADE?
Yes, the FREE PLAY SANDBOX allows users to build any question and inspect the raw JSON output. This feature is ideal for testing custom queries or exploring the decision model’s behavior in detail.
What are the preset scenarios available in the arcade?
The arcade includes preset scenarios like ticket triage, sentiment analysis, refund detection, resume grading, spam detection, and confidence calibration. Each scenario is designed to simulate real-world use cases for the Jev decision model.
